updated series_index to float and added UniqueConstraint to audiobook and audiobook_author

this code was moved to https://gitea.tmmworkshop.com/Nornsight/weave

.github/workflows/build_systems.yml

@@ -23,6 +23,6 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - name: Build default package
-        run: "nixos-rebuild build --flake ./#${{ matrix.system }}"
+        run: "nixos-rebuild build --accept-flake-config --flake ./#${{ matrix.system }}"
       - name: copy to nix-cache
         run: nix copy --accept-flake-config --to unix:///host-nix/var/nix/daemon-socket/socket .#nixosConfigurations.${{ matrix.system }}.config.system.build.toplevel

.github/workflows/fix_eval_warnings.yml

@@ -1,30 +0,0 @@
-name: fix_eval_warnings
-on:
-  workflow_run:
-    workflows: ["build_systems"]
-    types: [completed]
-
-jobs:
-  check-warnings:
-    if: >-
-      github.event.workflow_run.conclusion != 'cancelled' &&
-      github.event.workflow_run.head_branch == 'main' &&
-      (github.event.workflow_run.event == 'push' || github.event.workflow_run.event == 'schedule')
-    runs-on: self-hosted
-    permissions:
-      contents: write
-      pull-requests: write
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Fix eval warnings
-        env:
-          GH_TOKEN: ${{ secrets.GH_TOKEN_FOR_UPDATES }}
-        run: >-
-          nix develop .#devShells.x86_64-linux.default -c
-          python -m python.eval_warnings.main
-          --run-id "${{ github.event.workflow_run.id }}"
-          --repo "${{ github.repository }}"
-          --ollama-url "${{ secrets.OLLAMA_URL }}"
-          --run-url "${{ github.event.workflow_run.html_url }}"

.github/workflows/merge_flake_lock_update.yml

@@ -6,24 +6,18 @@ on:
 
 jobs:
   merge:
-    runs-on: ubuntu-latest
+    runs-on: self-hosted
 
     permissions:
       contents: write
       pull-requests: write
 
     steps:
-      - name: Checkout repository
-        uses: actions/checkout@v4
-
       - name: merge_flake_lock_update
-        run: |
-          pr_number=$(gh pr list --state open --author RichieCahill --label flake_lock_update --json number --jq '.[0].number')
-          echo "pr_number=$pr_number" >> $GITHUB_ENV
-          if [ -n "$pr_number" ]; then
-            gh pr merge "$pr_number" --rebase
-          else
-            echo "No open PR found with label flake_lock_update"
-          fi
+        run: >-
+          nix develop .#devShells.x86_64-linux.default -c
+          python -m python.gitea_flake_lock merge
+          --repo "${{ github.repository }}"
         env:
-          GITHUB_TOKEN: ${{ secrets.GH_TOKEN_FOR_UPDATES }}
+          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
+          GITEA_URL: https://gitea.tmmworkshop.com

.github/workflows/pytest.yml

@@ -1,13 +1,13 @@
 name: pytest
 
 on:
+  workflow_dispatch:
   push:
     branches:
       - main
   pull_request:
     branches:
       - main
-  merge_group:
 
 jobs:
   pytest:

.github/workflows/update-flake-lock.yml

@@ -6,18 +6,21 @@ on:
 
 jobs:
   lockfile:
-    runs-on: ubuntu-latest
+    runs-on: self-hosted
+    permissions:
+      actions: write
+      contents: write
+      pull-requests: write
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
-      - name: Install Nix
-        uses: DeterminateSystems/nix-installer-action@main
       - name: Update flake.lock
-        uses: DeterminateSystems/update-flake-lock@main
-        with:
-          token: ${{ secrets.GH_TOKEN_FOR_UPDATES }}
-          pr-title: "Update flake.lock"
-          pr-labels: |
-            dependencies
-            automated
-            flake_lock_update
+        run: nix flake update
+      - name: Create or update flake.lock PR
+        env:
+          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
+          GITEA_URL: https://gitea.tmmworkshop.com
+        run: >-
+          nix develop .#devShells.x86_64-linux.default -c
+          python -m python.gitea_flake_lock update
+          --repo "${{ github.repository }}"

.gitignore

@@ -169,3 +169,6 @@ test.*
 # Frontend build output
 frontend/dist/
 frontend/node_modules/
+
+# data from testing llms
+data/*

.vscode/settings.json

@@ -40,7 +40,6 @@
     "cgroupdriver",
     "charliermarsh",
     "Checkpointing",
-    "cloudflared",
     "codellama",
     "codezombiech",
     "compactmode",
@@ -204,6 +203,7 @@
     "peerconnection",
     "PESKYFOX",
     "PGID",
+    "pgvector",
     "pipewire",
     "pkgs",
     "plugdev",
@@ -308,6 +308,7 @@
     "usernamehw",
     "userprefs",
     "vaninventory",
+    "vdev",
     "vfat",
     "victron",
     "virt",

AGENTS.md

@@ -1,12 +0,0 @@
-## Dev environment tips
-
-- use treefmt to format all files
-- make python code ruff compliant
-- use pytest to test python code
-- always use the minimum amount of complexity
-- if judgment calls are easy to reverse make them. if not ask me first
-- Match existing code style.
-- Use builtin helpers getenv() over os.environ.get.
-- Prefer single-purpose functions over “do everything” helpers.
-- Avoid compatibility branches like PG_USER and POSTGRESQL_URL unless requested.
-- Keep helpers only if reused or they simplify the code otherwise inline.

common/global/default.nix

@@ -23,7 +23,10 @@
   boot = {
     tmp.useTmpfs = true;
     kernelPackages = lib.mkDefault pkgs.linuxPackages_6_12;
-    zfs.package = lib.mkDefault pkgs.zfs_2_4;
+    zfs = {
+      package = lib.mkDefault pkgs.zfs_2_4;
+      forceImportRoot = lib.mkDefault false;
+    };
   };
 
   hardware.enableRedistributableFirmware = true;
@@ -37,10 +40,17 @@
 
   nixpkgs = {
     overlays = builtins.attrValues outputs.overlays;
-    config.allowUnfree = true;
+    config = {
+      allowUnfree = true;
+      permittedInsecurePackages = [
+        "openssl-1.1.1w" # This is for discord-canary
+      ];
+    };
   };
 
   services = {
+    dbus.implementation = "dbus";
+
     # firmware update
     fwupd.enable = true;
 

common/global/nix.nix

@@ -34,6 +34,7 @@ in
       warn-dirty = false;
       flake-registry = ""; # disable global flake registries
       connect-timeout = 10;
+      download-buffer-size = 536870912;
       fallback = true;
     };
 

common/optional/monitoring-agent.nix

@@ -0,0 +1,256 @@
+{
+  config,
+  lib,
+  pkgs,
+  ...
+}:
+let
+  monitoringInterface = "ztwfunumly";
+  nodeTextfileDir = "/var/lib/prometheus-node-exporter-textfile";
+
+  mkProcessNameTemplate =
+    perPid: template: if perPid then "${template}:{{.PID}}:{{.StartTime}}" else template;
+
+  mkProcessMatchers = perPid: [
+    {
+      name = mkProcessNameTemplate perPid "{{.Username}}:{{.Matches.Module}}";
+      cmdline = [ "^/nix/store[^ ]*/bin/python[^ ]* -m (?P<Module>[^ ]+)" ];
+    }
+    {
+      name = mkProcessNameTemplate perPid "{{.Username}}:{{.Matches.Wrapped}}";
+      cmdline = [
+        "^/nix/store[^ ]*/bin/python[^ ]* /nix/store[^ ]*/bin/\\.?(?P<Wrapped>[^ /]+?)(?:-wrapped)?(?:\\s|$)"
+      ];
+    }
+    {
+      name = mkProcessNameTemplate perPid "{{.Username}}:{{.Matches.Wrapped}}";
+      cmdline = [
+        "^/nix/store[^ ]*/bin/node /nix/store[^ ]*-(?P<Wrapped>[A-Za-z0-9._+-]+)-[0-9][^ /]*/"
+      ];
+    }
+    {
+      name = mkProcessNameTemplate perPid "{{.Username}}:{{.Matches.Wrapped}}";
+      cmdline = [ "^/nix/store[^ ]*/(?:bin/|lib/[^ ]*/)?\\.?(?P<Wrapped>[^ /]+?)(?:-wrapped)?(?:\\s|$)" ];
+    }
+    {
+      name = mkProcessNameTemplate perPid "{{.Username}}:{{.ExeBase}}";
+      cmdline = [ ".+" ];
+    }
+  ];
+
+  perPidConfig = pkgs.writeText "process-exporter-per-pid.yaml" (
+    builtins.toJSON {
+      process_names = mkProcessMatchers true;
+    }
+  );
+
+  zpoolLatencyScript = pkgs.writeShellScript "zpool-latency-exporter" ''
+        set -euo pipefail
+
+        out_dir=${lib.escapeShellArg nodeTextfileDir}
+        host=${lib.escapeShellArg config.networking.hostName}
+        tmp_file="$(mktemp "$out_dir/zpool.prom.XXXXXX")"
+        trap 'rm -f "$tmp_file"' EXIT
+
+        pools="$(zpool list -H -o name | paste -sd, -)"
+
+        cat >"$tmp_file" <<'EOF'
+    # HELP zpool_iostat_total_wait_read_ns Average total read wait time reported by zpool iostat.
+    # TYPE zpool_iostat_total_wait_read_ns gauge
+    # HELP zpool_iostat_total_wait_write_ns Average total write wait time reported by zpool iostat.
+    # TYPE zpool_iostat_total_wait_write_ns gauge
+    # HELP zpool_iostat_disk_wait_read_ns Average disk read wait time reported by zpool iostat.
+    # TYPE zpool_iostat_disk_wait_read_ns gauge
+    # HELP zpool_iostat_disk_wait_write_ns Average disk write wait time reported by zpool iostat.
+    # TYPE zpool_iostat_disk_wait_write_ns gauge
+    # HELP zpool_iostat_syncq_wait_read_ns Average synchronous queue read wait time reported by zpool iostat.
+    # TYPE zpool_iostat_syncq_wait_read_ns gauge
+    # HELP zpool_iostat_syncq_wait_write_ns Average synchronous queue write wait time reported by zpool iostat.
+    # TYPE zpool_iostat_syncq_wait_write_ns gauge
+    # HELP zpool_iostat_asyncq_wait_read_ns Average asynchronous queue read wait time reported by zpool iostat.
+    # TYPE zpool_iostat_asyncq_wait_read_ns gauge
+    # HELP zpool_iostat_asyncq_wait_write_ns Average asynchronous queue write wait time reported by zpool iostat.
+    # TYPE zpool_iostat_asyncq_wait_write_ns gauge
+    EOF
+
+        zpool iostat -Hplvy -y 1 1 | awk -F '\t' -v host="$host" -v pools="$pools" '
+          function esc(str, out) {
+            out = str
+            gsub(/\\/, "\\\\", out)
+            gsub(/"/, "\\\"", out)
+            return out
+          }
+
+          function emit(metric, pool, vdev, value) {
+            if (value == "" || value == "-") {
+              return
+            }
+
+            printf "%s{host=\"%s\",pool=\"%s\",vdev=\"%s\"} %s\n",
+              metric,
+              esc(host),
+              esc(pool),
+              esc(vdev),
+              value
+          }
+
+          BEGIN {
+            split(pools, pool_names, ",")
+            for (idx in pool_names) {
+              if (pool_names[idx] != "") {
+                known_pools[pool_names[idx]] = 1
+              }
+            }
+          }
+
+          NF == 0 {
+            next
+          }
+
+          {
+            row_name = $1
+
+            if (row_name in known_pools) {
+              current_pool = row_name
+              current_vdev = "_pool"
+            } else if (current_pool == "") {
+              next
+            } else {
+              current_vdev = row_name
+            }
+
+            emit("zpool_iostat_total_wait_read_ns", current_pool, current_vdev, $8)
+            emit("zpool_iostat_total_wait_write_ns", current_pool, current_vdev, $9)
+            emit("zpool_iostat_disk_wait_read_ns", current_pool, current_vdev, $10)
+            emit("zpool_iostat_disk_wait_write_ns", current_pool, current_vdev, $11)
+            emit("zpool_iostat_syncq_wait_read_ns", current_pool, current_vdev, $12)
+            emit("zpool_iostat_syncq_wait_write_ns", current_pool, current_vdev, $13)
+            emit("zpool_iostat_asyncq_wait_read_ns", current_pool, current_vdev, $14)
+            emit("zpool_iostat_asyncq_wait_write_ns", current_pool, current_vdev, $15)
+          }
+        ' >>"$tmp_file"
+
+        mv "$tmp_file" "$out_dir/zpool.prom"
+        trap - EXIT
+  '';
+in
+{
+  networking.firewall.interfaces.${monitoringInterface}.allowedTCPPorts = [
+    9100
+    9134
+    9256
+    9257
+    9633
+  ];
+
+  services.prometheus.exporters = {
+    node = {
+      enable = true;
+      enabledCollectors = [
+        "pressure"
+        "processes"
+        "systemd"
+      ];
+      extraFlags = [ "--collector.textfile.directory=${nodeTextfileDir}" ];
+    };
+
+    process = {
+      enable = true;
+      user = "root";
+      group = "root";
+      settings.process_names = mkProcessMatchers false;
+      extraFlags = [
+        "-gather-smaps=false"
+        "-remove-empty-groups=true"
+        "-threads=false"
+      ];
+    };
+
+    smartctl.enable = true;
+    zfs.enable = true;
+  };
+
+  programs.atop = {
+    enable = true;
+    atopService.enable = true;
+    atopRotateTimer.enable = true;
+    atopacctService.enable = true;
+    settings.interval = 30;
+  };
+
+  systemd = {
+    services = {
+      prometheus-process-pid-exporter = {
+        description = "Prometheus process exporter with per-PID naming";
+        wantedBy = [ "multi-user.target" ];
+        after = [ "network.target" ];
+        serviceConfig = {
+          ExecStart = ''
+            ${pkgs.prometheus-process-exporter}/bin/process-exporter \
+              --web.listen-address 0.0.0.0:9257 \
+              --config.path ${perPidConfig} \
+              -children=false \
+              -gather-smaps=false \
+              -remove-empty-groups=true \
+              -threads=false
+          '';
+          User = "root";
+          Group = "root";
+          Restart = "always";
+          WorkingDirectory = "/tmp";
+          CapabilityBoundingSet = [ "" ];
+          DeviceAllow = [ "" ];
+          LockPersonality = true;
+          MemoryDenyWriteExecute = true;
+          NoNewPrivileges = true;
+          PrivateDevices = true;
+          PrivateTmp = true;
+          ProtectClock = true;
+          ProtectControlGroups = true;
+          ProtectHome = true;
+          ProtectHostname = true;
+          ProtectKernelLogs = true;
+          ProtectKernelModules = true;
+          ProtectKernelTunables = true;
+          ProtectSystem = "strict";
+          RemoveIPC = true;
+          RestrictAddressFamilies = [
+            "AF_INET"
+            "AF_INET6"
+          ];
+          RestrictNamespaces = true;
+          RestrictRealtime = true;
+          RestrictSUIDSGID = true;
+          SystemCallArchitectures = "native";
+          UMask = "0077";
+        };
+      };
+
+      zpool-latency-exporter = {
+        description = "Exports ZFS latency metrics for node_exporter textfile collection";
+        after = [ "zfs-import.target" ];
+        requires = [ "zfs-import.target" ];
+        path = [
+          config.boot.zfs.package
+          pkgs.coreutils
+          pkgs.gawk
+        ];
+        serviceConfig = {
+          Type = "oneshot";
+          ExecStart = zpoolLatencyScript;
+        };
+      };
+    };
+
+    timers.zpool-latency-exporter = {
+      wantedBy = [ "timers.target" ];
+      timerConfig = {
+        OnBootSec = "2m";
+        OnUnitActiveSec = "60s";
+        Unit = "zpool-latency-exporter.service";
+      };
+    };
+
+    tmpfiles.rules = [ "d ${nodeTextfileDir} 0755 root root - -" ];
+  };
+}

common/optional/syncthing_base.nix

@@ -12,7 +12,7 @@
       brain.id = "SSCGIPI-IV3VYKB-TRNIJE3-COV4T2H-CDBER7F-I2CGHYA-NWOEUDU-3T5QAAN"; # cspell:disable-line
       ipad.id = "KI76T3X-SFUGV2L-VSNYTKR-TSIUV5L-SHWD3HE-GQRGRCN-GY4UFMD-CW6Z6AX"; # cspell:disable-line
       jeeves.id = "ICRHXZW-ECYJCUZ-I4CZ64R-3XRK7CG-LL2HAAK-FGOHD22-BQA4AI6-5OAL6AG"; # cspell:disable-line
-      phone.id = "TBRULKD-7DZPGGZ-F6LLB7J-MSO54AY-7KLPBIN-QOFK6PX-W2HBEWI-PHM2CQI"; # cspell:disable-line
+      phone.id = "JPVQKQW-CFXOJXT-Q5G5F3H-QIDHDRE-GKHPTQB-GXZUQSP-U7FR7F7-INP3AAH"; # cspell:disable-line
       rhapsody-in-green.id = "ASL3KC4-3XEN6PA-7BQBRKE-A7JXLI6-DJT43BY-Q4WPOER-7UALUAZ-VTPQ6Q4"; # cspell:disable-line
     };
   };

common/optional/update.nix

@@ -4,7 +4,7 @@
     flags = [ "--accept-flake-config" ];
     randomizedDelaySec = "1h";
     persistent = true;
-    flake = "github:RichieCahill/dotfiles";
+    flake = "git+https://gitea.tmmworkshop.com/richie/dotfiles?ref=main";
     allowReboot = true;
     dates = "Sat *-*-* 06:00:00";
   };

docs/zfs_failed_root_import_recovry.md

@@ -0,0 +1,76 @@
+# ZFS failed root import recovery
+
+## Fast path
+
+If the machine fails to boot because ZFS refuses to import `root_pool`:
+
+### GRUB
+
+1. At the bootloader menu, select the normal NixOS entry.
+2. Press `e`.
+3. Find the line that starts with `linux`.
+4. Append this to the end of that line:
+
+```text
+zfs_force=1
+```
+
+5. Boot once with `Ctrl+x` or `F10`.
+
+### systemd-boot
+
+1. At the bootloader menu, highlight the normal NixOS entry.
+2. Press `e`.
+3. Append this to the end of the options line:
+
+```text
+zfs_force=1
+```
+
+4. Press `Enter` to boot once.
+
+## After boot
+
+Run:
+
+```bash
+sudo zpool status
+sudo zpool import
+journalctl -b | rg "ZFS|zfs|import|root_pool"
+```
+
+## Expected result
+
+`sudo zpool status` should show `root_pool` as `ONLINE`.
+
+## Reboot test
+
+Run:
+
+```bash
+sudo reboot
+```
+
+Do not add `zfs_force=1` the second time.
+
+## If it still fails
+
+Boot once more with:
+
+```text
+zfs_force=1
+```
+
+Then run:
+
+```bash
+sudo zpool status -v
+sudo zpool history | tail -n 50
+journalctl -b | rg "ZFS|zfs|import|root_pool"
+```
+
+## Notes
+
+- Root pool name is `root_pool`.
+- This is a one-time recovery path after disk moves, controller changes, dirty exports, or interrupted imports.
+- Some hosts also need the LUKS unlock USB key inserted before boot.

flake.lock

@@ -8,11 +8,11 @@
       },
       "locked": {
         "dir": "pkgs/firefox-addons",
-        "lastModified": 1773979456,
-        "narHash": "sha256-9kBMJ5IvxqNlkkj/swmE8uK1Sc7TL/LIRUI958m7uBM=",
+        "lastModified": 1781150628,
+        "narHash": "sha256-b4mp8l3qWuSCyYYo9HSngDtcB3PpecYiOXjULrjwwlw=",
         "owner": "rycee",
         "repo": "nur-expressions",
-        "rev": "81e28f47ac18d9e89513929c77e711e657b64851",
+        "rev": "753319310f4673a2dabbfab87482187b40bf9bac",
         "type": "gitlab"
       },
       "original": {
@@ -29,11 +29,11 @@
         ]
       },
       "locked": {
-        "lastModified": 1774007980,
-        "narHash": "sha256-FOnZjElEI8pqqCvB6K/1JRHTE8o4rer8driivTpq2uo=",
+        "lastModified": 1781189114,
+        "narHash": "sha256-5inaamLgUMWy+MOBE9ChF9QAF1o/74LFuHkI0W/9rqc=",
         "owner": "nix-community",
         "repo": "home-manager",
-        "rev": "9670de2921812bc4e0452f6e3efd8c859696c183",
+        "rev": "486595d2cf49cfcd649b58a284fa11ac0e34da22",
         "type": "github"
       },
       "original": {
@@ -43,12 +43,15 @@
       }
     },
     "nixos-hardware": {
+      "inputs": {
+        "nixpkgs": "nixpkgs"
+      },
       "locked": {
-        "lastModified": 1774018263,
-        "narHash": "sha256-HHYEwK1A22aSaxv2ibhMMkKvrDGKGlA/qObG4smrSqc=",
+        "lastModified": 1781168557,
+        "narHash": "sha256-LOnLQ2tpYF9gqIDDr3+j3DbpJJr/QCH6zPRT2GzEUOE=",
         "owner": "nixos",
         "repo": "nixos-hardware",
-        "rev": "2d4b4717b2534fad5c715968c1cece04a172b365",
+        "rev": "6358ff76821101c178e3ab4919a62799bfe3652e",
         "type": "github"
       },
       "original": {
@@ -60,27 +63,24 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1773821835,
-        "narHash": "sha256-TJ3lSQtW0E2JrznGVm8hOQGVpXjJyXY2guAxku2O9A4=",
-        "owner": "nixos",
-        "repo": "nixpkgs",
-        "rev": "b40629efe5d6ec48dd1efba650c797ddbd39ace0",
-        "type": "github"
+        "lastModified": 1767892417,
+        "narHash": "sha256-8bW3q88CEg2u4hSP66Vf4lpbLonHz7hqDNBMcCY7E9U=",
+        "rev": "3497aa5c9457a9d88d71fa93a4a8368816fbeeba",
+        "type": "tarball",
+        "url": "https://releases.nixos.org/nixos/unstable/nixos-26.05pre924538.3497aa5c9457/nixexprs.tar.xz"
       },
       "original": {
-        "owner": "nixos",
-        "ref": "nixos-unstable",
-        "repo": "nixpkgs",
-        "type": "github"
+        "type": "tarball",
+        "url": "https://channels.nixos.org/nixos-unstable/nixexprs.tar.xz"
       }
     },
     "nixpkgs-master": {
       "locked": {
-        "lastModified": 1774051532,
-        "narHash": "sha256-d3CGMweyYIcPuTj5BKq+1Lx4zwlgL31nVtN647tOZKo=",
+        "lastModified": 1781229721,
+        "narHash": "sha256-ORvqDbb/LYxiJljGIejapjkc/kJbVote2N1WSb9W45I=",
         "owner": "nixos",
         "repo": "nixpkgs",
-        "rev": "8620c0b5cc8fbe76502442181be1d0514bc3a1b7",
+        "rev": "173d0ad7a974f8543a9ab01d2271b2e290341b33",
         "type": "github"
       },
       "original": {
@@ -106,12 +106,28 @@
         "type": "github"
       }
     },
+    "nixpkgs_2": {
+      "locked": {
+        "lastModified": 1781074563,
+        "narHash": "sha256-md8WlXOlfnIeHeOScMTTHFyf2d6iaTwPl2apR5EQ3P4=",
+        "owner": "nixos",
+        "repo": "nixpkgs",
+        "rev": "9ae611a455b90cf061d8f332b977e387bda8e1ca",
+        "type": "github"
+      },
+      "original": {
+        "owner": "nixos",
+        "ref": "nixos-unstable",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
     "root": {
       "inputs": {
         "firefox-addons": "firefox-addons",
         "home-manager": "home-manager",
         "nixos-hardware": "nixos-hardware",
-        "nixpkgs": "nixpkgs",
+        "nixpkgs": "nixpkgs_2",
         "nixpkgs-master": "nixpkgs-master",
         "nixpkgs-stable": "nixpkgs-stable",
         "sops-nix": "sops-nix",
@@ -125,11 +141,11 @@
         ]
       },
       "locked": {
-        "lastModified": 1773889674,
-        "narHash": "sha256-+ycaiVAk3MEshJTg35cBTUa0MizGiS+bgpYw/f8ohkg=",
+        "lastModified": 1780547341,
+        "narHash": "sha256-Gq8KNx5A7hBB3uGJaj6eQfLDIz5YdLu92gqBcvHvoUo=",
         "owner": "Mic92",
         "repo": "sops-nix",
-        "rev": "29b6519f3e0780452bca0ac0be4584f04ac16cc5",
+        "rev": "9ed65852b6257fbeae4355bc24ecfea307ca759a",
         "type": "github"
       },
       "original": {

frontend/.gitignore

@@ -1,24 +0,0 @@
-# Logs
-logs
-*.log
-npm-debug.log*
-yarn-debug.log*
-yarn-error.log*
-pnpm-debug.log*
-lerna-debug.log*
-
-node_modules
-dist
-dist-ssr
-*.local
-
-# Editor directories and files
-.vscode/*
-!.vscode/extensions.json
-.idea
-.DS_Store
-*.suo
-*.ntvs*
-*.njsproj
-*.sln
-*.sw?

overlays/default.nix

@@ -21,7 +21,6 @@
         alembic
         apprise
         apscheduler
-        confluent-kafka
         fastapi
         fastapi-cli
         httpx
@@ -42,6 +41,7 @@
         sqlalchemy
         tenacity
         textual
+        tiktoken
         tinytuya
         typer
         websockets

pyproject.toml

@@ -26,6 +26,7 @@ dependencies = [
 [project.scripts]
 database = "python.database_cli:app"
 van-inventory = "python.van_inventory.main:serve"
+whisper-transcribe = "python.tools.whisper.transcribe:main"
 
 [dependency-groups]
 dev = [
@@ -50,6 +51,7 @@ lint.ignore = [
     "COM812", # (TEMP) conflicts when used with the formatter
     "ISC001", # (TEMP) conflicts when used with the formatter
     "S603",   # (PERM) This is known to cause a false positive
+    "S607",   # (PERM) This is becoming a consistent annoyance
 ]
 
 [tool.ruff.lint.per-file-ignores]
@@ -78,15 +80,10 @@ lint.ignore = [
 "python/congress_tracker/**" = [
     "TC003", # (perm) this creates issues because sqlalchemy uses these at runtime
 ]
-"python/eval_warnings/**" = [
-    "S607", # (perm) gh and git are expected on PATH in the runner environment
-]
+
 "python/alembic/**" = [
     "INP001", # (perm) this creates LSP issues for alembic
 ]
-"python/signal_bot/**" = [
-    "D107", # (perm) class docstrings cover __init__
-]
 
 [tool.ruff.lint.pydocstyle]
 convention = "google"

python/alembic/data_science_dev/versions/2026_03_24-adding_failedingestion_2f43120e3ffc.py

@@ -1,50 +0,0 @@
-"""adding FailedIngestion.
-
-Revision ID: 2f43120e3ffc
-Revises: f99be864fe69
-Create Date: 2026-03-24 23:46:17.277897
-
-"""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-import sqlalchemy as sa
-from alembic import op
-
-from python.orm import DataScienceDevBase
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-
-# revision identifiers, used by Alembic.
-revision: str = "2f43120e3ffc"
-down_revision: str | None = "f99be864fe69"
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-schema = DataScienceDevBase.schema_name
-
-
-def upgrade() -> None:
-    """Upgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.create_table(
-        "failed_ingestion",
-        sa.Column("raw_line", sa.Text(), nullable=False),
-        sa.Column("error", sa.Text(), nullable=False),
-        sa.Column("id", sa.Integer(), nullable=False),
-        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.PrimaryKeyConstraint("id", name=op.f("pk_failed_ingestion")),
-        schema=schema,
-    )
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    """Downgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_table("failed_ingestion", schema=schema)
-    # ### end Alembic commands ###

python/alembic/data_science_dev/versions/2026_03_25-attach_partitions_to_posts.py

@@ -1,80 +0,0 @@
-"""Attach all partition tables to the posts parent table.
-
-Alembic autogenerate creates partition tables as standalone tables but does not
-emit the ALTER TABLE ... ATTACH PARTITION statements needed for PostgreSQL to
-route inserts to the correct partition.
-
-Revision ID: a1b2c3d4e5f6
-Revises: 605b1794838f
-Create Date: 2026-03-25 10:00:00.000000
-
-"""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-from alembic import op
-from sqlalchemy import text
-
-from python.orm import DataScienceDevBase
-from python.orm.data_science_dev.posts.partitions import (
-    PARTITION_END_YEAR,
-    PARTITION_START_YEAR,
-    iso_weeks_in_year,
-    week_bounds,
-)
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-
-# revision identifiers, used by Alembic.
-revision: str = "a1b2c3d4e5f6"
-down_revision: str | None = "605b1794838f"
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-schema = DataScienceDevBase.schema_name
-
-ALREADY_ATTACHED_QUERY = text("""
-    SELECT inhrelid::regclass::text
-    FROM pg_inherits
-    WHERE inhparent = :parent::regclass
-""")
-
-
-def upgrade() -> None:
-    """Attach all weekly partition tables to the posts parent table."""
-    connection = op.get_bind()
-    already_attached = {
-        row[0]
-        for row in connection.execute(
-            ALREADY_ATTACHED_QUERY, {"parent": f"{schema}.posts"}
-        )
-    }
-
-    for year in range(PARTITION_START_YEAR, PARTITION_END_YEAR + 1):
-        for week in range(1, iso_weeks_in_year(year) + 1):
-            table_name = f"posts_{year}_{week:02d}"
-            qualified_name = f"{schema}.{table_name}"
-            if qualified_name in already_attached:
-                continue
-            start, end = week_bounds(year, week)
-            start_str = start.strftime("%Y-%m-%d %H:%M:%S")
-            end_str = end.strftime("%Y-%m-%d %H:%M:%S")
-            op.execute(
-                f"ALTER TABLE {schema}.posts "
-                f"ATTACH PARTITION {qualified_name} "
-                f"FOR VALUES FROM ('{start_str}') TO ('{end_str}')"
-            )
-
-
-def downgrade() -> None:
-    """Detach all weekly partition tables from the posts parent table."""
-    for year in range(PARTITION_START_YEAR, PARTITION_END_YEAR + 1):
-        for week in range(1, iso_weeks_in_year(year) + 1):
-            table_name = f"posts_{year}_{week:02d}"
-            op.execute(
-                f"ALTER TABLE {schema}.posts "
-                f"DETACH PARTITION {schema}.{table_name}"
-            )

python/alembic/env.py

@@ -3,7 +3,6 @@
 from __future__ import annotations
 
 import logging
-import re
 import sys
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Literal

python/alembic/richie/versions/2026_03_29-removed_ds_table_from_richie_db_c8a794340928.py

@@ -0,0 +1,187 @@
+"""removed ds table from richie DB.
+
+Revision ID: c8a794340928
+Revises: 6b275323f435
+Create Date: 2026-03-29 15:29:23.643146
+
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+import sqlalchemy as sa
+from alembic import op
+from sqlalchemy.dialects import postgresql
+
+from python.orm import RichieBase
+
+if TYPE_CHECKING:
+    from collections.abc import Sequence
+
+# revision identifiers, used by Alembic.
+revision: str = "c8a794340928"
+down_revision: str | None = "6b275323f435"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+schema = RichieBase.schema_name
+
+
+def upgrade() -> None:
+    """Upgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_table("vote_record", schema=schema)
+    op.drop_index(op.f("ix_vote_congress_chamber"), table_name="vote", schema=schema)
+    op.drop_index(op.f("ix_vote_date"), table_name="vote", schema=schema)
+    op.drop_index(op.f("ix_legislator_bioguide_id"), table_name="legislator", schema=schema)
+    op.drop_table("legislator", schema=schema)
+    op.drop_table("vote", schema=schema)
+    op.drop_index(op.f("ix_bill_congress"), table_name="bill", schema=schema)
+    op.drop_table("bill", schema=schema)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table(
+        "vote",
+        sa.Column("congress", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("chamber", sa.VARCHAR(), autoincrement=False, nullable=False),
+        sa.Column("session", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("number", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("vote_type", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("question", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("result", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("result_text", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("vote_date", sa.DATE(), autoincrement=False, nullable=False),
+        sa.Column("yea_count", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("nay_count", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("not_voting_count", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("present_count", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("bill_id", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("id", sa.INTEGER(), autoincrement=True, nullable=False),
+        sa.Column(
+            "created",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.Column(
+            "updated",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.ForeignKeyConstraint(["bill_id"], [f"{schema}.bill.id"], name=op.f("fk_vote_bill_id_bill")),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_vote")),
+        sa.UniqueConstraint(
+            "congress",
+            "chamber",
+            "session",
+            "number",
+            name=op.f("uq_vote_congress_chamber_session_number"),
+            postgresql_include=[],
+            postgresql_nulls_not_distinct=False,
+        ),
+        schema=schema,
+    )
+    op.create_index(op.f("ix_vote_date"), "vote", ["vote_date"], unique=False, schema=schema)
+    op.create_index(op.f("ix_vote_congress_chamber"), "vote", ["congress", "chamber"], unique=False, schema=schema)
+    op.create_table(
+        "vote_record",
+        sa.Column("vote_id", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("legislator_id", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("position", sa.VARCHAR(), autoincrement=False, nullable=False),
+        sa.ForeignKeyConstraint(
+            ["legislator_id"],
+            [f"{schema}.legislator.id"],
+            name=op.f("fk_vote_record_legislator_id_legislator"),
+            ondelete="CASCADE",
+        ),
+        sa.ForeignKeyConstraint(
+            ["vote_id"], [f"{schema}.vote.id"], name=op.f("fk_vote_record_vote_id_vote"), ondelete="CASCADE"
+        ),
+        sa.PrimaryKeyConstraint("vote_id", "legislator_id", name=op.f("pk_vote_record")),
+        schema=schema,
+    )
+    op.create_table(
+        "legislator",
+        sa.Column("bioguide_id", sa.TEXT(), autoincrement=False, nullable=False),
+        sa.Column("thomas_id", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("lis_id", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("govtrack_id", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("opensecrets_id", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("fec_ids", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("first_name", sa.VARCHAR(), autoincrement=False, nullable=False),
+        sa.Column("last_name", sa.VARCHAR(), autoincrement=False, nullable=False),
+        sa.Column("official_full_name", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("nickname", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("birthday", sa.DATE(), autoincrement=False, nullable=True),
+        sa.Column("gender", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("current_party", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("current_state", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("current_district", sa.INTEGER(), autoincrement=False, nullable=True),
+        sa.Column("current_chamber", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("id", sa.INTEGER(), autoincrement=True, nullable=False),
+        sa.Column(
+            "created",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.Column(
+            "updated",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_legislator")),
+        schema=schema,
+    )
+    op.create_index(op.f("ix_legislator_bioguide_id"), "legislator", ["bioguide_id"], unique=True, schema=schema)
+    op.create_table(
+        "bill",
+        sa.Column("congress", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("bill_type", sa.VARCHAR(), autoincrement=False, nullable=False),
+        sa.Column("number", sa.INTEGER(), autoincrement=False, nullable=False),
+        sa.Column("title", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("title_short", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("official_title", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("status", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("status_at", sa.DATE(), autoincrement=False, nullable=True),
+        sa.Column("sponsor_bioguide_id", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("subjects_top_term", sa.VARCHAR(), autoincrement=False, nullable=True),
+        sa.Column("id", sa.INTEGER(), autoincrement=True, nullable=False),
+        sa.Column(
+            "created",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.Column(
+            "updated",
+            postgresql.TIMESTAMP(timezone=True),
+            server_default=sa.text("now()"),
+            autoincrement=False,
+            nullable=False,
+        ),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_bill")),
+        sa.UniqueConstraint(
+            "congress",
+            "bill_type",
+            "number",
+            name=op.f("uq_bill_congress_type_number"),
+            postgresql_include=[],
+            postgresql_nulls_not_distinct=False,
+        ),
+        schema=schema,
+    )
+    op.create_index(op.f("ix_bill_congress"), "bill", ["congress"], unique=False, schema=schema)
+    # ### end Alembic commands ###

python/alembic/richie/versions/2026_06_03-adding_audiobook_libreary_metadata_d7864d1ffc17.py

@@ -0,0 +1,93 @@
+"""adding audiobook libreary metadata.
+
+Revision ID: d7864d1ffc17
+Revises: c8a794340928
+Create Date: 2026-06-03 20:24:09.200837
+
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+import sqlalchemy as sa
+from alembic import op
+
+from python.orm import RichieBase
+
+if TYPE_CHECKING:
+    from collections.abc import Sequence
+
+# revision identifiers, used by Alembic.
+revision: str = "d7864d1ffc17"
+down_revision: str | None = "c8a794340928"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+schema = RichieBase.schema_name
+
+
+def upgrade() -> None:
+    """Upgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table(
+        "audiobook_author",
+        sa.Column("name", sa.String(), nullable=False),
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_audiobook_author")),
+        sa.UniqueConstraint("name", name=op.f("uq_audiobook_author_name")),
+        schema=schema,
+    )
+    op.create_table(
+        "audiobook_series",
+        sa.Column("name", sa.String(), nullable=False),
+        sa.Column("author_id", sa.Integer(), nullable=False),
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["author_id"],
+            [f"{schema}.audiobook_author.id"],
+            name=op.f("fk_audiobook_series_author_id_audiobook_author"),
+            ondelete="CASCADE",
+        ),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_audiobook_series")),
+        sa.UniqueConstraint("author_id", "name", name=op.f("uq_audiobook_series_author_id")),
+        schema=schema,
+    )
+    op.create_table(
+        "audiobook",
+        sa.Column("title", sa.String(), nullable=False),
+        sa.Column("author_id", sa.Integer(), nullable=False),
+        sa.Column("series_id", sa.Integer(), nullable=True),
+        sa.Column("series_index", sa.Integer(), nullable=False),
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["author_id"],
+            [f"{schema}.audiobook_author.id"],
+            name=op.f("fk_audiobook_author_id_audiobook_author"),
+            ondelete="CASCADE",
+        ),
+        sa.ForeignKeyConstraint(
+            ["series_id"],
+            [f"{schema}.audiobook_series.id"],
+            name=op.f("fk_audiobook_series_id_audiobook_series"),
+            ondelete="SET NULL",
+        ),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_audiobook")),
+        schema=schema,
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_table("audiobook", schema=schema)
+    op.drop_table("audiobook_series", schema=schema)
+    op.drop_table("audiobook_author", schema=schema)
+    # ### end Alembic commands ###

python/alembic/richie/versions/2026_06_10-updated_series_index_to_float_and_added__b3c60cc5beb5.py

@@ -0,0 +1,63 @@
+"""updated series_index to float and added UniqueConstraint to audiobook and audiobook_author.
+
+Revision ID: b3c60cc5beb5
+Revises: d7864d1ffc17
+Create Date: 2026-06-10 20:02:43.073725
+
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+import sqlalchemy as sa
+from alembic import op
+
+from python.orm import RichieBase
+
+if TYPE_CHECKING:
+    from collections.abc import Sequence
+
+# revision identifiers, used by Alembic.
+revision: str = "b3c60cc5beb5"
+down_revision: str | None = "d7864d1ffc17"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+schema = RichieBase.schema_name
+
+
+def upgrade() -> None:
+    """Upgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column(
+        "audiobook",
+        "series_index",
+        existing_type=sa.INTEGER(),
+        type_=sa.Float(),
+        existing_nullable=False,
+        schema=schema,
+    )
+    op.create_unique_constraint(
+        op.f("uq_audiobook_author_id"),
+        "audiobook",
+        ["author_id", "series_id", "title"],
+        schema=schema,
+        postgresql_nulls_not_distinct=True,
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_constraint(op.f("uq_audiobook_author_id"), "audiobook", schema=schema, type_="unique")
+    op.alter_column(
+        "audiobook",
+        "series_index",
+        existing_type=sa.Float(),
+        type_=sa.INTEGER(),
+        existing_nullable=False,
+        schema=schema,
+    )
+    # ### end Alembic commands ###

python/alembic/signal_bot/versions/2026_03_17-seprating_signal_bot_database_6eaf696e07a5.py

@@ -1,100 +0,0 @@
-"""seprating signal_bot database.
-
-Revision ID: 6eaf696e07a5
-Revises:
-Create Date: 2026-03-17 21:35:37.612672
-
-"""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-import sqlalchemy as sa
-from alembic import op
-from sqlalchemy.dialects import postgresql
-
-from python.orm import SignalBotBase
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-
-# revision identifiers, used by Alembic.
-revision: str = "6eaf696e07a5"
-down_revision: str | None = None
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-schema = SignalBotBase.schema_name
-
-
-def upgrade() -> None:
-    """Upgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.create_table(
-        "dead_letter_message",
-        sa.Column("source", sa.String(), nullable=False),
-        sa.Column("message", sa.Text(), nullable=False),
-        sa.Column("received_at", sa.DateTime(timezone=True), nullable=False),
-        sa.Column(
-            "status", postgresql.ENUM("UNPROCESSED", "PROCESSED", name="message_status", schema=schema), nullable=False
-        ),
-        sa.Column("id", sa.Integer(), nullable=False),
-        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.PrimaryKeyConstraint("id", name=op.f("pk_dead_letter_message")),
-        schema=schema,
-    )
-    op.create_table(
-        "role",
-        sa.Column("name", sa.String(length=50), nullable=False),
-        sa.Column("id", sa.SmallInteger(), nullable=False),
-        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.PrimaryKeyConstraint("id", name=op.f("pk_role")),
-        sa.UniqueConstraint("name", name=op.f("uq_role_name")),
-        schema=schema,
-    )
-    op.create_table(
-        "signal_device",
-        sa.Column("phone_number", sa.String(length=50), nullable=False),
-        sa.Column("safety_number", sa.String(), nullable=True),
-        sa.Column(
-            "trust_level",
-            postgresql.ENUM("VERIFIED", "UNVERIFIED", "BLOCKED", name="trust_level", schema=schema),
-            nullable=False,
-        ),
-        sa.Column("last_seen", sa.DateTime(timezone=True), nullable=False),
-        sa.Column("id", sa.Integer(), nullable=False),
-        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.PrimaryKeyConstraint("id", name=op.f("pk_signal_device")),
-        sa.UniqueConstraint("phone_number", name=op.f("uq_signal_device_phone_number")),
-        schema=schema,
-    )
-    op.create_table(
-        "device_role",
-        sa.Column("device_id", sa.Integer(), nullable=False),
-        sa.Column("role_id", sa.SmallInteger(), nullable=False),
-        sa.Column("id", sa.Integer(), nullable=False),
-        sa.Column("created", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.Column("updated", sa.DateTime(timezone=True), server_default=sa.text("now()"), nullable=False),
-        sa.ForeignKeyConstraint(
-            ["device_id"], [f"{schema}.signal_device.id"], name=op.f("fk_device_role_device_id_signal_device")
-        ),
-        sa.ForeignKeyConstraint(["role_id"], [f"{schema}.role.id"], name=op.f("fk_device_role_role_id_role")),
-        sa.PrimaryKeyConstraint("id", name=op.f("pk_device_role")),
-        sa.UniqueConstraint("device_id", "role_id", name="uq_device_role_device_role"),
-        schema=schema,
-    )
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    """Downgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_table("device_role", schema=schema)
-    op.drop_table("signal_device", schema=schema)
-    op.drop_table("role", schema=schema)
-    op.drop_table("dead_letter_message", schema=schema)
-    # ### end Alembic commands ###

python/alembic/signal_bot/versions/2026_03_18-test_66bdd532bcab.py

@@ -1,72 +0,0 @@
-"""test.
-
-Revision ID: 66bdd532bcab
-Revises: 6eaf696e07a5
-Create Date: 2026-03-18 19:21:14.561568
-
-"""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-import sqlalchemy as sa
-from alembic import op
-from sqlalchemy.dialects import postgresql
-
-from python.orm import SignalBotBase
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-
-# revision identifiers, used by Alembic.
-revision: str = "66bdd532bcab"
-down_revision: str | None = "6eaf696e07a5"
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-schema = SignalBotBase.schema_name
-
-
-def upgrade() -> None:
-    """Upgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.alter_column(
-        "dead_letter_message",
-        "status",
-        existing_type=postgresql.ENUM("UNPROCESSED", "PROCESSED", name="message_status", schema=schema),
-        type_=sa.Enum("UNPROCESSED", "PROCESSED", name="message_status", native_enum=False),
-        existing_nullable=False,
-        schema=schema,
-    )
-    op.alter_column(
-        "signal_device",
-        "trust_level",
-        existing_type=postgresql.ENUM("VERIFIED", "UNVERIFIED", "BLOCKED", name="trust_level", schema=schema),
-        type_=sa.Enum("VERIFIED", "UNVERIFIED", "BLOCKED", name="trust_level", native_enum=False),
-        existing_nullable=False,
-        schema=schema,
-    )
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    """Downgrade."""
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.alter_column(
-        "signal_device",
-        "trust_level",
-        existing_type=sa.Enum("VERIFIED", "UNVERIFIED", "BLOCKED", name="trust_level", native_enum=False),
-        type_=postgresql.ENUM("VERIFIED", "UNVERIFIED", "BLOCKED", name="trust_level", schema=schema),
-        existing_nullable=False,
-        schema=schema,
-    )
-    op.alter_column(
-        "dead_letter_message",
-        "status",
-        existing_type=sa.Enum("UNPROCESSED", "PROCESSED", name="message_status", native_enum=False),
-        type_=postgresql.ENUM("UNPROCESSED", "PROCESSED", name="message_status", schema=schema),
-        existing_nullable=False,
-        schema=schema,
-    )
-    # ### end Alembic commands ###

python/data_science/__init__.py

@@ -1,3 +0,0 @@
-"""Data science CLI tools."""
-
-from __future__ import annotations

python/data_science/bluesky_ids.py

@@ -1,104 +0,0 @@
-"""Utilities for converting Bluesky identifiers to numeric database IDs.
-
-Handles DID-to-user_id hashing, TID-to-post_id decoding, and AT-URI parsing.
-"""
-
-from __future__ import annotations
-
-import hashlib
-
-TID_CHARSET = "234567abcdefghijklmnopqrstuvwxyz"
-_TID_LENGTH = 13
-_BIGINT_MASK = 0x7FFFFFFFFFFFFFFF
-_AT_URI_SEGMENT_COUNT = 3
-
-
-def did_to_user_id(did: str) -> int:
-    """Convert a DID string to a deterministic 63-bit integer for user_id.
-
-    Uses SHA-256, truncated to 63 bits (positive signed BigInteger range).
-    Collision probability is negligible at Bluesky's scale (~tens of millions of users).
-
-    Args:
-        did: A Bluesky DID string, e.g. "did:plc:abc123".
-
-    Returns:
-        A positive 63-bit integer suitable for BigInteger storage.
-    """
-    digest = hashlib.sha256(did.encode()).digest()
-    return int.from_bytes(digest[:8], "big") & _BIGINT_MASK
-
-
-def tid_to_integer(tid: str) -> int:
-    """Decode a Bluesky TID (base32-sortbase) into a 64-bit integer for post_id.
-
-    TIDs are 13-character, base32-sortbase encoded identifiers that encode a
-    microsecond timestamp plus a clock ID. They are globally unique by construction.
-
-    Args:
-        tid: A 13-character TID string, e.g. "3abc2defghijk".
-
-    Returns:
-        A positive integer suitable for BigInteger storage.
-
-    Raises:
-        ValueError: If the TID is malformed (wrong length or invalid characters).
-    """
-    if len(tid) != _TID_LENGTH:
-        message = f"TID must be {_TID_LENGTH} characters, got {len(tid)}: {tid!r}"
-        raise ValueError(message)
-
-    result = 0
-    for char in tid:
-        index = TID_CHARSET.find(char)
-        if index == -1:
-            message = f"Invalid character {char!r} in TID {tid!r}"
-            raise ValueError(message)
-        result = result * 32 + index
-    return result
-
-
-def parse_at_uri(uri: str) -> tuple[str, str, str]:
-    """Parse an AT-URI into its components.
-
-    Args:
-        uri: An AT-URI string, e.g. "at://did:plc:abc123/app.bsky.feed.post/3abc2defghijk".
-
-    Returns:
-        A tuple of (did, collection, rkey).
-
-    Raises:
-        ValueError: If the URI doesn't have the expected format.
-    """
-    stripped = uri.removeprefix("at://")
-    parts = stripped.split("/", maxsplit=2)
-    if len(parts) != _AT_URI_SEGMENT_COUNT:
-        message = f"Expected {_AT_URI_SEGMENT_COUNT} path segments in AT-URI, got {len(parts)}: {uri!r}"
-        raise ValueError(message)
-    return parts[0], parts[1], parts[2]
-
-
-def post_id_from_uri(uri: str) -> int:
-    """Extract and decode the post_id (TID) from an AT-URI.
-
-    Args:
-        uri: An AT-URI pointing to a post.
-
-    Returns:
-        The post_id as an integer.
-    """
-    _did, _collection, rkey = parse_at_uri(uri)
-    return tid_to_integer(rkey)
-
-
-def user_id_from_uri(uri: str) -> int:
-    """Extract and hash the user_id (DID) from an AT-URI.
-
-    Args:
-        uri: An AT-URI pointing to a post.
-
-    Returns:
-        The user_id as an integer.
-    """
-    did, _collection, _rkey = parse_at_uri(uri)
-    return did_to_user_id(did)

python/data_science/bluesky_transform.py

@@ -1,143 +0,0 @@
-"""Transform Bluesky Jetstream messages into rows matching the Posts table schema."""
-
-from __future__ import annotations
-
-import json
-import logging
-from datetime import datetime
-
-from python.data_science.bluesky_ids import (
-    did_to_user_id,
-    post_id_from_uri,
-    tid_to_integer,
-    user_id_from_uri,
-)
-
-logger = logging.getLogger(__name__)
-
-INSTANCE = "bsky"
-POST_COLLECTION = "app.bsky.feed.post"
-EMBED_RECORD_TYPE = "app.bsky.embed.record"
-EMBED_RECORD_WITH_MEDIA_TYPE = "app.bsky.embed.recordWithMedia"
-
-
-def transform_jetstream_post(message: dict) -> dict:
-    """Transform a Jetstream commit message into a dict matching Posts table columns.
-
-    Expects a Jetstream message with kind=commit, operation=create,
-    collection=app.bsky.feed.post.
-
-    Args:
-        message: The full Jetstream JSON message.
-
-    Returns:
-        A dict with keys matching the Posts table columns.
-    """
-    did = message["did"]
-    commit = message["commit"]
-    record = commit["record"]
-
-    row: dict = {
-        "post_id": tid_to_integer(commit["rkey"]),
-        "user_id": did_to_user_id(did),
-        "instance": INSTANCE,
-        "date": datetime.fromisoformat(record["createdAt"]),
-        "text": record.get("text", ""),
-        "langs": _extract_langs(record),
-        "like_count": 0,
-        "reply_count": 0,
-        "repost_count": 0,
-        "reply_to": None,
-        "replied_author": None,
-        "thread_root": None,
-        "thread_root_author": None,
-        "repost_from": None,
-        "reposted_author": None,
-        "quotes": None,
-        "quoted_author": None,
-        "labels": _extract_labels(record),
-        "sent_label": None,
-        "sent_score": None,
-    }
-
-    _extract_reply_refs(record, row)
-    _extract_quote_refs(record, row)
-
-    return row
-
-
-def is_post_create(message: dict) -> bool:
-    """Check if a Jetstream message is a post creation event.
-
-    Args:
-        message: The full Jetstream JSON message.
-
-    Returns:
-        True if this is a create commit for app.bsky.feed.post.
-    """
-    if message.get("kind") != "commit":
-        return False
-    commit = message.get("commit", {})
-    return commit.get("operation") == "create" and commit.get("collection") == POST_COLLECTION
-
-
-def _extract_langs(record: dict) -> str | None:
-    """Extract langs array as a JSON string, or None if absent."""
-    langs = record.get("langs")
-    if langs is None:
-        return None
-    return json.dumps(langs)
-
-
-def _extract_labels(record: dict) -> str | None:
-    """Extract self-labels as a JSON string, or None if absent."""
-    labels_obj = record.get("labels")
-    if labels_obj is None:
-        return None
-    values = labels_obj.get("values", [])
-    if not values:
-        return None
-    label_strings = [label.get("val", "") for label in values]
-    return json.dumps(label_strings)
-
-
-def _extract_reply_refs(record: dict, row: dict) -> None:
-    """Populate reply_to, replied_author, thread_root, thread_root_author from record.reply."""
-    reply = record.get("reply")
-    if reply is None:
-        return
-
-    parent = reply.get("parent", {})
-    parent_uri = parent.get("uri")
-    if parent_uri:
-        row["reply_to"] = post_id_from_uri(parent_uri)
-        row["replied_author"] = user_id_from_uri(parent_uri)
-
-    root = reply.get("root", {})
-    root_uri = root.get("uri")
-    if root_uri:
-        row["thread_root"] = post_id_from_uri(root_uri)
-        row["thread_root_author"] = user_id_from_uri(root_uri)
-
-
-def _extract_quote_refs(record: dict, row: dict) -> None:
-    """Populate quotes and quoted_author from embed record references."""
-    embed = record.get("embed")
-    if embed is None:
-        return
-
-    embed_type = embed.get("$type", "")
-
-    if embed_type == EMBED_RECORD_TYPE:
-        _set_quote_from_record(embed.get("record", {}), row)
-    elif embed_type == EMBED_RECORD_WITH_MEDIA_TYPE:
-        inner_record = embed.get("record", {}).get("record", {})
-        _set_quote_from_record(inner_record, row)
-
-
-def _set_quote_from_record(record_ref: dict, row: dict) -> None:
-    """Set quotes and quoted_author from a record reference object."""
-    uri = record_ref.get("uri")
-    if uri and POST_COLLECTION in uri:
-        row["quotes"] = post_id_from_uri(uri)
-        row["quoted_author"] = user_id_from_uri(uri)

python/data_science/firehose_consumer.py

@@ -1,203 +0,0 @@
-"""Kafka consumer that ingests Bluesky posts into the partitioned Posts table.
-
-Consumes Jetstream messages from Kafka, transforms them into Posts rows,
-and batch-inserts them into PostgreSQL with manual offset commits.
-
-Usage:
-    firehose-consumer
-    firehose-consumer --kafka-servers kafka:9092 --batch-size 500
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import signal
-from os import getenv
-from threading import Event
-from typing import Annotated
-
-import typer
-from confluent_kafka import Consumer, KafkaError, KafkaException
-from sqlalchemy.orm import Session
-
-from python.data_science.bluesky_transform import is_post_create, transform_jetstream_post
-from python.data_science.ingest_posts import ingest_batch
-from python.orm.common import get_postgres_engine
-from python.orm.data_science_dev.posts.failed_ingestion import FailedIngestion
-
-logger = logging.getLogger(__name__)
-
-DEFAULT_TOPIC = "bluesky.firehose.posts"
-DEFAULT_KAFKA_SERVERS = "localhost:9092"
-DEFAULT_GROUP_ID = "bluesky-posts-ingestor"
-DEFAULT_BATCH_SIZE = 500
-POLL_TIMEOUT_SECONDS = 5.0
-
-shutdown_event = Event()
-
-app = typer.Typer(help="Consume Bluesky posts from Kafka and ingest into PostgreSQL.")
-
-
-@app.command()
-def main(
-    kafka_servers: Annotated[str, typer.Option(help="Kafka bootstrap servers")] = "",
-    topic: Annotated[str, typer.Option(help="Kafka topic to consume from")] = "",
-    group_id: Annotated[str, typer.Option(help="Kafka consumer group ID")] = "",
-    batch_size: Annotated[int, typer.Option(help="Messages per DB insert batch")] = DEFAULT_BATCH_SIZE,
-) -> None:
-    """Consume Bluesky posts from Kafka and ingest into the partitioned posts table."""
-    logging.basicConfig(
-        level=logging.INFO,
-        format="%(asctime)s %(levelname)s %(message)s",
-        datefmt="%H:%M:%S",
-    )
-
-    servers = kafka_servers or getenv("KAFKA_BOOTSTRAP_SERVERS", DEFAULT_KAFKA_SERVERS)
-    topic_name = topic or getenv("BLUESKY_FIREHOSE_TOPIC", DEFAULT_TOPIC)
-    group = group_id or getenv("KAFKA_GROUP_ID", DEFAULT_GROUP_ID)
-
-    signal.signal(signal.SIGTERM, _handle_shutdown)
-    signal.signal(signal.SIGINT, _handle_shutdown)
-
-    consumer = _create_consumer(servers, group)
-    consumer.subscribe([topic_name])
-
-    engine = get_postgres_engine(name="DATA_SCIENCE_DEV")
-    total_inserted = 0
-
-    logger.info("Starting firehose consumer: topic=%s group=%s batch_size=%d", topic_name, group, batch_size)
-
-    try:
-        with Session(engine) as session:
-            while not shutdown_event.is_set():
-                inserted = _consume_batch(consumer, session, batch_size)
-                total_inserted += inserted
-                if inserted > 0:
-                    logger.info("Batch inserted %d rows (total: %d)", inserted, total_inserted)
-    except KafkaException:
-        logger.exception("Fatal Kafka error")
-    finally:
-        logger.info("Closing consumer (total inserted: %d)", total_inserted)
-        consumer.close()
-
-
-def _consume_batch(consumer: Consumer, session: Session, batch_size: int) -> int:
-    """Poll a batch of messages, transform, and insert into the database.
-
-    Args:
-        consumer: The Kafka consumer instance.
-        session: SQLAlchemy database session.
-        batch_size: Maximum number of messages to consume per batch.
-
-    Returns:
-        Number of rows successfully inserted.
-    """
-    messages = consumer.consume(num_messages=batch_size, timeout=POLL_TIMEOUT_SECONDS)
-    if not messages:
-        return 0
-
-    rows: list[dict] = []
-    for message in messages:
-        error = message.error()
-        if error is not None:
-            if error.code() == KafkaError._PARTITION_EOF:  # noqa: SLF001 — confluent-kafka exposes this as a pseudo-private constant; no public alternative exists
-                continue
-            logger.error("Consumer error: %s", error)
-            continue
-
-        row = _safe_transform(message.value(), session)
-        if row is not None:
-            rows.append(row)
-
-    if not rows:
-        consumer.commit(asynchronous=False)
-        return 0
-
-    inserted = ingest_batch(session, rows)
-    consumer.commit(asynchronous=False)
-    return inserted
-
-
-def _safe_transform(raw_value: bytes | None, session: Session) -> dict | None:
-    """Transform a Kafka message value into a Posts row, logging failures.
-
-    Args:
-        raw_value: Raw message bytes from Kafka.
-        session: SQLAlchemy session for logging failures.
-
-    Returns:
-        A transformed row dict, or None if transformation failed.
-    """
-    if raw_value is None:
-        return None
-
-    try:
-        message = json.loads(raw_value)
-    except (json.JSONDecodeError, UnicodeDecodeError):
-        logger.exception("Failed to decode Kafka message")
-        _log_failed_ingestion(session, raw_value, "JSON decode error")
-        return None
-
-    if not is_post_create(message):
-        return None
-
-    try:
-        return transform_jetstream_post(message)
-    except (KeyError, ValueError, TypeError):
-        logger.exception("Failed to transform Jetstream message")
-        _log_failed_ingestion(session, raw_value, "Transform error")
-        return None
-
-
-def _log_failed_ingestion(session: Session, raw_value: bytes, error: str) -> None:
-    """Log a failed ingestion to the FailedIngestion table.
-
-    Args:
-        session: SQLAlchemy session.
-        raw_value: The raw message bytes.
-        error: Description of the error.
-    """
-    try:
-        session.add(
-            FailedIngestion(
-                raw_line=raw_value.decode(errors="replace")[:10000],
-                error=error,
-            )
-        )
-        session.commit()
-    except Exception:
-        session.rollback()
-        logger.exception("Failed to log ingestion failure")
-
-
-def _create_consumer(servers: str, group: str) -> Consumer:
-    """Create a configured Kafka consumer.
-
-    Args:
-        servers: Kafka bootstrap servers string.
-        group: Consumer group ID.
-
-    Returns:
-        A configured confluent_kafka.Consumer.
-    """
-    config = {
-        "bootstrap.servers": servers,
-        "group.id": group,
-        "auto.offset.reset": "earliest",
-        "enable.auto.commit": False,
-        "max.poll.interval.ms": 300000,
-        "fetch.min.bytes": 1024,
-        "session.timeout.ms": 30000,
-    }
-    return Consumer(config)
-
-
-def _handle_shutdown(_signum: int, _frame: object) -> None:
-    """Signal handler to trigger graceful shutdown."""
-    logger.info("Shutdown signal received")
-    shutdown_event.set()
-
-
-if __name__ == "__main__":
-    app()

python/data_science/firehose_producer.py

@@ -1,230 +0,0 @@
-"""Bluesky Jetstream firehose to Kafka producer.
-
-Connects to the Bluesky Jetstream WebSocket API with zstd compression,
-filters for post creation events, and produces them to a Kafka topic.
-
-Usage:
-    firehose-producer
-    firehose-producer --kafka-servers kafka:9092 --topic bluesky.firehose.posts
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-import signal
-from os import getenv
-from threading import Event
-from typing import Annotated
-
-import typer
-from compression import zstd
-from confluent_kafka import KafkaError, KafkaException, Producer
-from websockets.exceptions import ConnectionClosed
-from websockets.sync.client import connect
-
-logger = logging.getLogger(__name__)
-
-JETSTREAM_URL = "wss://jetstream2.us-east.bsky.network/subscribe"
-DEFAULT_TOPIC = "bluesky.firehose.posts"
-DEFAULT_KAFKA_SERVERS = "localhost:9092"
-POLL_INTERVAL = 100
-POST_COLLECTION = "app.bsky.feed.post"
-
-shutdown_event = Event()
-
-app = typer.Typer(help="Stream Bluesky firehose posts into Kafka.")
-
-
-@app.command()
-def main(
-    kafka_servers: Annotated[str, typer.Option(help="Kafka bootstrap servers")] = "",
-    topic: Annotated[str, typer.Option(help="Kafka topic to produce to")] = "",
-    collections: Annotated[str, typer.Option(help="Comma-separated collections to subscribe to")] = POST_COLLECTION,
-) -> None:
-    """Connect to Bluesky Jetstream and produce post events to Kafka."""
-    logging.basicConfig(
-        level=logging.INFO,
-        format="%(asctime)s %(levelname)s %(message)s",
-        datefmt="%H:%M:%S",
-    )
-
-    servers = kafka_servers or getenv("KAFKA_BOOTSTRAP_SERVERS", DEFAULT_KAFKA_SERVERS)
-    topic_name = topic or getenv("BLUESKY_FIREHOSE_TOPIC", DEFAULT_TOPIC)
-
-    signal.signal(signal.SIGTERM, _handle_shutdown)
-    signal.signal(signal.SIGINT, _handle_shutdown)
-
-    producer = _create_producer(servers)
-    cursor: int | None = None
-
-    logger.info("Starting firehose producer → %s on %s", topic_name, servers)
-
-    while not shutdown_event.is_set():
-        try:
-            cursor = _stream_loop(producer, topic_name, collections, cursor)
-        except (ConnectionClosed, OSError):
-            logger.exception("WebSocket disconnected, reconnecting")
-        except KafkaException:
-            logger.exception("Kafka error, reconnecting")
-
-        if not shutdown_event.is_set():
-            logger.info("Reconnecting in 5 seconds (cursor=%s)", cursor)
-            shutdown_event.wait(timeout=5)
-
-    logger.info("Shutting down, flushing producer")
-    producer.flush(timeout=30)
-    logger.info("Producer shutdown complete")
-
-
-def _stream_loop(
-    producer: Producer,
-    topic: str,
-    collections: str,
-    cursor: int | None,
-) -> int | None:
-    """Connect to Jetstream and stream messages to Kafka until disconnected.
-
-    Args:
-        producer: The Kafka producer instance.
-        topic: Kafka topic name.
-        collections: Comma-separated AT Protocol collections to subscribe to.
-        cursor: Optional microsecond timestamp to resume from.
-
-    Returns:
-        The last processed time_us cursor value.
-    """
-    url = _build_jetstream_url(collections, cursor)
-    logger.info("Connecting to %s", url)
-
-    message_count = 0
-    last_cursor = cursor
-
-    with connect(url, additional_headers={"Accept-Encoding": "zstd"}) as websocket:
-        logger.info("Connected to Jetstream")
-
-        while not shutdown_event.is_set():
-            try:
-                raw_frame = websocket.recv(timeout=10)
-            except TimeoutError:
-                producer.poll(0)
-                continue
-
-            text = _decode_frame(raw_frame)
-            message = json.loads(text)
-
-            time_us = message.get("time_us")
-            if time_us is not None:
-                last_cursor = time_us
-
-            if not _is_post_create(message):
-                continue
-
-            did = message.get("did", "")
-
-            try:
-                producer.produce(
-                    topic,
-                    key=did.encode(),
-                    value=text.encode() if isinstance(text, str) else text,
-                    callback=_delivery_callback,
-                )
-            except BufferError:
-                logger.warning("Producer buffer full, flushing")
-                producer.flush(timeout=10)
-                producer.produce(
-                    topic,
-                    key=did.encode(),
-                    value=text.encode() if isinstance(text, str) else text,
-                    callback=_delivery_callback,
-                )
-
-            message_count += 1
-            if message_count % POLL_INTERVAL == 0:
-                producer.poll(0)
-
-            if message_count % 10000 == 0:
-                logger.info("Produced %d messages (cursor=%s)", message_count, last_cursor)
-
-    return last_cursor
-
-
-def _build_jetstream_url(collections: str, cursor: int | None) -> str:
-    """Build the Jetstream WebSocket URL with query parameters.
-
-    Args:
-        collections: Comma-separated collection names.
-        cursor: Optional microsecond timestamp for resumption.
-
-    Returns:
-        The full WebSocket URL.
-    """
-    params = ["compress=true"]
-    for raw_collection in collections.split(","):
-        cleaned = raw_collection.strip()
-        if cleaned:
-            params.append(f"wantedCollections={cleaned}")
-    if cursor is not None:
-        params.append(f"cursor={cursor}")
-    return f"{JETSTREAM_URL}?{'&'.join(params)}"
-
-
-def _decode_frame(frame: str | bytes) -> str:
-    """Decode a WebSocket frame, decompressing zstd if binary.
-
-    Jetstream with compress=true sends zstd-compressed binary frames.
-
-    Args:
-        frame: Raw WebSocket frame data.
-
-    Returns:
-        The decoded JSON string.
-    """
-    if isinstance(frame, bytes):
-        return zstd.decompress(frame).decode()
-    return frame
-
-
-def _is_post_create(message: dict) -> bool:
-    """Check if a Jetstream message is a post creation commit."""
-    if message.get("kind") != "commit":
-        return False
-    commit = message.get("commit", {})
-    return commit.get("operation") == "create" and commit.get("collection") == POST_COLLECTION
-
-
-def _create_producer(servers: str) -> Producer:
-    """Create a configured Kafka producer.
-
-    Args:
-        servers: Kafka bootstrap servers string.
-
-    Returns:
-        A configured confluent_kafka.Producer.
-    """
-    config = {
-        "bootstrap.servers": servers,
-        "linger.ms": 50,
-        "batch.size": 65536,
-        "compression.type": "zstd",
-        "acks": "all",
-        "retries": 5,
-        "retry.backoff.ms": 500,
-    }
-    return Producer(config)
-
-
-def _delivery_callback(error: KafkaError | None, _message: object) -> None:
-    """Log delivery failures from the Kafka producer."""
-    if error is not None:
-        logger.error("Kafka delivery failed: %s", error)
-
-
-def _handle_shutdown(_signum: int, _frame: object) -> None:
-    """Signal handler to trigger graceful shutdown."""
-    logger.info("Shutdown signal received")
-    shutdown_event.set()
-
-
-if __name__ == "__main__":
-    app()

python/data_science/ingest_posts.py

@@ -1,247 +0,0 @@
-"""Ingestion pipeline for loading JSONL post files into the weekly-partitioned posts table.
-
-Usage:
-    ingest-posts /path/to/files/
-    ingest-posts /path/to/single_file.jsonl
-    ingest-posts /data/dir/ --workers 4 --batch-size 5000
-"""
-
-from __future__ import annotations
-
-import logging
-from datetime import UTC, datetime
-from pathlib import Path  # noqa: TC003 this is needed for typer
-from typing import TYPE_CHECKING, Annotated
-
-import orjson
-import psycopg
-import typer
-
-from python.common import configure_logger
-from python.orm.common import get_connection_info
-from python.parallelize import parallelize_process
-
-if TYPE_CHECKING:
-    from collections.abc import Iterator
-
-logger = logging.getLogger(__name__)
-
-
-app = typer.Typer(help="Ingest JSONL post files into the partitioned posts table.")
-
-
-@app.command()
-def main(
-    path: Annotated[Path, typer.Argument(help="Directory containing JSONL files, or a single JSONL file")],
-    batch_size: Annotated[int, typer.Option(help="Rows per INSERT batch")] = 10000,
-    workers: Annotated[int, typer.Option(help="Parallel workers for multi-file ingestion")] = 4,
-    pattern: Annotated[str, typer.Option(help="Glob pattern for JSONL files")] = "*.jsonl",
-) -> None:
-    """Ingest JSONL post files into the weekly-partitioned posts table."""
-    configure_logger(level="INFO")
-
-    logger.info("starting ingest-posts")
-    logger.info("path=%s batch_size=%d workers=%d pattern=%s", path, batch_size, workers, pattern)
-    if path.is_file():
-        ingest_file(path, batch_size=batch_size)
-    elif path.is_dir():
-        ingest_directory(path, batch_size=batch_size, max_workers=workers, pattern=pattern)
-    else:
-        typer.echo(f"Path does not exist: {path}", err=True)
-        raise typer.Exit(code=1)
-
-    logger.info("ingest-posts done")
-
-
-def ingest_directory(
-    directory: Path,
-    *,
-    batch_size: int,
-    max_workers: int,
-    pattern: str = "*.jsonl",
-) -> None:
-    """Ingest all JSONL files in a directory using parallel workers."""
-    files = sorted(directory.glob(pattern))
-    if not files:
-        logger.warning("No JSONL files found in %s", directory)
-        return
-
-    logger.info("Found %d JSONL files to ingest", len(files))
-
-    kwargs_list = [{"path": fp, "batch_size": batch_size} for fp in files]
-    parallelize_process(ingest_file, kwargs_list, max_workers=max_workers)
-
-
-SCHEMA = "main"
-
-COLUMNS = (
-    "post_id",
-    "user_id",
-    "instance",
-    "date",
-    "text",
-    "langs",
-    "like_count",
-    "reply_count",
-    "repost_count",
-    "reply_to",
-    "replied_author",
-    "thread_root",
-    "thread_root_author",
-    "repost_from",
-    "reposted_author",
-    "quotes",
-    "quoted_author",
-    "labels",
-    "sent_label",
-    "sent_score",
-)
-
-INSERT_FROM_STAGING = f"""
-    INSERT INTO {SCHEMA}.posts ({", ".join(COLUMNS)})
-    SELECT {", ".join(COLUMNS)} FROM pg_temp.staging
-    ON CONFLICT (post_id, date) DO NOTHING
-"""  # noqa: S608
-
-FAILED_INSERT = f"""
-    INSERT INTO {SCHEMA}.failed_ingestion (raw_line, error)
-    VALUES (%(raw_line)s, %(error)s)
-"""  # noqa: S608
-
-
-def get_psycopg_connection() -> psycopg.Connection:
-    """Create a raw psycopg3 connection from environment variables."""
-    database, host, port, username, password = get_connection_info("DATA_SCIENCE_DEV")
-    return psycopg.connect(
-        dbname=database,
-        host=host,
-        port=int(port),
-        user=username,
-        password=password,
-        autocommit=False,
-    )
-
-
-def ingest_file(path: Path, *, batch_size: int) -> None:
-    """Ingest a single JSONL file into the posts table."""
-    log_trigger = max(100_000 // batch_size, 1)
-    failed_lines: list[dict] = []
-    try:
-        with get_psycopg_connection() as connection:
-            for index, batch in enumerate(read_jsonl_batches(path, batch_size, failed_lines), 1):
-                ingest_batch(connection, batch)
-                if index % log_trigger == 0:
-                    logger.info("Ingested %d batches (%d rows) from %s", index, index * batch_size, path)
-
-            if failed_lines:
-                logger.warning("Recording %d malformed lines from %s", len(failed_lines), path.name)
-                with connection.cursor() as cursor:
-                    cursor.executemany(FAILED_INSERT, failed_lines)
-                connection.commit()
-    except Exception:
-        logger.exception("Failed to ingest file: %s", path)
-        raise
-
-
-def ingest_batch(connection: psycopg.Connection, batch: list[dict]) -> None:
-    """COPY batch into a temp staging table, then INSERT ... ON CONFLICT into posts."""
-    if not batch:
-        return
-
-    try:
-        with connection.cursor() as cursor:
-            cursor.execute(f"""
-                CREATE TEMP TABLE IF NOT EXISTS staging
-                (LIKE {SCHEMA}.posts INCLUDING DEFAULTS)
-                ON COMMIT DELETE ROWS
-            """)
-            cursor.execute("TRUNCATE pg_temp.staging")
-
-            with cursor.copy(f"COPY pg_temp.staging ({', '.join(COLUMNS)}) FROM STDIN") as copy:
-                for row in batch:
-                    copy.write_row(tuple(row.get(column) for column in COLUMNS))
-
-            cursor.execute(INSERT_FROM_STAGING)
-        connection.commit()
-    except Exception as error:
-        connection.rollback()
-
-        if len(batch) == 1:
-            logger.exception("Skipping bad row post_id=%s", batch[0].get("post_id"))
-            with connection.cursor() as cursor:
-                cursor.execute(
-                    FAILED_INSERT,
-                    {
-                        "raw_line": orjson.dumps(batch[0], default=str).decode(),
-                        "error": str(error),
-                    },
-                )
-            connection.commit()
-            return
-
-        midpoint = len(batch) // 2
-        ingest_batch(connection, batch[:midpoint])
-        ingest_batch(connection, batch[midpoint:])
-
-
-def read_jsonl_batches(file_path: Path, batch_size: int, failed_lines: list[dict]) -> Iterator[list[dict]]:
-    """Stream a JSONL file and yield batches of transformed rows."""
-    batch: list[dict] = []
-    with file_path.open("r", encoding="utf-8") as handle:
-        for raw_line in handle:
-            line = raw_line.strip()
-            if not line:
-                continue
-            batch.extend(parse_line(line, file_path, failed_lines))
-            if len(batch) >= batch_size:
-                yield batch
-                batch = []
-    if batch:
-        yield batch
-
-
-def parse_line(line: str, file_path: Path, failed_lines: list[dict]) -> Iterator[dict]:
-    """Parse a JSONL line, handling concatenated JSON objects."""
-    try:
-        yield transform_row(orjson.loads(line))
-    except orjson.JSONDecodeError:
-        if "}{" not in line:
-            logger.warning("Skipping malformed line in %s: %s", file_path.name, line[:120])
-            failed_lines.append({"raw_line": line, "error": "malformed JSON"})
-            return
-        fragments = line.replace("}{", "}\n{").split("\n")
-        for fragment in fragments:
-            try:
-                yield transform_row(orjson.loads(fragment))
-            except (orjson.JSONDecodeError, KeyError, ValueError) as error:
-                logger.warning("Skipping malformed fragment in %s: %s", file_path.name, fragment[:120])
-                failed_lines.append({"raw_line": fragment, "error": str(error)})
-    except Exception as error:
-        logger.exception("Skipping bad row in %s: %s", file_path.name, line[:120])
-        failed_lines.append({"raw_line": line, "error": str(error)})
-
-
-def transform_row(raw: dict) -> dict:
-    """Transform a raw JSONL row into a dict matching the Posts table columns."""
-    raw["date"] = parse_date(raw["date"])
-    if raw.get("langs") is not None:
-        raw["langs"] = orjson.dumps(raw["langs"])
-    if raw.get("text") is not None:
-        raw["text"] = raw["text"].replace("\x00", "")
-    return raw
-
-
-def parse_date(raw_date: int) -> datetime:
-    """Parse compact YYYYMMDDHHmm integer into a naive datetime (input is UTC by spec)."""
-    return datetime(
-        raw_date // 100000000,
-        (raw_date // 1000000) % 100,
-        (raw_date // 10000) % 100,
-        (raw_date // 100) % 100,
-        raw_date % 100,
-        tzinfo=UTC,
-    )
-
-
-if __name__ == "__main__":
-    app()

python/database_cli.py

@@ -83,20 +83,6 @@ DATABASES: dict[str, DatabaseConfig] = {
         base_class_name="VanInventoryBase",
         models_module="python.orm.van_inventory.models",
     ),
-    "signal_bot": DatabaseConfig(
-        env_prefix="SIGNALBOT",
-        version_location="python/alembic/signal_bot/versions",
-        base_module="python.orm.signal_bot.base",
-        base_class_name="SignalBotBase",
-        models_module="python.orm.signal_bot.models",
-    ),
-    "data_science_dev": DatabaseConfig(
-        env_prefix="DATA_SCIENCE_DEV",
-        version_location="python/alembic/data_science_dev/versions",
-        base_module="python.orm.data_science_dev.base",
-        base_class_name="DataScienceDevBase",
-        models_module="python.orm.data_science_dev.models",
-    ),
 }
 
 

python/gitea.py

@@ -0,0 +1,347 @@
+"""Small Gitea API client for repository automation."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Self
+from urllib.parse import quote
+
+import httpx
+
+DEFAULT_PAGE_SIZE = 100
+EXPECTED_NO_CONTENT = 204
+EXPECTED_CREATED = 201
+EXPECTED_OK = 200
+
+
+@dataclass(frozen=True)
+class CreatedIssue:
+    """Issue data returned by Gitea."""
+
+    number: int | None
+    html_url: str | None
+    title: str
+
+
+@dataclass(frozen=True)
+class PullRequest:
+    """Pull request data returned by Gitea."""
+
+    number: int
+    title: str
+    html_url: str | None
+    labels: tuple[str, ...]
+    head_branch: str | None
+    base_branch: str | None
+
+
+@dataclass(frozen=True)
+class WorkflowJob:
+    """Workflow job data returned by Gitea Actions."""
+
+    id: int
+    name: str
+    run_id: int | None
+    status: str | None
+    conclusion: str | None
+
+
+class GiteaError(RuntimeError):
+    """Raised when Gitea rejects an API request."""
+
+
+def split_repo_name(repo: str) -> tuple[str, str]:
+    """Split an owner/repo string into its parts."""
+    owner, separator, repo_name = repo.partition("/")
+    if not separator or not owner or not repo_name:
+        msg = f"Invalid repository name: {repo}"
+        raise ValueError(msg)
+    return owner, repo_name
+
+
+class GiteaClient:
+    """HTTP client for the subset of Gitea APIs used in this repository."""
+
+    def __init__(
+        self,
+        *,
+        base_url: str,
+        token: str,
+        timeout: int = 30,
+        transport: httpx.BaseTransport | None = None,
+    ) -> None:
+        """Initialize the Gitea client."""
+        self._client = httpx.Client(
+            base_url=base_url.rstrip("/"),
+            timeout=timeout,
+            headers={"Authorization": f"token {token}"},
+            transport=transport,
+        )
+
+    def create_issue(
+        self,
+        *,
+        owner: str,
+        repo: str,
+        title: str,
+        body: str,
+        labels: list[int] | None = None,
+    ) -> CreatedIssue:
+        """Create a Gitea issue."""
+        payload: dict[str, object] = {"title": title, "body": body, "labels": labels or []}
+        response = self._request(
+            "POST",
+            f"/api/v1/repos/{owner}/{repo}/issues",
+            expected_statuses={EXPECTED_CREATED},
+            json=payload,
+        )
+        data = response.json()
+        return CreatedIssue(
+            number=_optional_int(data.get("number")),
+            html_url=_optional_str(data.get("html_url")),
+            title=str(data.get("title", title)),
+        )
+
+    def resolve_label_ids(self, *, owner: str, repo: str, labels: list[str]) -> list[int]:
+        """Resolve label names to Gitea label IDs."""
+        if not labels:
+            return []
+
+        available_labels: dict[str, int] = {}
+        page = 1
+        while True:
+            response = self._request(
+                "GET",
+                f"/api/v1/repos/{owner}/{repo}/labels",
+                params={"page": page, "limit": DEFAULT_PAGE_SIZE},
+            )
+            batch = response.json()
+            if not batch:
+                break
+            for label in batch:
+                label_name = str(label.get("name", ""))
+                label_id = _optional_int(label.get("id"))
+                if label_name and label_id is not None:
+                    available_labels[label_name] = label_id
+            if len(batch) < DEFAULT_PAGE_SIZE:
+                break
+            page += 1
+
+        missing = [label for label in labels if label not in available_labels]
+        if missing:
+            missing_names = ", ".join(sorted(missing))
+            msg = f"Missing Gitea labels: {missing_names}"
+            raise GiteaError(msg)
+
+        return [available_labels[label] for label in labels]
+
+    def list_open_pull_requests(
+        self,
+        *,
+        owner: str,
+        repo: str,
+        labels: list[str] | None = None,
+        head: str | None = None,
+    ) -> list[PullRequest]:
+        """List open pull requests for a repository."""
+        expected_labels = set(labels or [])
+        pull_requests: list[PullRequest] = []
+        page = 1
+        while True:
+            response = self._request(
+                "GET",
+                f"/api/v1/repos/{owner}/{repo}/pulls",
+                params={"state": "open", "page": page, "limit": DEFAULT_PAGE_SIZE},
+            )
+            batch = response.json()
+            if not batch:
+                break
+
+            for item in batch:
+                pull_request = _pull_request_from_api(item)
+                if head and pull_request.head_branch != head:
+                    continue
+                if expected_labels and not expected_labels.issubset(set(pull_request.labels)):
+                    continue
+                pull_requests.append(pull_request)
+
+            if len(batch) < DEFAULT_PAGE_SIZE:
+                break
+            page += 1
+
+        return pull_requests
+
+    def create_pull_request(
+        self,
+        *,
+        owner: str,
+        repo: str,
+        title: str,
+        body: str,
+        head: str,
+        base: str,
+        labels: list[str] | None = None,
+    ) -> PullRequest:
+        """Create a pull request."""
+        payload: dict[str, object] = {
+            "title": title,
+            "body": body,
+            "head": head,
+            "base": base,
+        }
+        if labels:
+            payload["labels"] = self.resolve_label_ids(owner=owner, repo=repo, labels=labels)
+
+        response = self._request(
+            "POST",
+            f"/api/v1/repos/{owner}/{repo}/pulls",
+            expected_statuses={EXPECTED_CREATED},
+            json=payload,
+        )
+        return _pull_request_from_api(response.json())
+
+    def merge_pull_request(
+        self,
+        *,
+        owner: str,
+        repo: str,
+        number: int,
+        merge_method: str = "rebase",
+        head_commit_id: str | None = None,
+        delete_branch_after_merge: bool = False,
+    ) -> None:
+        """Merge a pull request."""
+        payload: dict[str, object] = {
+            "Do": merge_method,
+            "delete_branch_after_merge": delete_branch_after_merge,
+        }
+        if head_commit_id:
+            payload["head_commit_id"] = head_commit_id
+
+        self._request(
+            "POST",
+            f"/api/v1/repos/{owner}/{repo}/pulls/{number}/merge",
+            json=payload,
+        )
+
+    def dispatch_workflow(self, *, owner: str, repo: str, workflow_id: str, ref: str) -> None:
+        """Trigger a workflow_dispatch run."""
+        workflow_path = quote(workflow_id, safe="")
+        self._request(
+            "POST",
+            f"/api/v1/repos/{owner}/{repo}/actions/workflows/{workflow_path}/dispatches",
+            expected_statuses={EXPECTED_OK, EXPECTED_NO_CONTENT},
+            json={"ref": ref},
+        )
+
+    def list_run_jobs(self, *, owner: str, repo: str, run_id: str | int) -> list[WorkflowJob]:
+        """List workflow jobs for a specific run."""
+        jobs: list[WorkflowJob] = []
+        page = 1
+        while True:
+            response = self._request(
+                "GET",
+                f"/api/v1/repos/{owner}/{repo}/actions/jobs",
+                params={"page": page, "limit": DEFAULT_PAGE_SIZE},
+            )
+            payload = response.json()
+            batch = payload.get("jobs", [])
+            if not batch:
+                break
+
+            for item in batch:
+                if str(item.get("run_id")) != str(run_id):
+                    continue
+                jobs.append(_workflow_job_from_api(item))
+
+            if len(batch) < DEFAULT_PAGE_SIZE:
+                break
+            page += 1
+
+        return jobs
+
+    def download_job_logs(self, *, owner: str, repo: str, job_id: int) -> str:
+        """Download logs for a workflow job."""
+        response = self._request(
+            "GET",
+            f"/api/v1/repos/{owner}/{repo}/actions/jobs/{job_id}/logs",
+        )
+        return response.text
+
+    def close(self) -> None:
+        """Close the underlying HTTP client."""
+        self._client.close()
+
+    def __enter__(self) -> Self:
+        """Enter the context manager."""
+        return self
+
+    def __exit__(self, *args: object) -> None:
+        """Close the HTTP client."""
+        self.close()
+
+    def _request(
+        self,
+        method: str,
+        path: str,
+        *,
+        expected_statuses: set[int] | None = None,
+        **kwargs: object,
+    ) -> httpx.Response:
+        """Send an HTTP request and validate the response status."""
+        response = self._client.request(method, path, **kwargs)
+        statuses = expected_statuses or {EXPECTED_OK}
+        if response.status_code not in statuses:
+            msg = f"Gitea request failed ({response.status_code}): {response.text}"
+            raise GiteaError(msg)
+        return response
+
+
+def _pull_request_from_api(data: dict[str, object]) -> PullRequest:
+    """Convert Gitea API pull-request data into a dataclass."""
+    number = _optional_int(data.get("number")) or _optional_int(data.get("index"))
+    if number is None:
+        msg = "Gitea pull request payload is missing a number"
+        raise GiteaError(msg)
+
+    labels = tuple(str(label.get("name", "")) for label in data.get("labels", []))
+    head = data.get("head", {})
+    base = data.get("base", {})
+    return PullRequest(
+        number=number,
+        title=str(data.get("title", "")),
+        html_url=_optional_str(data.get("html_url")),
+        labels=tuple(label for label in labels if label),
+        head_branch=_optional_str(head.get("ref")) or _optional_str(data.get("head_branch")),
+        base_branch=_optional_str(base.get("ref")) or _optional_str(data.get("base_branch")),
+    )
+
+
+def _workflow_job_from_api(data: dict[str, object]) -> WorkflowJob:
+    """Convert Gitea API workflow-job data into a dataclass."""
+    job_id = _optional_int(data.get("id"))
+    if job_id is None:
+        msg = "Gitea workflow job payload is missing an ID"
+        raise GiteaError(msg)
+
+    return WorkflowJob(
+        id=job_id,
+        name=str(data.get("name", "")),
+        run_id=_optional_int(data.get("run_id")),
+        status=_optional_str(data.get("status")),
+        conclusion=_optional_str(data.get("conclusion")),
+    )
+
+
+def _optional_int(value: object) -> int | None:
+    """Convert an API value to an integer when present."""
+    if value is None:
+        return None
+    return int(value)
+
+
+def _optional_str(value: object) -> str | None:
+    """Convert an API value to a string when present."""
+    if value is None:
+        return None
+    return str(value)

python/gitea_flake_lock.py

@@ -0,0 +1,148 @@
+"""Automation helpers for flake.lock pull requests on Gitea."""
+
+from __future__ import annotations
+
+import subprocess
+from os import getenv
+from typing import Annotated
+
+import typer
+
+from python.gitea import GiteaClient, PullRequest, split_repo_name
+
+DEFAULT_BASE_BRANCH = "main"
+DEFAULT_BRANCH = "automation/update-flake-lock"
+DEFAULT_GITEA_URL = "https://gitea.tmmworkshop.com"
+PR_LABELS = ["dependencies", "automated", "flake_lock_update"]
+PR_CHECK_WORKFLOWS = ["build_systems.yml", "treefmt.yml", "pytest.yml"]
+PR_TITLE = "Update flake.lock"
+PR_BODY = "Automated flake.lock update."
+
+app = typer.Typer(add_completion=False)
+
+
+def run_cmd(cmd: list[str], *, check: bool = True) -> subprocess.CompletedProcess[str]:
+    """Run a subprocess command."""
+    return subprocess.run(cmd, capture_output=True, text=True, check=check)
+
+
+def ensure_flake_lock_pull_request(
+    client: GiteaClient,
+    *,
+    owner: str,
+    repo: str,
+    branch: str,
+    base: str,
+) -> PullRequest:
+    """Return an existing flake.lock PR for the branch or create one."""
+    pull_requests = client.list_open_pull_requests(owner=owner, repo=repo, head=branch)
+    if pull_requests:
+        return pull_requests[0]
+
+    return client.create_pull_request(
+        owner=owner,
+        repo=repo,
+        title=PR_TITLE,
+        body=PR_BODY,
+        head=branch,
+        base=base,
+        labels=PR_LABELS,
+    )
+
+
+def find_flake_lock_pull_request(client: GiteaClient, *, owner: str, repo: str) -> PullRequest | None:
+    """Find the first open flake.lock pull request."""
+    pull_requests = client.list_open_pull_requests(owner=owner, repo=repo, labels=["flake_lock_update"])
+    if not pull_requests:
+        return None
+    return pull_requests[0]
+
+
+def dispatch_pull_request_checks(client: GiteaClient, *, owner: str, repo: str, branch: str) -> None:
+    """Dispatch the workflows that normally run for pull requests."""
+    for workflow in PR_CHECK_WORKFLOWS:
+        client.dispatch_workflow(owner=owner, repo=repo, workflow_id=workflow, ref=branch)
+
+
+def has_worktree_changes() -> bool:
+    """Return whether `flake.lock` has worktree changes."""
+    result = run_cmd(["git", "diff", "--quiet", "--", "flake.lock"], check=False)
+    return result.returncode != 0
+
+
+def commit_flake_lock_update(*, branch: str) -> None:
+    """Commit the updated lock file to the automation branch."""
+    run_cmd(["git", "config", "user.name", "gitea-actions[bot]"])
+    run_cmd(["git", "config", "user.email", "gitea-actions@tmmworkshop.com"])
+    run_cmd(["git", "checkout", "-B", branch])
+    run_cmd(["git", "add", "flake.lock"])
+    run_cmd(["git", "commit", "-m", "chore: update flake.lock"])
+
+
+def push_branch(*, branch: str) -> None:
+    """Push the automation branch to origin."""
+    run_cmd(["git", "push", "origin", f"HEAD:{branch}", "--force"])
+
+
+def _required_gitea_token() -> str:
+    """Read the required Gitea token from the environment."""
+    token = getenv("GITEA_TOKEN")
+    if token:
+        return token
+
+    msg = "GITEA_TOKEN environment variable is required"
+    raise RuntimeError(msg)
+
+
+@app.command()
+def update(
+    repo: Annotated[str, typer.Option("--repo", help="Gitea repository in owner/repo form")],
+    base: Annotated[str, typer.Option("--base", help="Base branch")] = DEFAULT_BASE_BRANCH,
+    branch: Annotated[str, typer.Option("--branch", help="Automation branch")] = DEFAULT_BRANCH,
+) -> None:
+    """Commit flake.lock changes and ensure a pull request exists."""
+    if not has_worktree_changes():
+        typer.echo("No flake.lock changes detected")
+        return
+
+    commit_flake_lock_update(branch=branch)
+    push_branch(branch=branch)
+
+    owner, repo_name = split_repo_name(repo)
+    with GiteaClient(
+        base_url=getenv("GITEA_URL", DEFAULT_GITEA_URL),
+        token=_required_gitea_token(),
+    ) as client:
+        pull_request = ensure_flake_lock_pull_request(
+            client,
+            owner=owner,
+            repo=repo_name,
+            branch=branch,
+            base=base,
+        )
+        # We can remove this if Gitea fixes the following issue:
+        # https://github.com/go-gitea/gitea/issues/33963
+        dispatch_pull_request_checks(client, owner=owner, repo=repo_name, branch=branch)
+    typer.echo(pull_request.html_url or f"Pull request #{pull_request.number}")
+
+
+@app.command()
+def merge(
+    repo: Annotated[str, typer.Option("--repo", help="Gitea repository in owner/repo form")],
+) -> None:
+    """Merge the first open flake.lock pull request."""
+    owner, repo_name = split_repo_name(repo)
+    with GiteaClient(
+        base_url=getenv("GITEA_URL", DEFAULT_GITEA_URL),
+        token=_required_gitea_token(),
+    ) as client:
+        pull_request = find_flake_lock_pull_request(client, owner=owner, repo=repo_name)
+        if not pull_request:
+            typer.echo("No open PR found with label flake_lock_update")
+            return
+        client.merge_pull_request(owner=owner, repo=repo_name, number=pull_request.number, merge_method="rebase")
+    typer.echo(f"Merged PR #{pull_request.number}")
+
+
+if __name__ == "__main__":
+    app()

python/orm/__init__.py

@@ -1,13 +1,9 @@
 """ORM package exports."""
 
-from python.orm.data_science_dev.base import DataScienceDevBase
 from python.orm.richie.base import RichieBase
-from python.orm.signal_bot.base import SignalBotBase
 from python.orm.van_inventory.base import VanInventoryBase
 
 __all__ = [
-    "DataScienceDevBase",
     "RichieBase",
-    "SignalBotBase",
     "VanInventoryBase",
 ]

python/orm/data_science_dev/__init__.py

@@ -1,11 +0,0 @@
-"""Data science dev database ORM exports."""
-
-from __future__ import annotations
-
-from python.orm.data_science_dev.base import DataScienceDevBase, DataScienceDevTableBase, DataScienceDevTableBaseBig
-
-__all__ = [
-    "DataScienceDevBase",
-    "DataScienceDevTableBase",
-    "DataScienceDevTableBaseBig",
-]

python/orm/data_science_dev/base.py

@@ -1,52 +0,0 @@
-"""Data science dev database ORM base."""
-
-from __future__ import annotations
-
-from datetime import datetime
-
-from sqlalchemy import BigInteger, DateTime, MetaData, func
-from sqlalchemy.ext.declarative import AbstractConcreteBase
-from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column
-
-from python.orm.common import NAMING_CONVENTION
-
-
-class DataScienceDevBase(DeclarativeBase):
-    """Base class for data_science_dev database ORM models."""
-
-    schema_name = "main"
-
-    metadata = MetaData(
-        schema=schema_name,
-        naming_convention=NAMING_CONVENTION,
-    )
-
-
-class _TableMixin:
-    """Shared timestamp columns for all table bases."""
-
-    created: Mapped[datetime] = mapped_column(
-        DateTime(timezone=True),
-        server_default=func.now(),
-    )
-    updated: Mapped[datetime] = mapped_column(
-        DateTime(timezone=True),
-        server_default=func.now(),
-        onupdate=func.now(),
-    )
-
-
-class DataScienceDevTableBase(_TableMixin, AbstractConcreteBase, DataScienceDevBase):
-    """Table with Integer primary key."""
-
-    __abstract__ = True
-
-    id: Mapped[int] = mapped_column(primary_key=True)
-
-
-class DataScienceDevTableBaseBig(_TableMixin, AbstractConcreteBase, DataScienceDevBase):
-    """Table with BigInteger primary key."""
-
-    __abstract__ = True
-
-    id: Mapped[int] = mapped_column(BigInteger, primary_key=True)

python/orm/data_science_dev/models.py

@@ -1,10 +0,0 @@
-"""Data science dev database ORM models."""
-
-from __future__ import annotations
-
-from python.orm.data_science_dev.posts import partitions  # noqa: F401 — registers partition classes in metadata
-from python.orm.data_science_dev.posts.tables import Posts
-
-__all__ = [
-    "Posts",
-]

python/orm/data_science_dev/posts/__init__.py

@@ -1,11 +0,0 @@
-"""Posts module — weekly-partitioned posts table and partition ORM models."""
-
-from __future__ import annotations
-
-from python.orm.data_science_dev.posts.failed_ingestion import FailedIngestion
-from python.orm.data_science_dev.posts.tables import Posts
-
-__all__ = [
-    "FailedIngestion",
-    "Posts",
-]

python/orm/data_science_dev/posts/columns.py

@@ -1,33 +0,0 @@
-"""Shared column definitions for the posts partitioned table family."""
-
-from __future__ import annotations
-
-from datetime import datetime
-
-from sqlalchemy import BigInteger, SmallInteger, Text
-from sqlalchemy.orm import Mapped, mapped_column
-
-
-class PostsColumns:
-    """Mixin providing all posts columns. Used by both the parent table and partitions."""
-
-    post_id: Mapped[int] = mapped_column(BigInteger, primary_key=True)
-    user_id: Mapped[int] = mapped_column(BigInteger)
-    instance: Mapped[str]
-    date: Mapped[datetime] = mapped_column(primary_key=True)
-    text: Mapped[str] = mapped_column(Text)
-    langs: Mapped[str | None]
-    like_count: Mapped[int]
-    reply_count: Mapped[int]
-    repost_count: Mapped[int]
-    reply_to: Mapped[int | None] = mapped_column(BigInteger)
-    replied_author: Mapped[int | None] = mapped_column(BigInteger)
-    thread_root: Mapped[int | None] = mapped_column(BigInteger)
-    thread_root_author: Mapped[int | None] = mapped_column(BigInteger)
-    repost_from: Mapped[int | None] = mapped_column(BigInteger)
-    reposted_author: Mapped[int | None] = mapped_column(BigInteger)
-    quotes: Mapped[int | None] = mapped_column(BigInteger)
-    quoted_author: Mapped[int | None] = mapped_column(BigInteger)
-    labels: Mapped[str | None]
-    sent_label: Mapped[int | None] = mapped_column(SmallInteger)
-    sent_score: Mapped[float | None]

python/orm/data_science_dev/posts/failed_ingestion.py

@@ -1,17 +0,0 @@
-"""Table for storing JSONL lines that failed during post ingestion."""
-
-from __future__ import annotations
-
-from sqlalchemy import Text
-from sqlalchemy.orm import Mapped, mapped_column
-
-from python.orm.data_science_dev.base import DataScienceDevTableBase
-
-
-class FailedIngestion(DataScienceDevTableBase):
-    """Stores raw JSONL lines and their error messages when ingestion fails."""
-
-    __tablename__ = "failed_ingestion"
-
-    raw_line: Mapped[str] = mapped_column(Text)
-    error: Mapped[str] = mapped_column(Text)

python/orm/data_science_dev/posts/partitions.py

@@ -1,71 +0,0 @@
-"""Dynamically generated ORM classes for each weekly partition of the posts table.
-
-Each class maps to a PostgreSQL partition table (e.g. posts_2024_01).
-These are real ORM models tracked by Alembic autogenerate.
-
-Uses ISO week numbering (datetime.isocalendar().week). ISO years can have
-52 or 53 weeks, and week boundaries are always Monday to Monday.
-"""
-
-from __future__ import annotations
-
-import sys
-from datetime import UTC, datetime
-
-from python.orm.data_science_dev.base import DataScienceDevBase
-from python.orm.data_science_dev.posts.columns import PostsColumns
-
-PARTITION_START_YEAR = 2023
-PARTITION_END_YEAR = 2026
-
-_current_module = sys.modules[__name__]
-
-
-def iso_weeks_in_year(year: int) -> int:
-    """Return the number of ISO weeks in a given year (52 or 53)."""
-    dec_28 = datetime(year, 12, 28, tzinfo=UTC)
-    return dec_28.isocalendar().week
-
-
-def week_bounds(year: int, week: int) -> tuple[datetime, datetime]:
-    """Return (start, end) datetimes for an ISO week.
-
-    Start = Monday 00:00:00 UTC of the given ISO week.
-    End   = Monday 00:00:00 UTC of the following ISO week.
-    """
-    start = datetime.fromisocalendar(year, week, 1).replace(tzinfo=UTC)
-    if week < iso_weeks_in_year(year):
-        end = datetime.fromisocalendar(year, week + 1, 1).replace(tzinfo=UTC)
-    else:
-        end = datetime.fromisocalendar(year + 1, 1, 1).replace(tzinfo=UTC)
-    return start, end
-
-
-def _build_partition_classes() -> dict[str, type]:
-    """Generate one ORM class per ISO week partition."""
-    classes: dict[str, type] = {}
-
-    for year in range(PARTITION_START_YEAR, PARTITION_END_YEAR + 1):
-        for week in range(1, iso_weeks_in_year(year) + 1):
-            class_name = f"PostsWeek{year}W{week:02d}"
-            table_name = f"posts_{year}_{week:02d}"
-
-            partition_class = type(
-                class_name,
-                (PostsColumns, DataScienceDevBase),
-                {
-                    "__tablename__": table_name,
-                    "__table_args__": ({"implicit_returning": False},),
-                },
-            )
-
-            classes[class_name] = partition_class
-
-    return classes
-
-
-# Generate all partition classes and register them on this module
-_partition_classes = _build_partition_classes()
-for _name, _cls in _partition_classes.items():
-    setattr(_current_module, _name, _cls)
-__all__ = list(_partition_classes.keys())

python/orm/data_science_dev/posts/tables.py

@@ -1,13 +0,0 @@
-"""Posts parent table with PostgreSQL weekly range partitioning on date column."""
-
-from __future__ import annotations
-
-from python.orm.data_science_dev.base import DataScienceDevBase
-from python.orm.data_science_dev.posts.columns import PostsColumns
-
-
-class Posts(PostsColumns, DataScienceDevBase):
-    """Parent partitioned table for posts, partitioned by week on `date`."""
-
-    __tablename__ = "posts"
-    __table_args__ = ({"postgresql_partition_by": "RANGE (date)"},)

python/orm/richie/__init__.py

@@ -2,8 +2,8 @@
 
 from __future__ import annotations
 
+from python.orm.richie.audiobook import Audiobook, AudiobookAuthor, AudiobookSeries
 from python.orm.richie.base import RichieBase, TableBase, TableBaseBig, TableBaseSmall
-from python.orm.richie.congress import Bill, Legislator, Vote, VoteRecord
 from python.orm.richie.contact import (
     Contact,
     ContactNeed,
@@ -13,17 +13,16 @@ from python.orm.richie.contact import (
 )
 
 __all__ = [
-    "Bill",
+    "Audiobook",
+    "AudiobookAuthor",
+    "AudiobookSeries",
     "Contact",
     "ContactNeed",
     "ContactRelationship",
-    "Legislator",
     "Need",
     "RelationshipType",
     "RichieBase",
     "TableBase",
     "TableBaseBig",
     "TableBaseSmall",
-    "Vote",
-    "VoteRecord",
 ]

python/orm/richie/audiobook.py

@@ -0,0 +1,55 @@
+"""Audiobook catalog models."""
+
+from __future__ import annotations
+
+from sqlalchemy import ForeignKey, String, UniqueConstraint
+from sqlalchemy.orm import Mapped, mapped_column, relationship
+
+from python.orm.richie.base import TableBase
+
+
+class AudiobookAuthor(TableBase):
+    """Canonical audiobook author."""
+
+    __tablename__ = "audiobook_author"
+    __table_args__ = (UniqueConstraint("name"),)
+
+    name: Mapped[str] = mapped_column(String, unique=True)
+
+    books: Mapped[list[Audiobook]] = relationship("Audiobook", back_populates="author")
+    series: Mapped[list[AudiobookSeries]] = relationship("AudiobookSeries", back_populates="author")
+
+
+class AudiobookSeries(TableBase):
+    """Canonical audiobook series."""
+
+    __tablename__ = "audiobook_series"
+    __table_args__ = (UniqueConstraint("author_id", "name"),)
+
+    name: Mapped[str] = mapped_column(String)
+    author_id: Mapped[int] = mapped_column(ForeignKey("main.audiobook_author.id", ondelete="CASCADE"))
+
+    author: Mapped[AudiobookAuthor] = relationship("AudiobookAuthor", back_populates="series")
+    books: Mapped[list[Audiobook]] = relationship("Audiobook", back_populates="series")
+
+
+class Audiobook(TableBase):
+    """Canonical audiobook title."""
+
+    __tablename__ = "audiobook"
+    __table_args__ = (
+        UniqueConstraint(
+            "author_id",
+            "series_id",
+            "title",
+            postgresql_nulls_not_distinct=True,
+        ),
+    )
+
+    title: Mapped[str] = mapped_column(String)
+    author_id: Mapped[int] = mapped_column(ForeignKey("main.audiobook_author.id", ondelete="CASCADE"))
+    series_id: Mapped[int | None] = mapped_column(ForeignKey("main.audiobook_series.id", ondelete="SET NULL"))
+    series_index: Mapped[float] = mapped_column(default=0.0)
+
+    author: Mapped[AudiobookAuthor] = relationship("AudiobookAuthor", back_populates="books")
+    series: Mapped[AudiobookSeries | None] = relationship("AudiobookSeries", back_populates="books")

python/orm/richie/congress.py

@@ -1,150 +0,0 @@
-"""Congress Tracker database models."""
-
-from __future__ import annotations
-
-from datetime import date
-
-from sqlalchemy import ForeignKey, Index, Text, UniqueConstraint
-from sqlalchemy.orm import Mapped, mapped_column, relationship
-
-from python.orm.richie.base import RichieBase, TableBase
-
-
-class Legislator(TableBase):
-    """Legislator model - members of Congress."""
-
-    __tablename__ = "legislator"
-
-    # Natural key - bioguide ID is the authoritative identifier
-    bioguide_id: Mapped[str] = mapped_column(Text, unique=True, index=True)
-
-    # Other IDs for cross-referencing
-    thomas_id: Mapped[str | None]
-    lis_id: Mapped[str | None]
-    govtrack_id: Mapped[int | None]
-    opensecrets_id: Mapped[str | None]
-    fec_ids: Mapped[str | None]  # JSON array stored as string
-
-    # Name info
-    first_name: Mapped[str]
-    last_name: Mapped[str]
-    official_full_name: Mapped[str | None]
-    nickname: Mapped[str | None]
-
-    # Bio
-    birthday: Mapped[date | None]
-    gender: Mapped[str | None]  # M/F
-
-    # Current term info (denormalized for query efficiency)
-    current_party: Mapped[str | None]
-    current_state: Mapped[str | None]
-    current_district: Mapped[int | None]  # House only
-    current_chamber: Mapped[str | None]  # rep/sen
-
-    # Relationships
-    vote_records: Mapped[list[VoteRecord]] = relationship(
-        "VoteRecord",
-        back_populates="legislator",
-        cascade="all, delete-orphan",
-    )
-
-
-class Bill(TableBase):
-    """Bill model - legislation introduced in Congress."""
-
-    __tablename__ = "bill"
-
-    # Composite natural key: congress + bill_type + number
-    congress: Mapped[int]
-    bill_type: Mapped[str]  # hr, s, hres, sres, hjres, sjres
-    number: Mapped[int]
-
-    # Bill info
-    title: Mapped[str | None]
-    title_short: Mapped[str | None]
-    official_title: Mapped[str | None]
-
-    # Status
-    status: Mapped[str | None]
-    status_at: Mapped[date | None]
-
-    # Sponsor
-    sponsor_bioguide_id: Mapped[str | None]
-
-    # Subjects
-    subjects_top_term: Mapped[str | None]
-
-    # Relationships
-    votes: Mapped[list[Vote]] = relationship(
-        "Vote",
-        back_populates="bill",
-    )
-
-    __table_args__ = (
-        UniqueConstraint("congress", "bill_type", "number", name="uq_bill_congress_type_number"),
-        Index("ix_bill_congress", "congress"),
-    )
-
-
-class Vote(TableBase):
-    """Vote model - roll call votes in Congress."""
-
-    __tablename__ = "vote"
-
-    # Composite natural key: congress + chamber + session + number
-    congress: Mapped[int]
-    chamber: Mapped[str]  # house/senate
-    session: Mapped[int]
-    number: Mapped[int]
-
-    # Vote details
-    vote_type: Mapped[str | None]
-    question: Mapped[str | None]
-    result: Mapped[str | None]
-    result_text: Mapped[str | None]
-
-    # Timing
-    vote_date: Mapped[date]
-
-    # Vote counts (denormalized for efficiency)
-    yea_count: Mapped[int | None]
-    nay_count: Mapped[int | None]
-    not_voting_count: Mapped[int | None]
-    present_count: Mapped[int | None]
-
-    # Related bill (optional - not all votes are on bills)
-    bill_id: Mapped[int | None] = mapped_column(ForeignKey("main.bill.id"))
-
-    # Relationships
-    bill: Mapped[Bill | None] = relationship("Bill", back_populates="votes")
-    vote_records: Mapped[list[VoteRecord]] = relationship(
-        "VoteRecord",
-        back_populates="vote",
-        cascade="all, delete-orphan",
-    )
-
-    __table_args__ = (
-        UniqueConstraint("congress", "chamber", "session", "number", name="uq_vote_congress_chamber_session_number"),
-        Index("ix_vote_date", "vote_date"),
-        Index("ix_vote_congress_chamber", "congress", "chamber"),
-    )
-
-
-class VoteRecord(RichieBase):
-    """Association table: Vote <-> Legislator with position."""
-
-    __tablename__ = "vote_record"
-
-    vote_id: Mapped[int] = mapped_column(
-        ForeignKey("main.vote.id", ondelete="CASCADE"),
-        primary_key=True,
-    )
-    legislator_id: Mapped[int] = mapped_column(
-        ForeignKey("main.legislator.id", ondelete="CASCADE"),
-        primary_key=True,
-    )
-    position: Mapped[str]  # Yea, Nay, Not Voting, Present
-
-    # Relationships
-    vote: Mapped[Vote] = relationship("Vote", back_populates="vote_records")
-    legislator: Mapped[Legislator] = relationship("Legislator", back_populates="vote_records")

python/orm/signal_bot/__init__.py

@@ -1,16 +0,0 @@
-"""Signal bot database ORM exports."""
-
-from __future__ import annotations
-
-from python.orm.signal_bot.base import SignalBotBase, SignalBotTableBase, SignalBotTableBaseSmall
-from python.orm.signal_bot.models import DeadLetterMessage, DeviceRole, RoleRecord, SignalDevice
-
-__all__ = [
-    "DeadLetterMessage",
-    "DeviceRole",
-    "RoleRecord",
-    "SignalBotBase",
-    "SignalBotTableBase",
-    "SignalBotTableBaseSmall",
-    "SignalDevice",
-]

python/orm/signal_bot/base.py

@@ -1,52 +0,0 @@
-"""Signal bot database ORM base."""
-
-from __future__ import annotations
-
-from datetime import datetime
-
-from sqlalchemy import DateTime, MetaData, SmallInteger, func
-from sqlalchemy.ext.declarative import AbstractConcreteBase
-from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column
-
-from python.orm.common import NAMING_CONVENTION
-
-
-class SignalBotBase(DeclarativeBase):
-    """Base class for signal_bot database ORM models."""
-
-    schema_name = "main"
-
-    metadata = MetaData(
-        schema=schema_name,
-        naming_convention=NAMING_CONVENTION,
-    )
-
-
-class _TableMixin:
-    """Shared timestamp columns for all table bases."""
-
-    created: Mapped[datetime] = mapped_column(
-        DateTime(timezone=True),
-        server_default=func.now(),
-    )
-    updated: Mapped[datetime] = mapped_column(
-        DateTime(timezone=True),
-        server_default=func.now(),
-        onupdate=func.now(),
-    )
-
-
-class SignalBotTableBaseSmall(_TableMixin, AbstractConcreteBase, SignalBotBase):
-    """Table with SmallInteger primary key."""
-
-    __abstract__ = True
-
-    id: Mapped[int] = mapped_column(SmallInteger, primary_key=True)
-
-
-class SignalBotTableBase(_TableMixin, AbstractConcreteBase, SignalBotBase):
-    """Table with Integer primary key."""
-
-    __abstract__ = True
-
-    id: Mapped[int] = mapped_column(primary_key=True)

python/orm/signal_bot/models.py

@@ -1,62 +0,0 @@
-"""Signal bot device, role, and dead letter ORM models."""
-
-from __future__ import annotations
-
-from datetime import datetime
-
-from sqlalchemy import DateTime, Enum, ForeignKey, SmallInteger, String, Text, UniqueConstraint
-from sqlalchemy.orm import Mapped, mapped_column, relationship
-
-from python.orm.signal_bot.base import SignalBotTableBase, SignalBotTableBaseSmall
-from python.signal_bot.models import MessageStatus, TrustLevel
-
-
-class RoleRecord(SignalBotTableBaseSmall):
-    """Lookup table for RBAC roles, keyed by smallint."""
-
-    __tablename__ = "role"
-
-    name: Mapped[str] = mapped_column(String(50), unique=True)
-
-
-class DeviceRole(SignalBotTableBase):
-    """Association between a device and a role."""
-
-    __tablename__ = "device_role"
-    __table_args__ = (
-        UniqueConstraint("device_id", "role_id", name="uq_device_role_device_role"),
-        {"schema": "main"},
-    )
-
-    device_id: Mapped[int] = mapped_column(ForeignKey("main.signal_device.id"))
-    role_id: Mapped[int] = mapped_column(SmallInteger, ForeignKey("main.role.id"))
-
-
-class SignalDevice(SignalBotTableBase):
-    """A Signal device tracked by phone number and safety number."""
-
-    __tablename__ = "signal_device"
-
-    phone_number: Mapped[str] = mapped_column(String(50), unique=True)
-    safety_number: Mapped[str | None]
-    trust_level: Mapped[TrustLevel] = mapped_column(
-        Enum(TrustLevel, name="trust_level", create_constraint=False, native_enum=False),
-        default=TrustLevel.UNVERIFIED,
-    )
-    last_seen: Mapped[datetime] = mapped_column(DateTime(timezone=True))
-
-    roles: Mapped[list[RoleRecord]] = relationship(secondary=DeviceRole.__table__)
-
-
-class DeadLetterMessage(SignalBotTableBase):
-    """A Signal message that failed processing and was sent to the dead letter queue."""
-
-    __tablename__ = "dead_letter_message"
-
-    source: Mapped[str]
-    message: Mapped[str] = mapped_column(Text)
-    received_at: Mapped[datetime] = mapped_column(DateTime(timezone=True))
-    status: Mapped[MessageStatus] = mapped_column(
-        Enum(MessageStatus, name="message_status", create_constraint=False, native_enum=False),
-        default=MessageStatus.UNPROCESSED,
-    )

python/signal_bot/__init__.py

@@ -1 +0,0 @@
-"""Signal command and control bot."""

python/signal_bot/commands/__init__.py

@@ -1 +0,0 @@
-"""Signal bot commands."""

python/signal_bot/commands/inventory.py

@@ -1,137 +0,0 @@
-"""Van inventory command — parse receipts and item lists via LLM, push to API."""
-
-from __future__ import annotations
-
-import json
-import logging
-from typing import TYPE_CHECKING, Any
-
-import httpx
-
-from python.signal_bot.models import InventoryItem, InventoryUpdate
-
-if TYPE_CHECKING:
-    from python.signal_bot.llm_client import LLMClient
-    from python.signal_bot.models import SignalMessage
-    from python.signal_bot.signal_client import SignalClient
-
-logger = logging.getLogger(__name__)
-
-SYSTEM_PROMPT = """\
-You are an inventory assistant. Extract items from the input and return ONLY
-a JSON array. Each element must have these fields:
-  - "name": item name (string)
-  - "quantity": numeric count or amount (default 1)
-  - "unit": unit of measure (e.g. "each", "lb", "oz", "gallon", "bag", "box")
-  - "category": category like "food", "tools", "supplies", etc.
-  - "notes": any extra detail (empty string if none)
-
-Example output:
-[{"name": "water bottles", "quantity": 6, "unit": "gallon", "category": "supplies", "notes": "1 gallon each"}]
-
-Return ONLY the JSON array, no other text.\
-"""
-
-IMAGE_PROMPT = "Extract all items from this receipt or inventory photo."
-TEXT_PROMPT = "Extract all items from this inventory list."
-
-
-def parse_llm_response(raw: str) -> list[InventoryItem]:
-    """Parse the LLM JSON response into InventoryItem list."""
-    text = raw.strip()
-    # Strip markdown code fences if present
-    if text.startswith("```"):
-        lines = text.split("\n")
-        lines = [line for line in lines if not line.startswith("```")]
-        text = "\n".join(lines)
-
-    items_data: list[dict[str, Any]] = json.loads(text)
-    return [InventoryItem.model_validate(item) for item in items_data]
-
-
-def _upsert_item(api_url: str, item: InventoryItem) -> None:
-    """Create or update an item via the van_inventory API.
-
-    Fetches existing items, and if one with the same name exists,
-    patches its quantity (summing). Otherwise creates a new item.
-    """
-    base = api_url.rstrip("/")
-    response = httpx.get(f"{base}/api/items", timeout=10)
-    response.raise_for_status()
-    existing: list[dict[str, Any]] = response.json()
-
-    match = next((e for e in existing if e["name"].lower() == item.name.lower()), None)
-
-    if match:
-        new_qty = match["quantity"] + item.quantity
-        patch = {"quantity": new_qty}
-        if item.category:
-            patch["category"] = item.category
-        response = httpx.patch(f"{base}/api/items/{match['id']}", json=patch, timeout=10)
-        response.raise_for_status()
-        return
-    payload = {
-        "name": item.name,
-        "quantity": item.quantity,
-        "unit": item.unit,
-        "category": item.category or None,
-    }
-    response = httpx.post(f"{base}/api/items", json=payload, timeout=10)
-    response.raise_for_status()
-
-
-def handle_inventory_update(
-    message: SignalMessage,
-    signal: SignalClient,
-    llm: LLMClient,
-    api_url: str,
-) -> InventoryUpdate:
-    """Process an inventory update from a Signal message.
-
-    Accepts either an image (receipt photo) or text list.
-    Uses the LLM to extract structured items, then pushes to the van_inventory API.
-    """
-    try:
-        logger.info(f"Processing inventory update from {message.source}")
-        if message.attachments:
-            image_data = signal.get_attachment(message.attachments[0])
-            raw_response = llm.chat(
-                IMAGE_PROMPT,
-                image_data=image_data,
-                system=SYSTEM_PROMPT,
-            )
-            source_type = "receipt_photo"
-        elif message.message.strip():
-            raw_response = llm.chat(
-                f"{TEXT_PROMPT}\n\n{message.message}",
-                system=SYSTEM_PROMPT,
-            )
-            source_type = "text_list"
-        else:
-            signal.reply(message, "Send a photo of a receipt or a text list of items to update inventory.")
-            return InventoryUpdate()
-
-        logger.info(f"{raw_response=}")
-
-        new_items = parse_llm_response(raw_response)
-
-        logger.info(f"{new_items=}")
-
-        for item in new_items:
-            _upsert_item(api_url, item)
-
-        summary = _format_summary(new_items)
-        signal.reply(message, f"Inventory updated with {len(new_items)} item(s):\n{summary}")
-
-        return InventoryUpdate(items=new_items, raw_response=raw_response, source_type=source_type)
-
-    except Exception:
-        logger.exception("Failed to process inventory update")
-        signal.reply(message, "Failed to process inventory update. Check logs for details.")
-        return InventoryUpdate()
-
-
-def _format_summary(items: list[InventoryItem]) -> str:
-    """Format items into a readable summary."""
-    lines = [f"  - {item.name} x{item.quantity} {item.unit} [{item.category}]" for item in items]
-    return "\n".join(lines)

python/signal_bot/commands/location.py

@@ -1,64 +0,0 @@
-"""Location command for the Signal bot."""
-
-from __future__ import annotations
-
-import logging
-from typing import TYPE_CHECKING, Any
-
-import httpx
-
-if TYPE_CHECKING:
-    from python.signal_bot.models import SignalMessage
-    from python.signal_bot.signal_client import SignalClient
-
-logger = logging.getLogger(__name__)
-
-
-def _get_entity_state(ha_url: str, ha_token: str, entity_id: str) -> dict[str, Any]:
-    """Fetch an entity's state from Home Assistant."""
-    entity_url = f"{ha_url}/api/states/{entity_id}"
-    logger.debug(f"Fetching {entity_url=}")
-    response = httpx.get(
-        entity_url,
-        headers={"Authorization": f"Bearer {ha_token}"},
-        timeout=30,
-    )
-    response.raise_for_status()
-    return response.json()
-
-
-def _format_location(latitude: str, longitude: str) -> str:
-    """Render a friendly location response."""
-    return f"Van location: {latitude}, {longitude}\nhttps://maps.google.com/?q={latitude},{longitude}"
-
-
-def handle_location_request(
-    message: SignalMessage,
-    signal: SignalClient,
-    ha_url: str | None,
-    ha_token: str | None,
-) -> None:
-    """Reply with van location from Home Assistant."""
-    if ha_url is None or ha_token is None:
-        signal.reply(message, "Location command is not configured (missing HA_URL or HA_TOKEN).")
-        return
-
-    lat_payload = None
-    lon_payload = None
-    try:
-        lat_payload = _get_entity_state(ha_url, ha_token, "sensor.van_last_known_latitude")
-        lon_payload = _get_entity_state(ha_url, ha_token, "sensor.van_last_known_longitude")
-    except httpx.HTTPError:
-        logger.exception("Couldn't fetch van location from Home Assistant right now.")
-        logger.debug(f"{ha_url=} {lat_payload=} {lon_payload=}")
-        signal.reply(message, "Couldn't fetch van location from Home Assistant right now.")
-        return
-
-    latitude = lat_payload.get("state", "")
-    longitude = lon_payload.get("state", "")
-
-    if not latitude or not longitude or latitude == "unavailable" or longitude == "unavailable":
-        signal.reply(message, "Van location is unavailable in Home Assistant right now.")
-        return
-
-    signal.reply(message, _format_location(latitude, longitude))

python/signal_bot/device_registry.py

@@ -1,286 +0,0 @@
-"""Device registry — tracks verified/unverified devices by safety number."""
-
-from __future__ import annotations
-
-import logging
-from datetime import datetime, timedelta
-from typing import TYPE_CHECKING, NamedTuple
-
-from sqlalchemy import delete, select
-from sqlalchemy.orm import Session
-
-from python.common import utcnow
-from python.orm.signal_bot.models import RoleRecord, SignalDevice
-from python.signal_bot.models import Role, TrustLevel
-
-if TYPE_CHECKING:
-    from sqlalchemy.engine import Engine
-
-    from python.signal_bot.signal_client import SignalClient
-
-logger = logging.getLogger(__name__)
-
-_BLOCKED_TTL = timedelta(minutes=60)
-_DEFAULT_TTL = timedelta(minutes=5)
-
-
-class _CacheEntry(NamedTuple):
-    expires: datetime
-    trust_level: TrustLevel
-    has_safety_number: bool
-    safety_number: str | None
-    roles: list[Role]
-
-
-class DeviceRegistry:
-    """Manage device trust based on Signal safety numbers.
-
-    Devices start as UNVERIFIED. An admin verifies them over SSH by calling
-    ``verify(phone_number)`` which marks the device VERIFIED and also tells
-    signal-cli to trust the identity.
-
-    Only VERIFIED devices may execute commands.
-    """
-
-    def __init__(self, signal_client: SignalClient, engine: Engine) -> None:
-        self.signal_client = signal_client
-        self.engine = engine
-        self._contact_cache: dict[str, _CacheEntry] = {}
-
-    def is_verified(self, phone_number: str) -> bool:
-        """Check if a phone number is verified."""
-        if entry := self._cached(phone_number):
-            return entry.trust_level == TrustLevel.VERIFIED
-        device = self._load_device(phone_number)
-        return device is not None and device.trust_level == TrustLevel.VERIFIED
-
-    def record_contact(self, phone_number: str, safety_number: str | None = None) -> None:
-        """Record seeing a device. Creates entry if new, updates last_seen."""
-        now = utcnow()
-
-        entry = self._cached(phone_number)
-        if entry and entry.safety_number == safety_number:
-            return
-
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if device:
-                if device.safety_number != safety_number and device.trust_level != TrustLevel.BLOCKED:
-                    logger.warning(f"Safety number changed for {phone_number}, resetting to UNVERIFIED")
-                    device.safety_number = safety_number
-                    device.trust_level = TrustLevel.UNVERIFIED
-                device.last_seen = now
-            else:
-                device = SignalDevice(
-                    phone_number=phone_number,
-                    safety_number=safety_number,
-                    trust_level=TrustLevel.UNVERIFIED,
-                    last_seen=now,
-                )
-                session.add(device)
-                logger.info(f"New device registered: {phone_number}")
-
-            session.commit()
-            self._update_cache(phone_number, device)
-
-    def has_safety_number(self, phone_number: str) -> bool:
-        """Check if a device has a safety number on file."""
-        if entry := self._cached(phone_number):
-            return entry.has_safety_number
-        device = self._load_device(phone_number)
-        return device is not None and device.safety_number is not None
-
-    def verify(self, phone_number: str) -> bool:
-        """Mark a device as verified. Called by admin over SSH.
-
-        Returns True if the device was found and verified.
-        """
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if not device:
-                logger.warning(f"Cannot verify unknown device: {phone_number}")
-                return False
-
-            device.trust_level = TrustLevel.VERIFIED
-            self.signal_client.trust_identity(phone_number, trust_all_known_keys=True)
-            session.commit()
-            self._update_cache(phone_number, device)
-            logger.info(f"Device verified: {phone_number}")
-            return True
-
-    def block(self, phone_number: str) -> bool:
-        """Block a device."""
-        return self._set_trust(phone_number, TrustLevel.BLOCKED, "Device blocked")
-
-    def unverify(self, phone_number: str) -> bool:
-        """Reset a device to unverified."""
-        return self._set_trust(phone_number, TrustLevel.UNVERIFIED)
-
-    # -- role management ------------------------------------------------------
-
-    def get_roles(self, phone_number: str) -> list[Role]:
-        """Return the roles for a device, defaulting to empty."""
-        if entry := self._cached(phone_number):
-            return entry.roles
-        device = self._load_device(phone_number)
-        return _extract_roles(device) if device else []
-
-    def has_role(self, phone_number: str, role: Role) -> bool:
-        """Check if a device has a specific role or is admin."""
-        roles = self.get_roles(phone_number)
-        return Role.ADMIN in roles or role in roles
-
-    def grant_role(self, phone_number: str, role: Role) -> bool:
-        """Add a role to a device. Called by admin over SSH."""
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if not device:
-                logger.warning(f"Cannot grant role for unknown device: {phone_number}")
-                return False
-
-            if any(record.name == role for record in device.roles):
-                return True
-
-            role_record = session.execute(select(RoleRecord).where(RoleRecord.name == role)).scalar_one_or_none()
-
-            if not role_record:
-                logger.warning(f"Unknown role: {role}")
-                return False
-
-            device.roles.append(role_record)
-            session.commit()
-            self._update_cache(phone_number, device)
-            logger.info(f"Device {phone_number} granted role {role}")
-            return True
-
-    def revoke_role(self, phone_number: str, role: Role) -> bool:
-        """Remove a role from a device. Called by admin over SSH."""
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if not device:
-                logger.warning(f"Cannot revoke role for unknown device: {phone_number}")
-                return False
-
-            device.roles = [record for record in device.roles if record.name != role]
-            session.commit()
-            self._update_cache(phone_number, device)
-            logger.info(f"Device {phone_number} revoked role {role}")
-            return True
-
-    def set_roles(self, phone_number: str, roles: list[Role]) -> bool:
-        """Replace all roles for a device. Called by admin over SSH."""
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if not device:
-                logger.warning(f"Cannot set roles for unknown device: {phone_number}")
-                return False
-
-            role_names = [str(role) for role in roles]
-            records = list(session.execute(select(RoleRecord).where(RoleRecord.name.in_(role_names))).scalars().all())
-            device.roles = records
-            session.commit()
-            self._update_cache(phone_number, device)
-            logger.info(f"Device {phone_number} roles set to {role_names}")
-            return True
-
-    # -- queries --------------------------------------------------------------
-
-    def list_devices(self) -> list[SignalDevice]:
-        """Return all known devices."""
-        with Session(self.engine) as session:
-            return list(session.execute(select(SignalDevice)).scalars().all())
-
-    def sync_identities(self) -> None:
-        """Pull identity list from signal-cli and record any new ones."""
-        identities = self.signal_client.get_identities()
-        for identity in identities:
-            number = identity.get("number", "")
-            safety = identity.get("safety_number", identity.get("fingerprint", ""))
-            if number:
-                self.record_contact(number, safety)
-
-    # -- internals ------------------------------------------------------------
-
-    def _cached(self, phone_number: str) -> _CacheEntry | None:
-        """Return the cache entry if it exists and hasn't expired."""
-        entry = self._contact_cache.get(phone_number)
-        if entry and utcnow() < entry.expires:
-            return entry
-        return None
-
-    def _load_device(self, phone_number: str) -> SignalDevice | None:
-        """Fetch a device by phone number (with joined roles)."""
-        with Session(self.engine) as session:
-            return session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-    def _update_cache(self, phone_number: str, device: SignalDevice) -> None:
-        """Refresh the cache entry for a device."""
-        ttl = _BLOCKED_TTL if device.trust_level == TrustLevel.BLOCKED else _DEFAULT_TTL
-        self._contact_cache[phone_number] = _CacheEntry(
-            expires=utcnow() + ttl,
-            trust_level=device.trust_level,
-            has_safety_number=device.safety_number is not None,
-            safety_number=device.safety_number,
-            roles=_extract_roles(device),
-        )
-
-    def _set_trust(self, phone_number: str, level: str, log_msg: str | None = None) -> bool:
-        """Update the trust level for a device."""
-        with Session(self.engine) as session:
-            device = session.execute(
-                select(SignalDevice).where(SignalDevice.phone_number == phone_number)
-            ).scalar_one_or_none()
-
-            if not device:
-                return False
-
-            device.trust_level = level
-            session.commit()
-            self._update_cache(phone_number, device)
-            if log_msg:
-                logger.info(f"{log_msg}: {phone_number}")
-            return True
-
-
-def _extract_roles(device: SignalDevice) -> list[Role]:
-    """Convert a device's RoleRecord objects to a list of Role enums."""
-    return [Role(record.name) for record in device.roles]
-
-
-def sync_roles(engine: Engine) -> None:
-    """Sync the Role enum to the role table, adding new and removing stale entries."""
-    expected = {role.value for role in Role}
-
-    with Session(engine) as session:
-        existing = {record.name for record in session.execute(select(RoleRecord)).scalars().all()}
-
-        to_add = expected - existing
-        to_remove = existing - expected
-
-        for name in to_add:
-            session.add(RoleRecord(name=name))
-            logger.info(f"Role added: {name}")
-
-        if to_remove:
-            session.execute(delete(RoleRecord).where(RoleRecord.name.in_(to_remove)))
-            for name in to_remove:
-                logger.info(f"Role removed: {name}")
-
-        session.commit()

python/signal_bot/llm_client.py

@@ -1,80 +0,0 @@
-"""Flexible LLM client for ollama backends."""
-
-from __future__ import annotations
-
-import base64
-import logging
-from typing import Any, Self
-
-import httpx
-
-logger = logging.getLogger(__name__)
-
-
-class LLMClient:
-    """Talk to an ollama instance.
-
-    Args:
-        model: Ollama model name.
-        host: Ollama host.
-        port: Ollama port.
-        temperature: Sampling temperature.
-    """
-
-    def __init__(
-        self,
-        *,
-        model: str,
-        host: str,
-        port: int = 11434,
-        temperature: float = 0.1,
-        timeout: int = 300,
-    ) -> None:
-        self.model = model
-        self.temperature = temperature
-        self._client = httpx.Client(base_url=f"http://{host}:{port}", timeout=timeout)
-
-    def chat(self, prompt: str, image_data: bytes | None = None, system: str | None = None) -> str:
-        """Send a text prompt and return the response."""
-        messages: list[dict[str, Any]] = []
-        if system:
-            messages.append({"role": "system", "content": system})
-
-        user_msg = {"role": "user", "content": prompt}
-        if image_data:
-            user_msg["images"] = [base64.b64encode(image_data).decode()]
-
-        messages.append(user_msg)
-        return self._generate(messages)
-
-    def _generate(self, messages: list[dict[str, Any]]) -> str:
-        """Call the ollama chat API."""
-        payload = {
-            "model": self.model,
-            "messages": messages,
-            "stream": False,
-            "options": {"temperature": self.temperature},
-        }
-        logger.info(f"LLM request to {self.model}")
-        response = self._client.post("/api/chat", json=payload)
-        response.raise_for_status()
-        data = response.json()
-        return data["message"]["content"]
-
-    def list_models(self) -> list[str]:
-        """List available models on the ollama instance."""
-        response = self._client.get("/api/tags")
-        response.raise_for_status()
-        return [m["name"] for m in response.json().get("models", [])]
-
-    def __enter__(self) -> Self:
-        """Enter the context manager."""
-        return self
-
-    def __exit__(self, *args: object) -> None:
-        """Close the HTTP client on exit."""
-        self.close()
-
-    def close(self) -> None:
-        """Close the HTTP client."""
-        self._client.close()

python/signal_bot/main.py

@@ -1,239 +0,0 @@
-"""Signal command and control bot — main entry point."""
-
-from __future__ import annotations
-
-import logging
-from dataclasses import dataclass
-from os import getenv
-from typing import TYPE_CHECKING, Annotated
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-
-import typer
-from alembic.command import upgrade
-from sqlalchemy.orm import Session
-from tenacity import before_sleep_log, retry, stop_after_attempt, wait_exponential
-
-from python.common import configure_logger, utcnow
-from python.database_cli import DATABASES
-from python.orm.common import get_postgres_engine
-from python.orm.signal_bot.models import DeadLetterMessage
-from python.signal_bot.commands.inventory import handle_inventory_update
-from python.signal_bot.commands.location import handle_location_request
-from python.signal_bot.device_registry import DeviceRegistry, sync_roles
-from python.signal_bot.llm_client import LLMClient
-from python.signal_bot.models import BotConfig, MessageStatus, Role, SignalMessage
-from python.signal_bot.signal_client import SignalClient
-
-logger = logging.getLogger(__name__)
-
-
-@dataclass(frozen=True, slots=True)
-class Command:
-    """A registered bot command."""
-
-    action: Callable[[SignalMessage, str], None]
-    help_text: str
-    role: Role | None  # None = no role required (always allowed)
-
-
-class Bot:
-    """Holds shared resources and dispatches incoming messages to command handlers."""
-
-    def __init__(
-        self,
-        signal: SignalClient,
-        llm: LLMClient,
-        registry: DeviceRegistry,
-        config: BotConfig,
-    ) -> None:
-        self.signal = signal
-        self.llm = llm
-        self.registry = registry
-        self.config = config
-        self.commands: dict[str, Command] = {
-            "help": Command(action=self._help, help_text="show this help message", role=None),
-            "status": Command(action=self._status, help_text="show bot status", role=Role.STATUS),
-            "inventory": Command(
-                action=self._inventory,
-                help_text="update van inventory from a text list or receipt photo",
-                role=Role.INVENTORY,
-            ),
-            "location": Command(
-                action=self._location,
-                help_text="get current van location",
-                role=Role.LOCATION,
-            ),
-        }
-
-    # -- actions --------------------------------------------------------------
-
-    def _help(self, message: SignalMessage, _cmd: str) -> None:
-        """Return help text filtered to the sender's roles."""
-        self.signal.reply(message, self._build_help(self.registry.get_roles(message.source)))
-
-    def _status(self, message: SignalMessage, _cmd: str) -> None:
-        """Return the status of the bot."""
-        models = self.llm.list_models()
-        model_list = ", ".join(models[:10])
-        device_count = len(self.registry.list_devices())
-        self.signal.reply(
-            message,
-            f"Bot online.\nLLM: {self.llm.model}\nAvailable models: {model_list}\nKnown devices: {device_count}",
-        )
-
-    def _inventory(self, message: SignalMessage, _cmd: str) -> None:
-        """Process an inventory update."""
-        handle_inventory_update(message, self.signal, self.llm, self.config.inventory_api_url)
-
-    def _location(self, message: SignalMessage, _cmd: str) -> None:
-        """Reply with current van location."""
-        handle_location_request(message, self.signal, self.config.ha_url, self.config.ha_token)
-
-    # -- dispatch -------------------------------------------------------------
-
-    def _build_help(self, roles: list[Role]) -> str:
-        """Build help text showing only the commands the user can access."""
-        is_admin = Role.ADMIN in roles
-        lines = ["Available commands:"]
-        for name, cmd in self.commands.items():
-            if cmd.role is None or is_admin or cmd.role in roles:
-                lines.append(f"  {name:20s} — {cmd.help_text}")
-        return "\n".join(lines)
-
-    def dispatch(self, message: SignalMessage) -> None:
-        """Route an incoming message to the right command handler."""
-        source = message.source
-
-        if not self.registry.is_verified(source):
-            logger.info(f"Device {source} not verified, ignoring message")
-            return
-
-        if not self.registry.has_safety_number(source) and self.registry.has_role(source, Role.ADMIN):
-            logger.warning(f"Admin device {source} missing safety number, ignoring message")
-            return
-
-        text = message.message.strip()
-        parts = text.split()
-
-        if not parts and not message.attachments:
-            return
-
-        cmd = parts[0].lower() if parts else ""
-
-        logger.info(f"f{source=} running {cmd=} with {message=}")
-
-        command = self.commands.get(cmd)
-        if command is None:
-            if message.attachments:
-                command = self.commands["inventory"]
-                cmd = "inventory"
-            else:
-                return
-
-        if command.role is not None and not self.registry.has_role(source, command.role):
-            logger.warning(f"Device {source} denied access to {cmd!r}")
-            self.signal.reply(message, f"Permission denied: you do not have the '{command.role}' role.")
-            return
-
-        command.action(message, cmd)
-
-    def process_message(self, message: SignalMessage) -> None:
-        """Process a single message, sending it to the dead letter queue after repeated failures."""
-        max_attempts = self.config.max_message_attempts
-        for attempt in range(1, max_attempts + 1):
-            try:
-                safety_number = self.signal.get_safety_number(message.source)
-                self.registry.record_contact(message.source, safety_number)
-                self.dispatch(message)
-            except Exception:
-                logger.exception(f"Failed to process message (attempt {attempt}/{max_attempts})")
-            else:
-                return
-
-        logger.error(f"Message from {message.source} failed {max_attempts} times, sending to dead letter queue")
-        with Session(self.config.engine) as session:
-            session.add(
-                DeadLetterMessage(
-                    source=message.source,
-                    message=message.message,
-                    received_at=utcnow(),
-                    status=MessageStatus.UNPROCESSED,
-                )
-            )
-            session.commit()
-
-    def run(self) -> None:
-        """Listen for messages via WebSocket, reconnecting on failure."""
-        logger.info("Bot started — listening via WebSocket")
-
-        @retry(
-            stop=stop_after_attempt(self.config.max_retries),
-            wait=wait_exponential(multiplier=self.config.reconnect_delay, max=self.config.max_reconnect_delay),
-            before_sleep=before_sleep_log(logger, logging.WARNING),
-            reraise=True,
-        )
-        def _listen() -> None:
-            for message in self.signal.listen():
-                logger.info(f"Message from {message.source}: {message.message[:80]}")
-                self.process_message(message)
-
-        try:
-            _listen()
-        except Exception:
-            logger.critical("Max retries exceeded, shutting down")
-            raise
-
-
-def main(
-    log_level: Annotated[str, typer.Option()] = "DEBUG",
-    llm_timeout: Annotated[int, typer.Option()] = 600,
-) -> None:
-    """Run the Signal command and control bot."""
-    configure_logger(log_level)
-    signal_api_url = getenv("SIGNAL_API_URL")
-    phone_number = getenv("SIGNAL_PHONE_NUMBER")
-    inventory_api_url = getenv("INVENTORY_API_URL")
-
-    if signal_api_url is None:
-        error = "SIGNAL_API_URL environment variable not set"
-        raise ValueError(error)
-    if phone_number is None:
-        error = "SIGNAL_PHONE_NUMBER environment variable not set"
-        raise ValueError(error)
-    if inventory_api_url is None:
-        error = "INVENTORY_API_URL environment variable not set"
-        raise ValueError(error)
-
-    signal_bot_config = DATABASES["signal_bot"].alembic_config()
-    upgrade(signal_bot_config, "head")
-    engine = get_postgres_engine(name="SIGNALBOT")
-    sync_roles(engine)
-    config = BotConfig(
-        signal_api_url=signal_api_url,
-        phone_number=phone_number,
-        inventory_api_url=inventory_api_url,
-        ha_url=getenv("HA_URL"),
-        ha_token=getenv("HA_TOKEN"),
-        engine=engine,
-    )
-
-    llm_host = getenv("LLM_HOST")
-    llm_model = getenv("LLM_MODEL", "qwen3-vl:32b")
-    llm_port = int(getenv("LLM_PORT", "11434"))
-    if llm_host is None:
-        error = "LLM_HOST environment variable not set"
-        raise ValueError(error)
-
-    with (
-        SignalClient(config.signal_api_url, config.phone_number) as signal,
-        LLMClient(model=llm_model, host=llm_host, port=llm_port, timeout=llm_timeout) as llm,
-    ):
-        registry = DeviceRegistry(signal, engine)
-        bot = Bot(signal, llm, registry, config)
-        bot.run()
-
-
-if __name__ == "__main__":
-    typer.run(main)

python/signal_bot/models.py

@@ -1,97 +0,0 @@
-"""Models for the Signal command and control bot."""
-
-from __future__ import annotations
-
-from datetime import datetime  # noqa: TC003 - pydantic needs this at runtime
-from enum import StrEnum
-from typing import Any
-
-from pydantic import BaseModel, ConfigDict
-from sqlalchemy.engine import Engine  # noqa: TC002 - pydantic needs this at runtime
-
-
-class TrustLevel(StrEnum):
-    """Device trust level."""
-
-    VERIFIED = "verified"
-    UNVERIFIED = "unverified"
-    BLOCKED = "blocked"
-
-
-class Role(StrEnum):
-    """RBAC roles — one per command, plus admin which grants all."""
-
-    ADMIN = "admin"
-    STATUS = "status"
-    INVENTORY = "inventory"
-    LOCATION = "location"
-
-
-class MessageStatus(StrEnum):
-    """Dead letter queue message status."""
-
-    UNPROCESSED = "unprocessed"
-    PROCESSED = "processed"
-
-
-class Device(BaseModel):
-    """A registered device tracked by safety number."""
-
-    phone_number: str
-    safety_number: str
-    trust_level: TrustLevel = TrustLevel.UNVERIFIED
-    first_seen: datetime
-    last_seen: datetime
-
-
-class SignalMessage(BaseModel):
-    """An incoming Signal message."""
-
-    source: str
-    timestamp: int
-    message: str = ""
-    attachments: list[str] = []
-    group_id: str | None = None
-    is_receipt: bool = False
-
-
-class SignalEnvelope(BaseModel):
-    """Raw envelope from signal-cli-rest-api."""
-
-    envelope: dict[str, Any]
-    account: str | None = None
-
-
-class InventoryItem(BaseModel):
-    """An item in the van inventory."""
-
-    name: str
-    quantity: float = 1
-    unit: str = "each"
-    category: str = ""
-    notes: str = ""
-
-
-class InventoryUpdate(BaseModel):
-    """Result of processing an inventory update."""
-
-    items: list[InventoryItem] = []
-    raw_response: str = ""
-    source_type: str = ""  # "receipt_photo" or "text_list"
-
-
-class BotConfig(BaseModel):
-    """Top-level bot configuration."""
-
-    model_config = ConfigDict(arbitrary_types_allowed=True)
-
-    signal_api_url: str
-    phone_number: str
-    inventory_api_url: str
-    ha_url: str | None = None
-    ha_token: str | None = None
-    engine: Engine
-    reconnect_delay: int = 5
-    max_reconnect_delay: int = 300
-    max_retries: int = 10
-    max_message_attempts: int = 3

python/signal_bot/signal_client.py

@@ -1,141 +0,0 @@
-"""Client for the signal-cli-rest-api."""
-
-from __future__ import annotations
-
-import json
-import logging
-from typing import TYPE_CHECKING, Any, Self
-
-import httpx
-import websockets.sync.client
-
-if TYPE_CHECKING:
-    from collections.abc import Generator
-
-from python.signal_bot.models import SignalMessage
-
-logger = logging.getLogger(__name__)
-
-
-def _parse_envelope(envelope: dict[str, Any]) -> SignalMessage | None:
-    """Parse a signal-cli envelope into a SignalMessage, or None if not a data message."""
-    data_message = envelope.get("dataMessage")
-    if not data_message:
-        return None
-
-    attachment_ids = [att["id"] for att in data_message.get("attachments", []) if "id" in att]
-
-    group_info = data_message.get("groupInfo")
-    group_id = group_info.get("groupId") if group_info else None
-
-    return SignalMessage(
-        source=envelope.get("source", ""),
-        timestamp=envelope.get("timestamp", 0),
-        message=data_message.get("message", "") or "",
-        attachments=attachment_ids,
-        group_id=group_id,
-    )
-
-
-class SignalClient:
-    """Communicate with signal-cli-rest-api.
-
-    Args:
-        base_url: URL of the signal-cli-rest-api (e.g. http://localhost:8989).
-        phone_number: The registered phone number to send/receive as.
-    """
-
-    def __init__(self, base_url: str, phone_number: str) -> None:
-        self.base_url = base_url.rstrip("/")
-        self.phone_number = phone_number
-        self._client = httpx.Client(base_url=self.base_url, timeout=30)
-
-    def _ws_url(self) -> str:
-        """Build the WebSocket URL from the base HTTP URL."""
-        url = self.base_url.replace("http://", "ws://").replace("https://", "wss://")
-        return f"{url}/v1/receive/{self.phone_number}"
-
-    def listen(self) -> Generator[SignalMessage]:
-        """Connect via WebSocket and yield messages as they arrive."""
-        ws_url = self._ws_url()
-        logger.info(f"Connecting to WebSocket: {ws_url}")
-
-        with websockets.sync.client.connect(ws_url) as ws:
-            for raw in ws:
-                try:
-                    data = json.loads(raw)
-                    envelope = data.get("envelope", {})
-                    message = _parse_envelope(envelope)
-                    if message:
-                        yield message
-                except json.JSONDecodeError:
-                    logger.warning(f"Non-JSON WebSocket frame: {raw[:200]}")
-
-    def send(self, recipient: str, message: str) -> None:
-        """Send a text message."""
-        payload = {
-            "message": message,
-            "number": self.phone_number,
-            "recipients": [recipient],
-        }
-        response = self._client.post("/v2/send", json=payload)
-        response.raise_for_status()
-
-    def send_to_group(self, group_id: str, message: str) -> None:
-        """Send a message to a group."""
-        payload = {
-            "message": message,
-            "number": self.phone_number,
-            "recipients": [group_id],
-        }
-        response = self._client.post("/v2/send", json=payload)
-        response.raise_for_status()
-
-    def get_attachment(self, attachment_id: str) -> bytes:
-        """Download an attachment by ID."""
-        response = self._client.get(f"/v1/attachments/{attachment_id}")
-        response.raise_for_status()
-        return response.content
-
-    def get_identities(self) -> list[dict[str, Any]]:
-        """List known identities and their trust levels."""
-        response = self._client.get(f"/v1/identities/{self.phone_number}")
-        response.raise_for_status()
-        return response.json()
-
-    def get_safety_number(self, phone_number: str) -> str | None:
-        """Look up the safety number for a contact from signal-cli's local store."""
-        for identity in self.get_identities():
-            if identity.get("number") == phone_number:
-                return identity.get("safety_number", identity.get("fingerprint", ""))
-        return None
-
-    def trust_identity(self, number_to_trust: str, *, trust_all_known_keys: bool = False) -> None:
-        """Trust an identity (verify safety number)."""
-        payload: dict[str, Any] = {}
-        if trust_all_known_keys:
-            payload["trust_all_known_keys"] = True
-        response = self._client.put(
-            f"/v1/identities/{self.phone_number}/trust/{number_to_trust}",
-            json=payload,
-        )
-        response.raise_for_status()
-
-    def reply(self, message: SignalMessage, text: str) -> None:
-        """Reply to a message, routing to group or individual."""
-        if message.group_id:
-            self.send_to_group(message.group_id, text)
-        else:
-            self.send(message.source, text)
-
-    def __enter__(self) -> Self:
-        """Enter the context manager."""
-        return self
-
-    def __exit__(self, *args: object) -> None:
-        """Close the HTTP client on exit."""
-        self.close()
-
-    def close(self) -> None:
-        """Close the HTTP client."""
-        self._client.close()

python/tools/audiobook/__init__.py

@@ -0,0 +1 @@
+"""Audiobook tools."""

python/tools/audiobook/audible_convert.py

@@ -0,0 +1,471 @@
+"""Convert Audible AAX downloads into Audiobookshelf-friendly M4B files."""
+
+from __future__ import annotations
+
+import json
+import logging
+import re
+import shutil
+import subprocess
+from concurrent.futures import ThreadPoolExecutor
+from dataclasses import asdict, dataclass
+from os import getenv
+from pathlib import Path  # noqa: TC003 This is required for the typer CLI
+from typing import TYPE_CHECKING, Annotated, Any
+from uuid import uuid7
+
+import typer
+
+from python.common import configure_logger
+from python.orm.common import get_postgres_engine
+from python.tools.audiobook.metadata_agent import (
+    AgentConfig,
+    StandardBookMetadata,
+    standard_book_metadata,
+    write_agent_log,
+)
+
+if TYPE_CHECKING:
+    from sqlalchemy.engine import Engine
+
+logger = logging.getLogger(__name__)
+
+SENSITIVE_COMMAND_ARGUMENTS = {"-activation_bytes"}
+BOOK_RANGE_PATTERN = re.compile(r"(?:^|-)books?-(?P<start>[1-9]\d*)-(?P<end>[1-9]\d*)(?:-|$)")
+
+
+@dataclass(frozen=True)
+class ConversionConfig:
+    """Runtime settings for one conversion command."""
+
+    resolved_output: Path
+    ollama_api_key: str
+    agent_config: AgentConfig
+    engine: Engine
+    activation_bytes: str | None
+    dry_run: bool
+    overwrite: bool
+    work_directory_name: str = ".audible_convert"
+    dry_run_directory_name: str = "dry-run"
+    temp_directory_name: str = "tmp"
+    log_directory_name: str = "logs"
+    review_directory_name: str = "review"
+
+
+@dataclass(frozen=True)
+class ConcurrentConversionResult:
+    """Result from running ffmpeg and metadata resolution together."""
+
+    metadata: StandardBookMetadata | None
+    conversion_error: Exception | None
+    metadata_error: Exception | None
+
+
+class CommandExecutionError(RuntimeError):
+    """Command failed without exposing sensitive arguments."""
+
+    def __init__(self, arguments: list[str], returncode: int) -> None:
+        """Create a redacted command failure."""
+        self.arguments = tuple(arguments)
+        self.returncode = returncode
+        command = " ".join(redact_command_arguments(arguments))
+        super().__init__(f"Command failed with exit code {returncode}: {command}")
+
+
+def main(
+    input_directory: Annotated[Path, typer.Argument(help="Directory audible-cli downloads AAX files into.")],
+    output_directory: Annotated[Path, typer.Argument(help="Audiobook output directory.")],
+    *,
+    dry_run: Annotated[
+        bool,
+        typer.Option("--dry-run", help="Print planned output files and write marker files without converting."),
+    ] = False,
+    overwrite: Annotated[bool, typer.Option("--overwrite", help="Overwrite existing M4B files.")] = False,
+) -> None:
+    """Convert AAX files from a download directory into M4B files."""
+    configure_logger()
+    resolved_input = input_directory.resolve(strict=True)
+    resolved_output = output_directory.resolve()
+    if not dry_run:
+        resolved_output.mkdir(parents=True, exist_ok=True)
+
+    ollama_api_key = getenv("OLLAMA_API_KEY")
+    if not ollama_api_key:
+        msg = "OLLAMA_API_KEY is required for audiobook metadata resolution"
+        raise RuntimeError(msg)
+
+    config = ConversionConfig(
+        resolved_output=resolved_output,
+        ollama_api_key=ollama_api_key,
+        agent_config=AgentConfig(),
+        engine=get_postgres_engine(name="RICHIE"),
+        activation_bytes=getenv("AUDIBLE_ACTIVATION_BYTES"),
+        dry_run=dry_run,
+        overwrite=overwrite,
+    )
+
+    aax_files = sorted(resolved_input.glob("*.aax"))
+    if not aax_files:
+        logger.info("No AAX files found in %s", resolved_input)
+        return
+    for aax_file in aax_files:
+        logger.info("Converting %s", aax_file)
+        convert_aax_file_with_agent(aax_file, config)
+
+
+def run_command(arguments: list[str], *, capture: bool = False) -> subprocess.CompletedProcess[str]:
+    """Run a command and return the completed process.
+
+    Args:
+        arguments: Command and arguments to run.
+        capture: Whether to capture stdout and stderr.
+
+    Returns:
+        The completed process.
+    """
+    logger.debug("%s", " ".join(redact_command_arguments(arguments)))
+    try:
+        return subprocess.run(arguments, check=True, capture_output=capture, text=True)
+    except subprocess.CalledProcessError as error:
+        raise CommandExecutionError(arguments, error.returncode) from error
+
+
+def redact_command_arguments(arguments: list[str]) -> list[str]:
+    """Return command arguments with sensitive values redacted."""
+    redacted = []
+    redact_next = False
+    for argument in arguments:
+        if redact_next:
+            redacted.append("<redacted>")
+            redact_next = False
+            continue
+
+        redacted.append(argument)
+        redact_next = argument in SENSITIVE_COMMAND_ARGUMENTS
+    return redacted
+
+
+def read_metadata(aax_file: Path) -> dict[str, str]:
+    """Read ffprobe format tags from an AAX file.
+
+    Args:
+        aax_file: AAX file to inspect.
+
+    Returns:
+        Lower-cased metadata tag names mapped to their values.
+    """
+    completed = run_command(
+        [
+            "ffprobe",
+            "-v",
+            "quiet",
+            "-print_format",
+            "json",
+            "-show_format",
+            str(aax_file),
+        ],
+        capture=True,
+    )
+    ffprobe_data: dict[str, Any] = json.loads(completed.stdout)
+    tags = ffprobe_data.get("format", {}).get("tags", {})
+    return {str(key).lower(): str(value) for key, value in tags.items()}
+
+
+def output_stem(metadata: StandardBookMetadata) -> str:
+    """Build the output stem for a book.
+
+    Args:
+        metadata: Book metadata.
+
+    Returns:
+        Output stem in author-series_01-title form.
+    """
+    index_slug = series_index_slug(metadata.series_index, metadata.title)
+    return f"{metadata.author}-{metadata.series}_{index_slug}-{metadata.title}"
+
+
+def series_index_slug(series_index: float, title: str = "") -> str:
+    """Return a filename-safe series index."""
+    if title_range := title_series_range_slug(series_index, title):
+        return title_range
+    index = float(series_index)
+    if index.is_integer():
+        return f"{int(index):02}"
+    return f"{int(index):02}.5"
+
+
+def title_series_range_slug(series_index: float, title: str) -> str | None:
+    """Return a series range slug found in an omnibus title."""
+    index = float(series_index)
+    if not index.is_integer():
+        return None
+    first_index = int(index)
+    for match in BOOK_RANGE_PATTERN.finditer(title):
+        start = int(match.group("start"))
+        end = int(match.group("end"))
+        if start == first_index and end > start:
+            return f"{start:02}-{end:02}"
+    return None
+
+
+def metadata_output_path(output_directory: Path, metadata: StandardBookMetadata) -> Path:
+    """Build the final M4B path from resolved metadata."""
+    stem = output_stem(metadata)
+    return output_directory / stem / f"{stem}.m4b"
+
+
+def convert_aax_file(
+    aax_file: Path,
+    destination: Path,
+    activation_bytes: str | None,
+    *,
+    overwrite: bool,
+) -> None:
+    """Convert an AAX file into an M4B file.
+
+    Args:
+        aax_file: Source AAX file.
+        destination: Destination M4B file.
+        activation_bytes: Optional Audible activation bytes for ffmpeg.
+        overwrite: Whether to overwrite an existing M4B.
+    """
+    if destination.exists() and not overwrite:
+        logger.info("Skipping existing file %s", destination)
+        return
+
+    destination.parent.mkdir(parents=True, exist_ok=True)
+    arguments = ["ffmpeg", "-hide_banner", "-y" if overwrite else "-n"]
+    if activation_bytes:
+        arguments.extend(["-activation_bytes", activation_bytes])
+    arguments.extend(["-i", str(aax_file), "-map_metadata", "0", "-c", "copy", str(destination)])
+    run_command(arguments)
+
+
+def write_review_file(
+    *,
+    destination: Path | None,
+    ffprobe_metadata: dict[str, str],
+    log_file: Path,
+    metadata: StandardBookMetadata | None,
+    reason: str,
+    review_file: Path,
+    source: Path,
+    temp_file: Path | None,
+) -> None:
+    """Write a manual review file for an unresolved conversion."""
+    review_file.parent.mkdir(parents=True, exist_ok=True)
+    payload = {
+        "destination": str(destination) if destination else None,
+        "ffprobe_metadata": ffprobe_metadata,
+        "metadata": asdict(metadata) if metadata else None,
+        "reason": reason,
+        "source": str(source),
+        "temp_file": str(temp_file) if temp_file else None,
+    }
+    review_file.write_text(json.dumps(payload, indent=2, sort_keys=True), encoding="utf-8")
+    write_agent_log(log_file, "review_written", path=str(review_file), reason=reason)
+
+
+def cleanup_temp_output(temp_file: Path) -> None:
+    """Remove a run's temporary output directory."""
+    shutil.rmtree(temp_file.parent, ignore_errors=True)
+
+
+def dry_run_aax_file_with_agent(
+    aax_file: Path,
+    ffprobe_metadata: dict[str, str],
+    engine: Engine,
+    config: ConversionConfig,
+    log_file: Path,
+    review_file: Path,
+) -> None:
+    """Resolve and print the planned output path without converting."""
+    metadata = standard_book_metadata(
+        aax_file.name,
+        ffprobe_metadata,
+        engine,
+        log_file,
+        config.ollama_api_key,
+        config.agent_config,
+    )
+    destination = None if metadata.needs_review else metadata_output_path(config.resolved_output, metadata)
+    if metadata.needs_review:
+        write_review_file(
+            destination=destination,
+            ffprobe_metadata=ffprobe_metadata,
+            log_file=log_file,
+            metadata=metadata,
+            reason="metadata_needs_review",
+            review_file=review_file,
+            source=aax_file,
+            temp_file=None,
+        )
+        typer.echo(f"{aax_file} -> REVIEW {review_file}")
+    else:
+        stem = output_stem(metadata)
+        dry_run_file = (
+            config.resolved_output / config.work_directory_name / config.dry_run_directory_name / stem / f"{stem}.m4b"
+        )
+        dry_run_file.parent.mkdir(parents=True, exist_ok=True)
+        dry_run_file.write_text(f"{destination}\n", encoding="utf-8")
+        write_agent_log(
+            log_file,
+            "dry_run_file_written",
+            destination=str(destination),
+            path=str(dry_run_file),
+        )
+        typer.echo(f"{aax_file} -> {destination}")
+
+
+def convert_temp_file_and_resolve_metadata(
+    aax_file: Path,
+    temp_file: Path,
+    ffprobe_metadata: dict[str, str],
+    config: ConversionConfig,
+    log_file: Path,
+) -> ConcurrentConversionResult:
+    """Run ffmpeg and metadata resolution in parallel."""
+    conversion_error: Exception | None = None
+    metadata_error: Exception | None = None
+    metadata: StandardBookMetadata | None = None
+
+    with ThreadPoolExecutor(max_workers=2) as executor:
+        conversion_future = executor.submit(
+            convert_aax_file,
+            aax_file,
+            temp_file,
+            config.activation_bytes,
+            overwrite=True,
+        )
+        metadata_future = executor.submit(
+            standard_book_metadata,
+            aax_file.name,
+            ffprobe_metadata,
+            config.engine,
+            log_file,
+            config.ollama_api_key,
+            config.agent_config,
+        )
+
+        conversion_error = conversion_future.exception()
+        if conversion_error is None:
+            conversion_future.result()
+
+        metadata_error = metadata_future.exception()
+        if metadata_error is None:
+            metadata = metadata_future.result()
+
+    return ConcurrentConversionResult(
+        metadata=metadata,
+        conversion_error=conversion_error,
+        metadata_error=metadata_error,
+    )
+
+
+def convert_aax_file_with_agent(aax_file: Path, config: ConversionConfig) -> None:
+    """Convert one AAX file using the metadata agent for the final path."""
+    run_id = uuid7().hex
+    log_file = config.resolved_output / config.work_directory_name / config.log_directory_name / f"{run_id}.jsonl"
+    review_file = config.resolved_output / config.work_directory_name / config.review_directory_name / f"{run_id}.json"
+    write_agent_log(log_file, "conversion_start", source=str(aax_file), dry_run=config.dry_run)
+    try:
+        ffprobe_metadata = read_metadata(aax_file)
+    except Exception as error:
+        logger.exception("ffprobe failed")
+        write_review_file(
+            destination=None,
+            ffprobe_metadata={},
+            log_file=log_file,
+            metadata=None,
+            reason=f"ffprobe_failed: {error}",
+            review_file=review_file,
+            source=aax_file,
+            temp_file=None,
+        )
+        return
+
+    if config.dry_run:
+        dry_run_aax_file_with_agent(
+            aax_file,
+            ffprobe_metadata,
+            config.engine,
+            config,
+            log_file,
+            review_file,
+        )
+        return
+
+    temp_file = (
+        config.resolved_output / config.work_directory_name / config.temp_directory_name / run_id / "converted.m4b"
+    )
+    temp_file.parent.mkdir(parents=True, exist_ok=True)
+
+    result = convert_temp_file_and_resolve_metadata(aax_file, temp_file, ffprobe_metadata, config, log_file)
+
+    if result.conversion_error:
+        reason = f"ffmpeg_failed: {result.conversion_error}"
+        write_review_file(
+            destination=None,
+            ffprobe_metadata=ffprobe_metadata,
+            log_file=log_file,
+            metadata=result.metadata,
+            reason=reason,
+            review_file=review_file,
+            source=aax_file,
+            temp_file=temp_file if temp_file.exists() else None,
+        )
+        return
+
+    if result.metadata_error:
+        write_review_file(
+            destination=None,
+            ffprobe_metadata=ffprobe_metadata,
+            log_file=log_file,
+            metadata=None,
+            reason=f"metadata_failed: {result.metadata_error}",
+            review_file=review_file,
+            source=aax_file,
+            temp_file=temp_file,
+        )
+        return
+
+    if result.metadata is None or result.metadata.needs_review:
+        write_review_file(
+            destination=None,
+            ffprobe_metadata=ffprobe_metadata,
+            log_file=log_file,
+            metadata=result.metadata,
+            reason="metadata_needs_review",
+            review_file=review_file,
+            source=aax_file,
+            temp_file=temp_file,
+        )
+        return
+
+    destination = metadata_output_path(config.resolved_output, result.metadata)
+    if destination.exists() and not config.overwrite:
+        write_agent_log(log_file, "destination_exists", destination=str(destination))
+        cleanup_temp_output(temp_file)
+        return
+
+    destination.parent.mkdir(parents=True, exist_ok=True)
+    try:
+        temp_file.replace(destination)
+    except Exception as error:  # noqa: BLE001
+        write_review_file(
+            destination=destination,
+            ffprobe_metadata=ffprobe_metadata,
+            log_file=log_file,
+            metadata=result.metadata,
+            reason=f"rename_failed: {error}",
+            review_file=review_file,
+            source=aax_file,
+            temp_file=temp_file if temp_file.exists() else None,
+        )
+    else:
+        cleanup_temp_output(temp_file)
+        write_agent_log(log_file, "conversion_complete", destination=str(destination))
+
+
+if __name__ == "__main__":
+    typer.run(main)

python/tools/audiobook/catalog.py

@@ -0,0 +1,176 @@
+"""Import audiobook catalog authors and series from CSV files."""
+
+from __future__ import annotations
+
+import csv
+import logging
+from pathlib import Path  # noqa: TC003 This is required for the typer CLI
+from typing import Annotated
+
+import typer
+from sqlalchemy import select
+from sqlalchemy.orm import Session
+
+from python.common import configure_logger
+from python.orm.common import get_postgres_engine
+from python.orm.richie import AudiobookAuthor, AudiobookSeries
+
+logger = logging.getLogger(__name__)
+
+AUTHOR_NAME_COLUMN = "author_name"
+ID_COLUMN = "id"
+NAME_COLUMN = "name"
+
+
+class CatalogImportError(ValueError):
+    """CSV catalog import failed validation."""
+
+
+def main(
+    authors_csv: Annotated[Path, typer.Argument(help="CSV with name and optional id.")],
+    series_csv: Annotated[Path, typer.Argument(help="CSV with name, author_name, and optional id.")],
+) -> None:
+    """Upsert audiobook authors and series from CSV files."""
+    configure_logger()
+    try:
+        engine = get_postgres_engine(name="RICHIE")
+        with Session(engine) as session:
+            author_count = upsert_authors_from_csv(session, authors_csv)
+            series_count = upsert_series_from_csv(session, series_csv)
+            session.commit()
+    except CatalogImportError as error:
+        typer.echo(str(error), err=True)
+        raise typer.Exit(code=1) from error
+
+    logger.info("Upserted %s authors and %s series", author_count, series_count)
+
+
+def upsert_authors_from_csv(session: Session, authors_csv: Path) -> int:
+    """Upsert authors from a CSV file."""
+    count = 0
+    for row_number, row in csv_rows(authors_csv):
+        name = required_csv_value(row, authors_csv, row_number, NAME_COLUMN)
+        upsert_author(session, name, csv_id(row, authors_csv, row_number))
+        count += 1
+    return count
+
+
+def upsert_series_from_csv(session: Session, series_csv: Path) -> int:
+    """Upsert series from a CSV file."""
+    count = 0
+    for row_number, row in csv_rows(series_csv):
+        series_name = required_csv_value(row, series_csv, row_number, NAME_COLUMN)
+        author_name = required_csv_value(row, series_csv, row_number, AUTHOR_NAME_COLUMN)
+        author = find_author_by_name(session, author_name)
+        if author is None:
+            msg = f"{series_csv}:{row_number}: author not found: {author_name}"
+            raise CatalogImportError(msg)
+        upsert_series(session, series_name, author, csv_id(row, series_csv, row_number))
+        count += 1
+    return count
+
+
+def upsert_author(session: Session, name: str, author_id: int | None) -> AudiobookAuthor:
+    """Upsert one author by id or exact name."""
+    if author_id is not None:
+        author = session.get(AudiobookAuthor, author_id)
+        if author is None:
+            author = AudiobookAuthor(id=author_id, name=name)
+            session.add(author)
+        else:
+            author.name = name
+        session.flush()
+        return author
+
+    author = find_author_by_name(session, name)
+    if author is None:
+        author = AudiobookAuthor(name=name)
+        session.add(author)
+        session.flush()
+    return author
+
+
+def upsert_series(
+    session: Session,
+    name: str,
+    author: AudiobookAuthor,
+    series_id: int | None,
+) -> AudiobookSeries:
+    """Upsert one series by id or exact author/name match."""
+    if series_id is not None:
+        series = session.get(AudiobookSeries, series_id)
+        if series is None:
+            series = AudiobookSeries(id=series_id, name=name, author=author)
+            session.add(series)
+        else:
+            series.name = name
+            series.author = author
+        session.flush()
+        return series
+
+    series = find_series_by_name_and_author(session, name, author.id)
+    if series is None:
+        series = AudiobookSeries(name=name, author=author)
+        session.add(series)
+        session.flush()
+    return series
+
+
+def find_author_by_name(session: Session, name: str) -> AudiobookAuthor | None:
+    """Find one author by exact name."""
+    return session.scalar(select(AudiobookAuthor).where(AudiobookAuthor.name == name))
+
+
+def find_series_by_name_and_author(
+    session: Session,
+    name: str,
+    author_id: int,
+) -> AudiobookSeries | None:
+    """Find one series by exact name and author."""
+    return session.scalar(
+        select(AudiobookSeries).where(
+            AudiobookSeries.name == name,
+            AudiobookSeries.author_id == author_id,
+        ),
+    )
+
+
+def csv_rows(csv_path: Path) -> list[tuple[int, dict[str, str | None]]]:
+    """Read a CSV file as numbered rows."""
+    with csv_path.open(newline="", encoding="utf-8") as file:
+        reader = csv.DictReader(file)
+        if reader.fieldnames is None:
+            msg = f"{csv_path}: missing CSV header"
+            raise CatalogImportError(msg)
+        return [(row_number, row) for row_number, row in enumerate(reader, start=2)]
+
+
+def required_csv_value(
+    row: dict[str, str | None],
+    csv_path: Path,
+    row_number: int,
+    column: str,
+) -> str:
+    """Read a required CSV value."""
+    value = row.get(column)
+    if value and value.strip():
+        return value.strip()
+    msg = f"{csv_path}:{row_number}: missing required column value: {column}"
+    raise CatalogImportError(msg)
+
+
+def csv_id(row: dict[str, str | None], csv_path: Path, row_number: int) -> int | None:
+    """Read an optional id field from a CSV row."""
+    value = row.get(ID_COLUMN)
+    if value is None or not value.strip():
+        return None
+    try:
+        return int(value)
+    except ValueError as error:
+        msg = f"{csv_path}:{row_number}: id must be an integer: {value}"
+        raise CatalogImportError(msg) from error
+    return None
+
+
+if __name__ == "__main__":
+    typer.run(main)

python/tools/audiobook/llm_tool_calling.py

@@ -0,0 +1,599 @@
+"""LLM tool calling support for audiobook metadata resolution."""
+
+from __future__ import annotations
+
+import json
+import re
+import time
+from collections.abc import Callable
+from dataclasses import dataclass
+from typing import TYPE_CHECKING
+
+from sqlalchemy import or_, select
+
+from python.orm.richie import Audiobook, AudiobookAuthor, AudiobookSeries
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    from sqlalchemy.orm import Session
+
+    from python.tools.audiobook.metadata_agent import AgentConfig
+
+CATALOG_SLUG_PATTERN = re.compile(r"^[a-z0-9]+(?:_[a-z0-9]+)*$")
+TITLE_SLUG_PATTERN = re.compile(r"^[a-z0-9]+(?:-[a-z0-9]+)*$")
+
+LogWriter = Callable[..., None]
+
+
+class MetadataResolutionError(ValueError):
+    """Metadata resolution failed validation."""
+
+
+@dataclass(frozen=True)
+class EnsuredBook:
+    """Book row plus whether it was created."""
+
+    book: Audiobook
+    action: str
+
+
+class CatalogToolRegistry:
+    """Controlled catalog tools exposed to the metadata model."""
+
+    def __init__(
+        self,
+        session: Session,
+        log_path: Path,
+        config: AgentConfig,
+        write_log: LogWriter,
+    ) -> None:
+        """Create a registry bound to one database session and audit log."""
+        self.session = session
+        self.log_path = log_path
+        self.config = config
+        self.write_log = write_log
+        self.seen_author_ids: set[int] = set()
+        self.seen_series_ids: set[int] = set()
+        self.seen_book_ids: set[int] = set()
+        self.created_author_ids: set[int] = set()
+        self.created_series_ids: set[int] = set()
+        self.created_book_ids: set[int] = set()
+
+    def tool_schemas(self) -> list[dict[str, object]]:
+        """Return Ollama tool schemas."""
+        schemas = [
+            {
+                "type": "function",
+                "function": {
+                    "name": "search_authors",
+                    "description": "Search canonical audiobook authors by slug or noisy source text.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"query": {"type": "string"}},
+                        "required": ["query"],
+                    },
+                },
+            },
+            {
+                "type": "function",
+                "function": {
+                    "name": "search_series",
+                    "description": "Search canonical audiobook series by slug or noisy source text.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "query": {"type": "string"},
+                            "author_id": {"type": ["integer", "null"]},
+                        },
+                        "required": ["query"],
+                    },
+                },
+            },
+            {
+                "type": "function",
+                "function": {
+                    "name": "search_books",
+                    "description": "Search canonical audiobook titles with optional author and series filters.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "query": {"type": "string"},
+                            "author_id": {"type": ["integer", "null"]},
+                            "series_id": {"type": ["integer", "null"]},
+                        },
+                        "required": ["query"],
+                    },
+                },
+            },
+            {
+                "type": "function",
+                "function": {
+                    "name": "ensure_author",
+                    "description": "Normalize an author name to a catalog slug, then return or create that author.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"name": {"type": "string"}},
+                        "required": ["name"],
+                    },
+                },
+            },
+            {
+                "type": "function",
+                "function": {
+                    "name": "ensure_series",
+                    "description": "Normalize a series name to a catalog slug, then return or create it for an author.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "name": {"type": "string"},
+                            "author_id": {"type": "integer"},
+                        },
+                        "required": ["name", "author_id"],
+                    },
+                },
+            },
+            {
+                "type": "function",
+                "function": {
+                    "name": "ensure_book",
+                    "description": "Normalize a title to a book slug, then return or create it for an author/series.",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "title": {"type": "string"},
+                            "author_id": {"type": "integer"},
+                            "series_id": {"type": ["integer", "null"]},
+                            "series_index": {"type": "number", "multipleOf": 0.5},
+                        },
+                        "required": ["title", "author_id", "series_id", "series_index"],
+                    },
+                },
+            },
+        ]
+        enabled_tool_names = set(self.config.tool_names)
+        return [schema for schema in schemas if schema["function"]["name"] in enabled_tool_names]
+
+    def run(self, name: str, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Run one catalog tool and audit the call."""
+        handlers = {
+            "search_authors": self.run_search_authors,
+            "search_series": self.run_search_series,
+            "search_books": self.run_search_books,
+            "ensure_author": self.run_ensure_author,
+            "ensure_series": self.run_ensure_series,
+            "ensure_book": self.run_ensure_book,
+        }
+        handler = handlers.get(name)
+        if handler is None:
+            self.write_log(self.log_path, "tool_error", tool=name, arguments=arguments, error="unknown_tool")
+            msg = f"Unknown audiobook metadata tool: {name}"
+            raise MetadataResolutionError(msg)
+        if name not in self.config.tool_names:
+            self.write_log(self.log_path, "tool_error", tool=name, arguments=arguments, error="tool_not_enabled")
+            msg = f"Audiobook metadata tool is not enabled: {name}"
+            raise MetadataResolutionError(msg)
+
+        started = time.perf_counter()
+        self.write_log(self.log_path, "tool_call", tool=name, arguments=arguments)
+        result = handler(arguments)
+        duration_ms = round((time.perf_counter() - started) * 1000, 3)
+        self.write_log(
+            self.log_path,
+            "tool_result",
+            tool=name,
+            duration_ms=duration_ms,
+            result_count=len(result),
+            preview=result[:3],
+        )
+        return result
+
+    def get_author(self, author_id: int) -> AudiobookAuthor | None:
+        """Return an author by id."""
+        return self.session.get(AudiobookAuthor, author_id)
+
+    def get_book(self, book_id: int) -> Audiobook | None:
+        """Return a book by id."""
+        return self.session.get(Audiobook, book_id)
+
+    def get_series(self, series_id: int) -> AudiobookSeries | None:
+        """Return a series by id."""
+        return self.session.get(AudiobookSeries, series_id)
+
+    def prune_unused_created_rows(self, *, author_id: int, book_id: int | None, series_id: int | None) -> None:
+        """Remove catalog rows created during this run but not used by final metadata."""
+        used_book_ids = {book_id} if book_id is not None else set()
+        for created_book_id in self.created_book_ids - used_book_ids:
+            if book := self.get_book(created_book_id):
+                self.session.delete(book)
+
+        self.session.flush()
+        used_series_ids = {series_id} if series_id is not None else set()
+        for created_series_id in self.created_series_ids - used_series_ids:
+            series = self.get_series(created_series_id)
+            if series and not series.books:
+                self.session.delete(series)
+
+        self.session.flush()
+        for created_author_id in self.created_author_ids - {author_id}:
+            author = self.get_author(created_author_id)
+            if author and not author.books and not author.series:
+                self.session.delete(author)
+
+    def run_search_authors(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Search authors from tool arguments and remember returned ids."""
+        query = required_string(arguments, "query")
+        statement = select(AudiobookAuthor).order_by(AudiobookAuthor.name).limit(self.config.max_tool_results)
+        if terms := query_terms(query):
+            statement = statement.where(or_(*(AudiobookAuthor.name.ilike(f"%{term}%") for term in terms)))
+
+        authors = self.session.scalars(statement).all()
+        self.seen_author_ids.update(author.id for author in authors)
+        return [{"id": author.id, "name": author.name} for author in authors]
+
+    def run_search_series(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Search series from tool arguments and remember returned ids."""
+        query = required_string(arguments, "query")
+        author_id = optional_int(arguments.get("author_id"), "author_id")
+        statement = select(AudiobookSeries).order_by(AudiobookSeries.name).limit(self.config.max_tool_results)
+        if terms := query_terms(query):
+            statement = statement.where(or_(*(AudiobookSeries.name.ilike(f"%{term}%") for term in terms)))
+        if author_id is not None:
+            statement = statement.where(AudiobookSeries.author_id == author_id)
+
+        series_rows = self.session.scalars(statement).all()
+        self.seen_series_ids.update(series.id for series in series_rows)
+        self.seen_author_ids.update(series.author_id for series in series_rows)
+        return [
+            {
+                "id": series.id,
+                "name": series.name,
+                "author_id": series.author_id,
+                "author": series.author.name,
+            }
+            for series in series_rows
+        ]
+
+    def run_search_books(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Search books from tool arguments and remember returned ids."""
+        query = required_string(arguments, "query")
+        author_id = optional_int(arguments.get("author_id"), "author_id")
+        series_id = optional_int(arguments.get("series_id"), "series_id")
+        statement = select(Audiobook).order_by(Audiobook.title).limit(self.config.max_tool_results)
+        if terms := query_terms(query):
+            statement = statement.where(or_(*(Audiobook.title.ilike(f"%{term}%") for term in terms)))
+        if author_id is not None:
+            statement = statement.where(Audiobook.author_id == author_id)
+        if series_id is not None:
+            statement = statement.where(Audiobook.series_id == series_id)
+
+        books = self.session.scalars(statement).all()
+        self.seen_book_ids.update(book.id for book in books)
+        self.seen_author_ids.update(book.author_id for book in books)
+        self.seen_series_ids.update(book.series_id for book in books if book.series_id is not None)
+        return [
+            {
+                "id": book.id,
+                "title": book.title,
+                "author_id": book.author_id,
+                "author": book.author.name,
+                "series_id": book.series_id,
+                "series": book.series.name if book.series else self.config.standalone_series,
+                "series_index": book.series_index,
+            }
+            for book in books
+        ]
+
+    def run_ensure_author(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Ensure an author from tool arguments and return a tool result."""
+        name = normalize_catalog_slug(required_string(arguments, "name"))
+        validate_catalog_slug(name, "author")
+        author = self.session.scalar(select(AudiobookAuthor).where(AudiobookAuthor.name == name))
+        action = "existing"
+        if author is None:
+            author = AudiobookAuthor(name=name)
+            self.session.add(author)
+            self.session.flush()
+            self.created_author_ids.add(author.id)
+            action = "created"
+
+        self.seen_author_ids.add(author.id)
+        return [{"id": author.id, "name": author.name, "action": action}]
+
+    def run_ensure_series(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Ensure a series from tool arguments and return a tool result."""
+        name = normalize_catalog_slug(required_string(arguments, "name"))
+        author_id = required_int(arguments, "author_id")
+        validate_catalog_slug(name, "series")
+        author = self.required_author(author_id)
+        series = self.find_series_by_catalog_slug(name, author.id)
+        action = "existing"
+        if series is None:
+            series = AudiobookSeries(name=name, author=author)
+            self.session.add(series)
+            self.session.flush()
+            self.created_series_ids.add(series.id)
+            action = "created"
+
+        self.seen_author_ids.add(author.id)
+        self.seen_series_ids.add(series.id)
+        return [self.series_result(series, action)]
+
+    def run_ensure_book(self, arguments: dict[str, object]) -> list[dict[str, object]]:
+        """Ensure a book from tool arguments and return a tool result."""
+        title = required_string(arguments, "title")
+        author_id = required_int(arguments, "author_id")
+        series_id = optional_int(arguments.get("series_id"), "series_id")
+        series_index = required_series_index(arguments, "series_index")
+        ensured = self.ensure_book(title, author_id, series_id, series_index)
+        return [self.book_result(ensured.book, ensured.action)]
+
+    def ensure_book(
+        self,
+        title: str,
+        author_id: int,
+        series_id: int | None,
+        series_index: float,
+    ) -> EnsuredBook:
+        """Return an existing book row, or create it after validating ownership."""
+        title = normalize_title_slug(title)
+        validate_title_slug(title)
+        author = self.required_author(author_id)
+        series = None
+        if series_id is None:
+            if series_index != 0:
+                msg = "standalone books must use series_index 0"
+                raise MetadataResolutionError(msg)
+        else:
+            series = self.required_series(series_id)
+            if series.author_id != author.id:
+                msg = f"series_id {series_id} does not belong to author_id {author_id}"
+                raise MetadataResolutionError(msg)
+            if series_index <= 0:
+                msg = "series books must use a positive series_index"
+                raise MetadataResolutionError(msg)
+
+        statement = select(Audiobook).where(
+            Audiobook.title == title,
+            Audiobook.author_id == author.id,
+        )
+        if series is None:
+            statement = statement.where(Audiobook.series_id.is_(None))
+        else:
+            statement = statement.where(Audiobook.series_id == series.id)
+        book = self.session.scalar(statement)
+        if book is None:
+            book = Audiobook(title=title, author=author, series=series, series_index=series_index)
+            self.session.add(book)
+            self.session.flush()
+            self.created_book_ids.add(book.id)
+            action = "created"
+        else:
+            action = "existing"
+
+        self.seen_book_ids.add(book.id)
+        self.seen_author_ids.add(author.id)
+        if book.series_id is not None:
+            self.seen_series_ids.add(book.series_id)
+        return EnsuredBook(book=book, action=action)
+
+    def required_author(self, author_id: int) -> AudiobookAuthor:
+        """Return an author or fail metadata resolution."""
+        author = self.get_author(author_id)
+        if author is None:
+            msg = f"author_id {author_id} does not exist"
+            raise MetadataResolutionError(msg)
+        return author
+
+    def required_series(self, series_id: int) -> AudiobookSeries:
+        """Return a series or fail metadata resolution."""
+        series = self.get_series(series_id)
+        if series is None:
+            msg = f"series_id {series_id} does not exist"
+            raise MetadataResolutionError(msg)
+        return series
+
+    def find_series_by_catalog_slug(self, name: str, author_id: int) -> AudiobookSeries | None:
+        """Return a series by exact slug or underscore-insensitive slug."""
+        exact = self.session.scalar(
+            select(AudiobookSeries).where(
+                AudiobookSeries.name == name,
+                AudiobookSeries.author_id == author_id,
+            ),
+        )
+        if exact is not None:
+            return exact
+
+        compact_name = compact_catalog_slug(name)
+        series_rows = self.session.scalars(
+            select(AudiobookSeries).where(AudiobookSeries.author_id == author_id).order_by(AudiobookSeries.name),
+        ).all()
+        for series in series_rows:
+            if compact_catalog_slug(series.name) == compact_name:
+                return series
+        return None
+
+    def series_result(self, series: AudiobookSeries, action: str) -> dict[str, object]:
+        """Build a normalized series tool result."""
+        return {
+            "id": series.id,
+            "name": series.name,
+            "author_id": series.author_id,
+            "author": series.author.name,
+            "action": action,
+        }
+
+    def book_result(self, book: Audiobook, action: str) -> dict[str, object]:
+        """Build a normalized book tool result."""
+        return {
+            "id": book.id,
+            "title": book.title,
+            "author_id": book.author_id,
+            "author": book.author.name,
+            "series_id": book.series_id,
+            "series": book.series.name if book.series else self.config.standalone_series,
+            "series_index": book.series_index,
+            "action": action,
+        }
+
+
+def run_tool_calls(
+    messages: list[dict[str, object]],
+    message: dict[str, object],
+    tool_calls: list[tuple[str, dict[str, object]]],
+    registry: CatalogToolRegistry,
+    log_path: Path,
+    write_log: LogWriter,
+) -> str | None:
+    """Run tool calls, append tool messages, and return fatal error text when stopped."""
+    messages.append(message)
+    for tool_name, arguments in tool_calls:
+        try:
+            tool_result = registry.run(tool_name, arguments)
+        except MetadataResolutionError as error:
+            if is_fatal_tool_error(error):
+                return str(error)
+            write_log(log_path, "tool_error", tool=tool_name, arguments=arguments, error=str(error))
+            messages.append(
+                {
+                    "role": "tool",
+                    "tool_name": tool_name,
+                    "content": json.dumps({"error": str(error)}, sort_keys=True),
+                },
+            )
+            continue
+        messages.append(
+            {
+                "role": "tool",
+                "tool_name": tool_name,
+                "content": json.dumps(tool_result, sort_keys=True),
+            },
+        )
+    return None
+
+
+def parse_tool_calls(message: dict[str, object]) -> list[tuple[str, dict[str, object]]]:
+    """Parse Ollama tool calls from a response message."""
+    raw_tool_calls = message.get("tool_calls") or []
+    if not isinstance(raw_tool_calls, list):
+        msg = "tool_calls must be a list"
+        raise MetadataResolutionError(msg)
+
+    tool_calls = []
+    for raw_call in raw_tool_calls:
+        if not isinstance(raw_call, dict):
+            msg = "tool call must be an object"
+            raise MetadataResolutionError(msg)
+        function = raw_call.get("function")
+        if not isinstance(function, dict):
+            msg = "tool call is missing function"
+            raise MetadataResolutionError(msg)
+        name = function.get("name")
+        if not isinstance(name, str) or not name:
+            msg = "tool call is missing function name"
+            raise MetadataResolutionError(msg)
+        arguments = parse_tool_arguments(function.get("arguments", {}))
+        tool_calls.append((name, arguments))
+    return tool_calls
+
+
+def parse_tool_arguments(raw_arguments: object) -> dict[str, object]:
+    """Parse tool call arguments returned by Ollama."""
+    if isinstance(raw_arguments, dict):
+        return {str(key): value for key, value in raw_arguments.items()}
+    if isinstance(raw_arguments, str):
+        parsed = json.loads(raw_arguments) if raw_arguments else {}
+        if isinstance(parsed, dict):
+            return {str(key): value for key, value in parsed.items()}
+    msg = "tool arguments must be an object"
+    raise MetadataResolutionError(msg)
+
+
+def validate_title_slug(title: str) -> None:
+    """Validate a canonical book title slug."""
+    if not TITLE_SLUG_PATTERN.fullmatch(title):
+        msg = f"title slug is invalid: {title}"
+        raise MetadataResolutionError(msg)
+
+
+def validate_catalog_slug(value: str, label: str) -> None:
+    """Validate a canonical catalog slug."""
+    if not CATALOG_SLUG_PATTERN.fullmatch(value):
+        msg = f"{label} slug is invalid: {value}"
+        raise MetadataResolutionError(msg)
+
+
+def normalize_catalog_slug(value: str) -> str:
+    """Normalize noisy catalog names into lower snake-case slugs."""
+    return re.sub(r"[^a-z0-9]+", "_", value.strip().casefold()).strip("_")
+
+
+def compact_catalog_slug(value: str) -> str:
+    """Return a catalog slug comparison key that ignores underscores."""
+    return normalize_catalog_slug(value).replace("_", "")
+
+
+def normalize_title_slug(value: str) -> str:
+    """Normalize noisy book titles into lower kebab-case slugs."""
+    return re.sub(r"[^a-z0-9]+", "-", value.strip().casefold()).strip("-")
+
+
+def is_fatal_tool_error(error: MetadataResolutionError) -> bool:
+    """Return whether a tool error should stop the agent immediately."""
+    message = str(error)
+    return message.startswith(
+        (
+            "Unknown audiobook metadata tool",
+            "Audiobook metadata tool is not enabled",
+        ),
+    )
+
+
+def query_terms(query: str) -> tuple[str, ...]:
+    """Return text variants useful for matching noisy audiobook metadata."""
+    normalized = query.strip().casefold()
+    underscore_slug = normalize_catalog_slug(normalized)
+    compact_slug = compact_catalog_slug(normalized)
+    hyphen_slug = normalize_title_slug(normalized)
+    return tuple(dict.fromkeys(term for term in (normalized, underscore_slug, compact_slug, hyphen_slug) if term))
+
+
+def required_string(data: dict[str, object], key: str) -> str:
+    """Read a required string field."""
+    value = data.get(key)
+    if not isinstance(value, str) or not value.strip():
+        msg = f"{key} must be a non-empty string"
+        raise MetadataResolutionError(msg)
+    return value.strip()
+
+
+def required_int(data: dict[str, object], key: str) -> int:
+    """Read a required integer field."""
+    value = data.get(key)
+    if isinstance(value, bool) or not isinstance(value, int):
+        msg = f"{key} must be an integer"
+        raise MetadataResolutionError(msg)
+    return value
+
+
+def required_series_index(data: dict[str, object], key: str) -> float:
+    """Read a required whole-number or half-number series index."""
+    value = data.get(key)
+    if isinstance(value, bool) or not isinstance(value, int | float):
+        msg = f"{key} must be a number"
+        raise MetadataResolutionError(msg)
+    series_index = float(value)
+    if not (series_index * 2).is_integer():
+        msg = f"{key} must be a whole number or .5 increment"
+        raise MetadataResolutionError(msg)
+    return series_index
+
+
+def optional_int(value: object, key: str) -> int | None:
+    """Read an optional integer field."""
+    if value is None:
+        return None
+    if isinstance(value, bool) or not isinstance(value, int):
+        msg = f"{key} must be an integer or null"
+        raise MetadataResolutionError(msg)
+    return value

python/tools/audiobook/metadata_agent.py

@@ -0,0 +1,575 @@
+"""Resolve audiobook metadata with a controlled Ollama tool loop."""
+
+from __future__ import annotations
+
+import json
+import re
+from dataclasses import asdict, dataclass, is_dataclass, replace
+from os import PathLike
+from typing import TYPE_CHECKING
+
+import httpx
+from sqlalchemy.orm import Session
+
+from python.common import utcnow
+from python.tools.audiobook.llm_tool_calling import (
+    CatalogToolRegistry,
+    MetadataResolutionError,
+    normalize_title_slug,
+    optional_int,
+    parse_tool_calls,
+    required_int,
+    required_series_index,
+    required_string,
+    run_tool_calls,
+    validate_catalog_slug,
+    validate_title_slug,
+)
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    from sqlalchemy.engine import Engine
+
+    from python.orm.richie import AudiobookAuthor
+
+FENCED_JSON_PATTERN = re.compile(r"^```(?:json)?\s*(?P<json>.*?)\s*```$", re.IGNORECASE | re.DOTALL)
+
+
+@dataclass(frozen=True)
+class AgentConfig:
+    """Runtime settings for the audiobook metadata agent."""
+
+    model: str = "deepseek-v4-flash:cloud"
+    ollama_chat_url: str = "https://ollama.com/api/chat"
+    http_timeout_seconds: int = 300
+    max_agent_turns: int = 8
+    max_tool_results: int = 10
+    min_confidence: float = 0.85
+    invalid_final_retries: int = 1
+    standalone_series: str = "standalone"
+    tool_names: tuple[str, ...] = (
+        "search_authors",
+        "search_series",
+        "search_books",
+        "ensure_author",
+        "ensure_series",
+        "ensure_book",
+    )
+
+
+@dataclass(frozen=True)
+class StandardBookMetadata:
+    """Canonical metadata for the final audiobook path."""
+
+    author_id: int
+    author: str
+    book_id: int | None
+    title: str
+    series_id: int | None
+    series: str
+    series_index: float
+    confidence: float
+    needs_review: bool
+    evidence: list[str]
+
+
+@dataclass(frozen=True)
+class FinalMetadataFields:
+    """Raw model fields after schema validation."""
+
+    author_id: int
+    book_id: int | None
+    title: str
+    series_id: int | None
+    series_index: float
+    confidence: float
+    evidence: list[str]
+
+
+@dataclass(frozen=True)
+class ResolvedBookFields:
+    """Book fields after optional catalog book resolution."""
+
+    book_id: int | None
+    title: str
+    series_id: int | None
+    series_index: float
+
+
+@dataclass(frozen=True)
+class AgentStepResult:
+    """Outcome from one model response."""
+
+    metadata: StandardBookMetadata | None
+    invalid_final_count: int
+    should_continue: bool
+
+
+def standard_book_metadata(
+    aax_file_name: str,
+    aax_metadata_from_ffprobe: dict[str, str],
+    engine: Engine,
+    log_path: Path,
+    ollama_api_key: str,
+    config: AgentConfig,
+) -> StandardBookMetadata:
+    """Resolve canonical audiobook metadata with the configured Ollama Cloud model."""
+    with Session(engine) as session:
+        registry = CatalogToolRegistry(session, log_path, config, write_agent_log)
+        agent = AudiobookMetadataAgent(
+            registry=registry, log_path=log_path, ollama_api_key=ollama_api_key, config=config
+        )
+        metadata = agent.run(aax_file_name, aax_metadata_from_ffprobe)
+        if metadata.needs_review:
+            session.rollback()
+        else:
+            registry.prune_unused_created_rows(
+                author_id=metadata.author_id,
+                book_id=metadata.book_id,
+                series_id=metadata.series_id,
+            )
+            session.commit()
+        return metadata
+
+
+class AudiobookMetadataAgent:
+    """Ollama-backed metadata resolver with a fixed local tool registry."""
+
+    def __init__(
+        self,
+        *,
+        registry: CatalogToolRegistry,
+        log_path: Path,
+        ollama_api_key: str,
+        config: AgentConfig,
+    ) -> None:
+        """Create an Ollama metadata agent."""
+        self._registry = registry
+        self._log_path = log_path
+        self._ollama_api_key = ollama_api_key
+        self._config = config
+
+    def run(self, aax_file_name: str, aax_metadata_from_ffprobe: dict[str, str]) -> StandardBookMetadata:
+        """Resolve metadata for one AAX file."""
+        messages = [
+            {"role": "system", "content": system_prompt()},
+            {"role": "user", "content": user_prompt(aax_file_name, aax_metadata_from_ffprobe)},
+        ]
+        invalid_final_count = 0
+        result: StandardBookMetadata | None = None
+
+        for turn in range(1, self._config.max_agent_turns + 1):
+            step = self.run_step(messages, turn, invalid_final_count)
+            invalid_final_count = step.invalid_final_count
+            if step.should_continue:
+                continue
+            result = step.metadata
+            break
+
+        if result is None:
+            return self.force_final_response(messages)
+        return result
+
+    def run_step(
+        self,
+        messages: list[dict[str, object]],
+        turn: int,
+        invalid_final_count: int,
+    ) -> AgentStepResult:
+        """Run one model turn and return the next agent-loop action."""
+        data = self.chat(messages, turn)
+        message = data.get("message")
+        if not isinstance(message, dict):
+            return AgentStepResult(
+                metadata=review_metadata("Ollama response did not include a message", self._config),
+                invalid_final_count=invalid_final_count,
+                should_continue=False,
+            )
+
+        try:
+            tool_calls = parse_tool_calls(message)
+        except (json.JSONDecodeError, MetadataResolutionError) as error:
+            return AgentStepResult(
+                metadata=review_metadata(str(error), self._config),
+                invalid_final_count=invalid_final_count,
+                should_continue=False,
+            )
+        if tool_calls:
+            fatal_error = run_tool_calls(messages, message, tool_calls, self._registry, self._log_path, write_agent_log)
+            if fatal_error is not None:
+                return AgentStepResult(
+                    metadata=review_metadata(fatal_error, self._config),
+                    invalid_final_count=invalid_final_count,
+                    should_continue=False,
+                )
+            return AgentStepResult(metadata=None, invalid_final_count=invalid_final_count, should_continue=True)
+        return self.handle_final_message(messages, message, invalid_final_count)
+
+    def handle_final_message(
+        self,
+        messages: list[dict[str, object]],
+        message: dict[str, object],
+        invalid_final_count: int,
+    ) -> AgentStepResult:
+        """Validate a final model message or request one retry."""
+        content = message.get("content")
+        if not isinstance(content, str):
+            return AgentStepResult(
+                metadata=review_metadata("Ollama final response did not include string content", self._config),
+                invalid_final_count=invalid_final_count,
+                should_continue=False,
+            )
+
+        try:
+            resolved = self.validate_final(parse_final_json_content(content))
+        except (json.JSONDecodeError, MetadataResolutionError) as error:
+            return self.handle_invalid_final(messages, error, invalid_final_count)
+
+        write_agent_log(self._log_path, "final_metadata", metadata=resolved)
+        return AgentStepResult(metadata=resolved, invalid_final_count=invalid_final_count, should_continue=False)
+
+    def handle_invalid_final(
+        self,
+        messages: list[dict[str, object]],
+        error: json.JSONDecodeError | MetadataResolutionError,
+        invalid_final_count: int,
+    ) -> AgentStepResult:
+        """Log invalid final JSON and either retry or return review metadata."""
+        invalid_final_count += 1
+        write_agent_log(
+            self._log_path,
+            "final_validation_error",
+            error=str(error),
+            invalid_final_count=invalid_final_count,
+        )
+        if invalid_final_count > self._config.invalid_final_retries:
+            return AgentStepResult(
+                metadata=review_metadata(str(error), self._config),
+                invalid_final_count=invalid_final_count,
+                should_continue=False,
+            )
+        messages.append(
+            {
+                "role": "user",
+                "content": (
+                    "Your previous final answer was invalid. Return only valid JSON matching the required "
+                    f"schema. Validation error: {error}"
+                ),
+            },
+        )
+        return AgentStepResult(metadata=None, invalid_final_count=invalid_final_count, should_continue=True)
+
+    def force_final_response(self, messages: list[dict[str, object]]) -> StandardBookMetadata:
+        """Request a no-tool final answer after the normal turn limit."""
+        messages.append({"role": "user", "content": forced_final_prompt()})
+        write_agent_log(self._log_path, "forced_final_request", reason="max_turns")
+        data = self.chat(messages, self._config.max_agent_turns + 1, tools_enabled=False)
+        message = data.get("message")
+        if not isinstance(message, dict):
+            return review_metadata("Ollama forced final response did not include a message", self._config)
+        content = message.get("content")
+        if not isinstance(content, str):
+            return review_metadata("Ollama forced final response did not include string content", self._config)
+        try:
+            resolved = self.validate_final(parse_final_json_content(content))
+        except (json.JSONDecodeError, MetadataResolutionError) as error:
+            return review_metadata(f"Ollama forced final response was invalid: {error}", self._config)
+        write_agent_log(self._log_path, "final_metadata", metadata=resolved)
+        return resolved
+
+    def chat(self, messages: list[dict[str, object]], turn: int, *, tools_enabled: bool = True) -> dict[str, object]:
+        """Send one chat request to Ollama and log the request and response."""
+        payload = {
+            "model": self._config.model,
+            "messages": messages,
+            "stream": False,
+            "options": {"temperature": 0.1},
+        }
+        tool_names = []
+        if tools_enabled:
+            payload["tools"] = self._registry.tool_schemas()
+            tool_names = self._config.tool_names
+        write_agent_log(
+            self._log_path,
+            "model_request",
+            model=self._config.model,
+            turn=turn,
+            message_count=len(messages),
+            tool_names=tool_names,
+            tools_enabled=tools_enabled,
+        )
+        write_agent_log(
+            self._log_path,
+            "llm_messages_sent",
+            model=self._config.model,
+            turn=turn,
+            messages=messages,
+            tools_enabled=tools_enabled,
+        )
+        response = httpx.post(
+            self._config.ollama_chat_url,
+            headers={"Authorization": f"Bearer {self._ollama_api_key}"},
+            json=payload,
+            timeout=self._config.http_timeout_seconds,
+        )
+        response.raise_for_status()
+        raw_data = response.json()
+        if not isinstance(raw_data, dict):
+            return {}
+        data = {str(key): value for key, value in raw_data.items()}
+        message = data.get("message", {})
+        content = message.get("content") if isinstance(message, dict) else ""
+        write_agent_log(
+            self._log_path,
+            "llm_message_received",
+            model=self._config.model,
+            turn=turn,
+            message=message,
+        )
+        write_agent_log(
+            self._log_path,
+            "model_response",
+            model=self._config.model,
+            turn=turn,
+            has_tool_calls=bool(isinstance(message, dict) and message.get("tool_calls")),
+            content_chars=len(content) if isinstance(content, str) else 0,
+        )
+        return data
+
+    def validate_final(self, raw_metadata: object) -> StandardBookMetadata:
+        """Validate final model metadata against catalog rows."""
+        fields = parse_final_metadata_fields(raw_metadata)
+        fields = replace(fields, title=normalize_title_slug(fields.title))
+        author = self.validate_author(fields.author_id)
+        validate_title_slug(fields.title)
+        book_fields = self.resolve_book_fields(fields)
+        series = self.validate_series(fields.author_id, book_fields.series_id, book_fields.series_index)
+
+        return StandardBookMetadata(
+            author_id=fields.author_id,
+            author=author.name,
+            book_id=book_fields.book_id,
+            title=book_fields.title,
+            series_id=book_fields.series_id,
+            series=series,
+            series_index=book_fields.series_index,
+            confidence=fields.confidence,
+            needs_review=fields.confidence < self._config.min_confidence,
+            evidence=fields.evidence,
+        )
+
+    def validate_author(self, author_id: int) -> AudiobookAuthor:
+        """Validate that an author id was seen and exists."""
+        if author_id not in self._registry.seen_author_ids:
+            msg = f"author_id {author_id} was not returned by search_authors"
+            raise MetadataResolutionError(msg)
+        author = self._registry.get_author(author_id)
+        if author is None:
+            msg = f"author_id {author_id} does not exist"
+            raise MetadataResolutionError(msg)
+        validate_catalog_slug(author.name, "author")
+        return author
+
+    def resolve_book_fields(self, fields: FinalMetadataFields) -> ResolvedBookFields:
+        """Resolve final book fields from a seen book id or created book."""
+        if fields.book_id is None:
+            ensured = self._registry.ensure_book(
+                fields.title,
+                fields.author_id,
+                fields.series_id,
+                fields.series_index,
+            )
+            return ResolvedBookFields(
+                book_id=ensured.book.id,
+                title=ensured.book.title,
+                series_id=ensured.book.series_id,
+                series_index=ensured.book.series_index,
+            )
+
+        if fields.book_id not in self._registry.seen_book_ids:
+            msg = f"book_id {fields.book_id} was not returned by search_books"
+            raise MetadataResolutionError(msg)
+        book = self._registry.get_book(fields.book_id)
+        if book is None:
+            msg = f"book_id {fields.book_id} does not exist"
+            raise MetadataResolutionError(msg)
+        if book.author_id != fields.author_id:
+            msg = f"book_id {fields.book_id} does not belong to author_id {fields.author_id}"
+            raise MetadataResolutionError(msg)
+        return ResolvedBookFields(
+            book_id=fields.book_id,
+            title=book.title,
+            series_id=book.series_id,
+            series_index=book.series_index,
+        )
+
+    def validate_series(self, author_id: int, series_id: int | None, series_index: float) -> str:
+        """Validate final series fields and return the canonical series slug."""
+        if series_id is None:
+            if series_index != 0:
+                msg = "standalone books must use series_index 0"
+                raise MetadataResolutionError(msg)
+            return self._config.standalone_series
+
+        if series_id not in self._registry.seen_series_ids:
+            msg = f"series_id {series_id} was not returned by search_series"
+            raise MetadataResolutionError(msg)
+        series = self._registry.get_series(series_id)
+        if series is None:
+            msg = f"series_id {series_id} does not exist"
+            raise MetadataResolutionError(msg)
+        if series.author_id != author_id:
+            msg = f"series_id {series_id} does not belong to author_id {author_id}"
+            raise MetadataResolutionError(msg)
+        if series_index <= 0:
+            msg = "series books must use a positive series_index"
+            raise MetadataResolutionError(msg)
+        validate_catalog_slug(series.name, "series")
+        return series.name
+
+
+def write_agent_log(log_path: Path, event: str, **fields: object) -> None:
+    """Append one JSONL audit event."""
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    record = {
+        "created": utcnow().isoformat(),
+        "event": event,
+        **{key: json_log_value(value) for key, value in fields.items()},
+    }
+    with log_path.open("a", encoding="utf-8") as file:
+        file.write(json.dumps(record, sort_keys=True))
+        file.write("\n")
+
+
+def json_log_value(value: object) -> object:
+    """Return a JSON-serializable value for audit logs."""
+    if is_dataclass(value) and not isinstance(value, type):
+        return json_log_value(asdict(value))
+    if isinstance(value, dict):
+        return {str(key): json_log_value(item) for key, item in value.items()}
+    if isinstance(value, list | tuple):
+        return [json_log_value(item) for item in value]
+    if isinstance(value, set):
+        return [json_log_value(item) for item in sorted(value, key=str)]
+    if isinstance(value, PathLike):
+        return str(value)
+    return value
+
+
+def system_prompt() -> str:
+    """Return the stable system prompt."""
+    return """You standardize Audible audiobook metadata against a private catalog.
+
+Rules:
+- You must use the provided tools before returning final metadata.
+- Only use author_id, series_id, or book_id values returned by tools.
+- Return final metadata as JSON only. Do not wrap it in Markdown.
+- The final JSON object must contain author_id, book_id, title, series_id, series_index, confidence, and evidence.
+- title must be a canonical title slug using lower-case words separated by hyphens.
+- Use series_id null and series_index 0 for standalone books.
+- If you use a series_id, series_index must be a whole number or .5 value greater than 0.
+- Treat series slugs that differ only by underscores as the same series. Prefer the existing catalog row instead of
+  creating a new series.
+- Detect omnibus or box-set editions that contain multiple numbered novels, books, or novellas.
+- For an omnibus, make a best-effort range from the filename, tags, and catalog rows. Keep series_index as the
+  first covered book number and include the range in the title when the source title includes it, for example
+  books-1-3.
+- Be careful with omnibuses of novels or novellas later published as one book: keep the omnibus as the audiobook's
+  book record unless catalog rows clearly identify a better match.
+- Do not create publisher collections or author collections as series unless the book metadata clearly gives a
+  numbered series.
+- Series belong to authors. Use a series_id only when it belongs to the selected author_id.
+- Always search for the author before creating one. If no exact author slug exists, call ensure_author.
+- Always search for a series with author_id before creating one. If no exact series slug exists, call ensure_series.
+- Always search for a book before creating one. If no exact title slug exists, call ensure_book.
+- If a tool returns an error, correct your tool arguments or final metadata before continuing.
+- confidence must be a number from 0 to 1.
+- evidence must be a short list of strings explaining which filename, tags, and catalog rows support the answer."""
+
+
+def forced_final_prompt() -> str:
+    """Return the no-tools finalization prompt."""
+    return (
+        "Stop calling tools. Return final metadata as JSON only using the tool results already provided. "
+        "If search_books returned no matching rows but author and series are known, use book_id null and resolve "
+        "the title slug from the AAX filename and ffprobe tags. The validator will create the missing book. "
+        "Use only author_id and series_id values returned by earlier tool results."
+    )
+
+
+def user_prompt(aax_file_name: str, metadata: dict[str, str]) -> str:
+    """Build the user prompt from source metadata."""
+    return (
+        "Resolve this Audible audiobook.\n\n"
+        f"AAX file name: {aax_file_name}\n\n"
+        "ffprobe format tags:\n"
+        f"{json.dumps(metadata, indent=2, sort_keys=True)}"
+    )
+
+
+def parse_final_json_content(content: str) -> object:
+    """Parse final model content, accepting bare or fenced JSON."""
+    stripped = content.strip()
+    if match := FENCED_JSON_PATTERN.fullmatch(stripped):
+        stripped = match.group("json").strip()
+    return json.loads(stripped)
+
+
+def parse_final_metadata_fields(raw_metadata: object) -> FinalMetadataFields:
+    """Parse the model's final JSON object into typed fields."""
+    if not isinstance(raw_metadata, dict):
+        msg = "Final metadata must be a JSON object"
+        raise MetadataResolutionError(msg)
+    data = {str(key): value for key, value in raw_metadata.items()}
+    return FinalMetadataFields(
+        author_id=required_int(data, "author_id"),
+        book_id=optional_int(data.get("book_id"), "book_id"),
+        title=required_string(data, "title"),
+        series_id=optional_int(data.get("series_id"), "series_id"),
+        series_index=required_series_index(data, "series_index"),
+        confidence=required_float(data, "confidence"),
+        evidence=required_string_list(data, "evidence"),
+    )
+
+
+def review_metadata(reason: str, config: AgentConfig) -> StandardBookMetadata:
+    """Return a metadata result that must be reviewed manually."""
+    return StandardBookMetadata(
+        author_id=0,
+        author="unknown_author",
+        book_id=None,
+        title="unknown-title",
+        series_id=None,
+        series=config.standalone_series,
+        series_index=0,
+        confidence=0,
+        needs_review=True,
+        evidence=[reason],
+    )
+
+
+def required_float(data: dict[str, object], key: str) -> float:
+    """Read a required float field."""
+    value = data.get(key)
+    if isinstance(value, bool) or not isinstance(value, int | float):
+        msg = f"{key} must be a number"
+        raise MetadataResolutionError(msg)
+    confidence = float(value)
+    if confidence < 0 or confidence > 1:
+        msg = f"{key} must be between 0 and 1"
+        raise MetadataResolutionError(msg)
+    return confidence
+
+
+def required_string_list(data: dict[str, object], key: str) -> list[str]:
+    """Read a required list of strings."""
+    value = data.get(key)
+    if not isinstance(value, list) or not value or not all(isinstance(item, str) for item in value):
+        msg = f"{key} must be a non-empty list of strings"
+        raise MetadataResolutionError(msg)
+    strings = [item.strip() for item in value if item.strip()]
+    if not strings:
+        msg = f"{key} must include at least one non-empty string"
+        raise MetadataResolutionError(msg)
+    return strings

python/tools/snapshot_manager.py

@@ -34,8 +34,9 @@ def main(config_file: Path) -> None:
                 logger.error(msg)
                 signal_alert(msg)
                 continue
-
-            get_snapshots_to_delete(dataset, get_count_lookup(config_file, dataset.name))
+            count_lookup = get_count_lookup(config_file, dataset.name)
+            logger.info(f"using {count_lookup} for {dataset.name}")
+            get_snapshots_to_delete(dataset, count_lookup)
     except Exception:
         logger.exception("snapshot_manager failed")
         signal_alert("snapshot_manager failed")
@@ -99,6 +100,7 @@ def get_snapshots_to_delete(
     """
     snapshots = dataset.get_snapshots()
 
+    logger.info(f"calculating snapshots for {dataset.name} to be deleted")
     if not snapshots:
         logger.info(f"{dataset.name} has no snapshots")
         return

python/tools/whisper/Dockerfile

@@ -0,0 +1,17 @@
+FROM nvidia/cuda:12.4.1-cudnn-runtime-ubuntu22.04
+
+ENV DEBIAN_FRONTEND=noninteractive \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1
+
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends python3 python3-pip ffmpeg \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN pip3 install --no-cache-dir --upgrade pip \
+    && pip3 install --no-cache-dir faster-whisper requests
+
+WORKDIR /app
+COPY python/tools/whisper/inference.py /app/inference.py
+
+ENTRYPOINT ["python3", "/app/inference.py"]

python/tools/whisper/Dockerfile.dockerignore

@@ -0,0 +1,2 @@
+*
+!python/tools/whisper/inference.py

python/tools/whisper/__init__.py

@@ -0,0 +1 @@
+"""Whisper transcription tools (host orchestrator and container entrypoint)."""

python/tools/whisper/inference.py

@@ -0,0 +1,136 @@
+"""Container entrypoint that transcribes a directory of audio files with faster-whisper.
+
+Run inside the whisper-transcribe docker image; segment timestamps are grouped
+into one-minute buckets so the output reads as ``[HH:MM:00] text``.
+"""
+
+from __future__ import annotations
+
+import argparse
+import logging
+from pathlib import Path
+
+from faster_whisper import WhisperModel
+
+logger = logging.getLogger(__name__)
+
+AUDIO_EXTENSIONS = {".mp3", ".wav", ".m4a", ".flac", ".ogg", ".opus", ".mp4", ".mkv", ".webm", ".aac"}
+BUCKET_SECONDS = 60
+BEAM_SIZE = 5
+SECONDS_PER_HOUR = 3600
+SECONDS_PER_MINUTE = 60
+
+
+def format_timestamp(total_seconds: float) -> str:
+    """Render a whole-minute timestamp as ``HH:MM:00``.
+
+    Args:
+        total_seconds: Offset in seconds from the start of the audio.
+
+    Returns:
+        A zero-padded ``HH:MM:00`` string.
+    """
+    hours = int(total_seconds // SECONDS_PER_HOUR)
+    minutes = int((total_seconds % SECONDS_PER_HOUR) // SECONDS_PER_MINUTE)
+    return f"{hours:02d}:{minutes:02d}:00"
+
+
+def transcribe_file(model: WhisperModel, audio_path: Path, output_path: Path) -> None:
+    """Transcribe one audio file and write the bucketed transcript to disk.
+
+    Args:
+        model: Loaded faster-whisper model.
+        audio_path: Source audio file.
+        output_path: Destination ``.txt`` path.
+    """
+    logger.info("Transcribing %s", audio_path)
+    segments, info = model.transcribe(
+        str(audio_path),
+        language="en",
+        beam_size=BEAM_SIZE,
+        vad_filter=True,
+    )
+    logger.info("Duration %.1fs", info.duration)
+
+    buckets: dict[int, list[str]] = {}
+    for segment in segments:
+        bucket = int(segment.start // BUCKET_SECONDS)
+        buckets.setdefault(bucket, []).append(segment.text.strip())
+
+    lines = [f"[{format_timestamp(bucket * BUCKET_SECONDS)}] {' '.join(buckets[bucket])}" for bucket in sorted(buckets)]
+    output_path.write_text("\n\n".join(lines) + "\n", encoding="utf-8")
+    logger.info("Wrote %s", output_path)
+
+
+def find_audio_files(input_directory: Path) -> list[Path]:
+    """Collect every audio file under ``input_directory``.
+
+    Args:
+        input_directory: Directory to walk recursively.
+
+    Returns:
+        Sorted list of audio file paths.
+    """
+    return sorted(
+        path for path in input_directory.rglob("*") if path.is_file() and path.suffix.lower() in AUDIO_EXTENSIONS
+    )
+
+
+def configure_container_logger() -> None:
+    """Configure logging for the container (stdout, INFO)."""
+    logging.basicConfig(
+        level=logging.INFO,
+        format="%(asctime)s %(levelname)s %(message)s",
+    )
+
+
+def parse_arguments() -> argparse.Namespace:
+    """Parse CLI arguments for the container entrypoint.
+
+    Returns:
+        Parsed argparse namespace.
+    """
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--input", type=Path, default=Path("/audio"))
+    parser.add_argument("--output", type=Path, default=Path("/output"))
+    parser.add_argument("--model", default="large-v3")
+    parser.add_argument(
+        "--download-only",
+        action="store_true",
+        help="Download the model into the cache volume and exit without transcribing.",
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    """Load the model, then either exit (download-only) or transcribe the directory."""
+    configure_container_logger()
+    arguments = parse_arguments()
+
+    logger.info("Loading model %s on CUDA", arguments.model)
+    model = WhisperModel(arguments.model, device="cuda", compute_type="float16")
+
+    if arguments.download_only:
+        logger.info("Model ready; exiting (download-only mode)")
+        return
+
+    arguments.output.mkdir(parents=True, exist_ok=True)
+
+    audio_files = find_audio_files(arguments.input)
+    if not audio_files:
+        logger.warning("No audio files found in %s", arguments.input)
+        return
+
+    logger.info("Found %d audio file(s)", len(audio_files))
+    for audio_path in audio_files:
+        relative = audio_path.relative_to(arguments.input)
+        output_path = arguments.output / relative.with_suffix(".txt")
+        output_path.parent.mkdir(parents=True, exist_ok=True)
+        if output_path.exists():
+            logger.info("Skip %s (already transcribed)", relative)
+            continue
+        transcribe_file(model, audio_path, output_path)
+
+
+if __name__ == "__main__":
+    main()

python/tools/whisper/transcribe.py

@@ -0,0 +1,167 @@
+"""Build and run the whisper transcription docker container on demand.
+
+The container is started fresh for each invocation and removed on exit
+(``docker run --rm``). The model is cached in a named docker volume so
+only the first run pays the download cost.
+"""
+
+from __future__ import annotations
+
+import logging
+import subprocess
+from pathlib import Path
+from typing import Annotated
+
+import typer
+
+from python.common import configure_logger
+
+logger = logging.getLogger(__name__)
+
+
+class Config:
+    """Paths and names for the whisper-transcribe Docker workflow."""
+
+    image_tag = "whisper-transcribe:latest"
+    model_volume = "whisper-models"
+    repo_root = Path(__file__).resolve().parents[3]
+    dockerfile = Path(__file__).resolve().parent / "Dockerfile"
+    huggingface_cache = "/root/.cache/huggingface"
+
+
+def run_docker(arguments: list[str]) -> None:
+    """Run a docker subcommand, streaming output and raising on failure.
+
+    Args:
+        arguments: Arguments to pass to the ``docker`` binary.
+
+    Raises:
+        subprocess.CalledProcessError: If docker exits non-zero.
+    """
+    logger.info("docker %s", " ".join(arguments))
+    subprocess.run(["docker", *arguments], check=True)
+
+
+def build_image() -> None:
+    """Build the whisper-transcribe image using the repo root as build context."""
+    logger.info("Building image %s", Config.image_tag)
+    run_docker(
+        [
+            "build",
+            "--tag",
+            Config.image_tag,
+            "--file",
+            str(Config.dockerfile),
+            str(Config.repo_root),
+        ],
+    )
+
+
+def model_cache_present(model: str) -> bool:
+    """Check whether the given model is already downloaded in the cache volume.
+
+    Args:
+        model: faster-whisper model name (e.g. ``large-v3``).
+
+    Returns:
+        True if the HuggingFace cache directory for the model exists in the volume.
+    """
+    cache_directory = f"hub/models--Systran--faster-whisper-{model}"
+    completed = subprocess.run(
+        [
+            "docker",
+            "run",
+            "--rm",
+            "--volume",
+            f"{Config.model_volume}:/cache",
+            "alpine",
+            "test",
+            "-d",
+            f"/cache/{cache_directory}",
+        ],
+        check=False,
+    )
+    return completed.returncode == 0
+
+
+def download_model(model: str) -> None:
+    """Download the model into the cache volume and exit.
+
+    Args:
+        model: faster-whisper model name.
+    """
+    logger.info("Downloading model %s into volume %s", model, Config.model_volume)
+    run_docker(
+        [
+            "run",
+            "--rm",
+            "--device=nvidia.com/gpu=all",
+            "--ipc=host",
+            "--volume",
+            f"{Config.model_volume}:{Config.huggingface_cache}",
+            Config.image_tag,
+            "--model",
+            model,
+            "--download-only",
+        ],
+    )
+
+
+def transcribe(input_directory: Path, output_directory: Path, model: str) -> None:
+    """Run transcription on every audio file under ``input_directory``.
+
+    Args:
+        input_directory: Host path containing audio files (mounted read-only).
+        output_directory: Host path for ``.txt`` transcripts.
+        model: faster-whisper model name.
+    """
+    logger.info("Transcribing %s -> %s (model=%s)", input_directory, output_directory, model)
+    run_docker(
+        [
+            "run",
+            "--rm",
+            "--device=nvidia.com/gpu=all",
+            "--ipc=host",
+            "--volume",
+            f"{input_directory}:/audio:ro",
+            "--volume",
+            f"{output_directory}:/output",
+            "--volume",
+            f"{Config.model_volume}:{Config.huggingface_cache}",
+            Config.image_tag,
+            "--model",
+            model,
+        ],
+    )
+
+
+def main(
+    input_directory: Annotated[Path, typer.Argument(help="Directory of audio files to transcribe.")],
+    output_directory: Annotated[Path, typer.Argument(help="Directory to write .txt transcripts to.")],
+    model: Annotated[str, typer.Option(help="faster-whisper model name.")] = "large-v3",
+    *,
+    force_download: Annotated[
+        bool,
+        typer.Option("--force-download", help="Re-download the model even if already cached."),
+    ] = False,
+) -> None:
+    """Build the image, ensure the model is cached, then transcribe and stop."""
+    configure_logger()
+
+    resolved_input = input_directory.resolve(strict=True)
+    output_directory.mkdir(parents=True, exist_ok=True)
+    resolved_output = output_directory.resolve()
+
+    build_image()
+
+    if force_download or not model_cache_present(model):
+        download_model(model)
+    else:
+        logger.info("Model %s already cached in volume %s", model, Config.model_volume)
+
+    transcribe(resolved_input, resolved_output, model)
+    logger.info("Done. Container stopped.")
+
+
+if __name__ == "__main__":
+    typer.run(main)

systems/bob/default.nix

@@ -1,28 +1,39 @@
-{ inputs, ... }:
+{ inputs, pkgs, ... }:
 {
   imports = [
+    "${inputs.self}/users/math"
     "${inputs.self}/users/richie"
+    "${inputs.self}/users/steve"
     "${inputs.self}/common/global"
-    "${inputs.self}/common/optional/desktop.nix"
     "${inputs.self}/common/optional/docker.nix"
     "${inputs.self}/common/optional/scanner.nix"
+    "${inputs.self}/common/optional/monitoring-agent.nix"
     "${inputs.self}/common/optional/steam.nix"
     "${inputs.self}/common/optional/syncthing_base.nix"
     "${inputs.self}/common/optional/systemd-boot.nix"
     "${inputs.self}/common/optional/update.nix"
     "${inputs.self}/common/optional/yubikey.nix"
     "${inputs.self}/common/optional/zerotier.nix"
-    "${inputs.self}/common/optional/brain_substituter.nix"
     "${inputs.self}/common/optional/nvidia.nix"
     ./hardware.nix
     ./syncthing.nix
     ./llms.nix
   ];
 
+  boot = {
+    kernelPackages = pkgs.linuxPackages_6_18;
+    zfs.package = pkgs.zfs_2_4;
+  };
+
   networking = {
     hostName = "bob";
     hostId = "7c678a41";
-    firewall.enable = true;
+    firewall = {
+      enable = true;
+      allowedTCPPorts = [
+        8000
+      ];
+    };
     networkmanager.enable = true;
   };
 

systems/bob/hardware.nix

@@ -28,9 +28,13 @@
         allowDiscards = true;
         keyFileSize = 4096;
         keyFile = "/dev/disk/by-id/usb-Samsung_Flash_Drive_FIT_0374620080067131-0:0";
-        fallbackToPassword = true;
       };
     };
+
+    zfs.extraPools = [
+      "storage"
+    ];
+
     kernelModules = [ "kvm-amd" ];
     extraModulePackages = [ ];
   };

systems/bob/llms.nix

@@ -23,6 +23,7 @@
       "magistral:24b"
       "ministral-3:14b"
       "nemotron-3-nano:30b"
+      "nemotron-3-nano:4b"
       "nemotron-cascade-2:30b"
       "qwen3-coder:30b"
       "qwen3-embedding:0.6b"
@@ -41,11 +42,14 @@
       "qwen3:8b"
       "qwen3.5:27b"
       "qwen3.5:35b"
+      "qwen3.6:27b"
+      "qwen3.6:35b"
+      "rinex20/translategemma3:12b"
       "translategemma:12b"
       "translategemma:27b"
       "translategemma:4b"
     ];
-    models = "/zfs/models";
+    models = "/zfs/storage/models";
     openFirewall = true;
   };
 }

systems/bob/syncthing.nix

@@ -31,5 +31,15 @@
       ];
       fsWatcherEnabled = true;
     };
+    "recordings" = {
+      path = "/home/richie/recordings";
+      devices = [
+        "jeeves"
+        "phone"
+        "rhapsody-in-green"
+      ];
+      fsWatcherEnabled = true;
+    };
+
   };
 }

systems/brain/hardware.nix

@@ -26,7 +26,6 @@
         allowDiscards = true;
         keyFileSize = 4096;
         keyFile = "/dev/disk/by-id/usb-USB_SanDisk_3.2Gen1_03021630090925173333-0:0";
-        fallbackToPassword = true;
       };
     };
     kernelModules = [ "kvm-intel" ];

systems/jeeves/default.nix

@@ -4,17 +4,21 @@ let
 in
 {
   imports = [
-    "${inputs.self}/users/richie"
-    "${inputs.self}/users/math"
     "${inputs.self}/users/dov"
+    "${inputs.self}/users/math"
+    "${inputs.self}/users/richie"
+    "${inputs.self}/users/steve"
     "${inputs.self}/common/global"
     "${inputs.self}/common/optional/docker.nix"
+    "${inputs.self}/common/optional/monitoring-agent.nix"
     "${inputs.self}/common/optional/ssh_decrypt.nix"
     "${inputs.self}/common/optional/syncthing_base.nix"
     "${inputs.self}/common/optional/update.nix"
     "${inputs.self}/common/optional/zerotier.nix"
+    ./monitoring
     ./docker
     ./services
+    ./web_services
     ./hardware.nix
     ./networking.nix
     ./programs.nix
@@ -35,5 +39,10 @@ in
     zerotierone.joinNetworks = [ "a09acf02330d37b9" ];
   };
 
+  users.groups = {
+    nornsight = { };
+    nornsight-admin = { };
+  };
+
   system.stateVersion = "24.05";
 }

systems/jeeves/hardware.nix

@@ -9,7 +9,6 @@ let
     inherit device;
     keyFileSize = 4096;
     keyFile = "/dev/disk/by-id/usb-XIAO_USB_Drive_24587CE29074-0:0";
-    fallbackToPassword = true;
   };
   makeLuksSSD =
     device:

systems/jeeves/monitoring/dashboards/overview.json

@@ -0,0 +1,426 @@
+{
+  "annotations": {
+    "list": [
+      {
+        "builtIn": 1,
+        "datasource": {
+          "type": "grafana",
+          "uid": "-- Grafana --"
+        },
+        "enable": true,
+        "hide": true,
+        "iconColor": "rgba(0, 211, 255, 1)",
+        "name": "Annotations & Alerts",
+        "type": "dashboard"
+      }
+    ]
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "links": [],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "100 * (1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m])))",
+          "legendFormat": "{{instance}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "CPU Used",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 6,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "100 * (1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes))",
+          "legendFormat": "{{instance}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "RAM Used",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 12,
+        "y": 0
+      },
+      "id": 3,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "100 * (1 - (node_memory_SwapFree_bytes / node_memory_SwapTotal_bytes))",
+          "legendFormat": "{{instance}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Swap Used",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 6,
+        "x": 18,
+        "y": 0
+      },
+      "id": 4,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "node_load1",
+          "legendFormat": "{{instance}} load1",
+          "range": true,
+          "refId": "A"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "node_load5",
+          "legendFormat": "{{instance}} load5",
+          "range": true,
+          "refId": "B"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "node_load15",
+          "legendFormat": "{{instance}} load15",
+          "range": true,
+          "refId": "C"
+        }
+      ],
+      "title": "Load",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 0,
+        "y": 8
+      },
+      "id": 5,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "sum by (instance) (rate(node_disk_read_bytes_total[5m]))",
+          "legendFormat": "{{instance}} read",
+          "range": true,
+          "refId": "A"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "sum by (instance) (rate(node_disk_written_bytes_total[5m]))",
+          "legendFormat": "{{instance}} write",
+          "range": true,
+          "refId": "B"
+        }
+      ],
+      "title": "Disk Throughput",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 12,
+        "y": 8
+      },
+      "id": 6,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "100 * (1 - (node_filesystem_avail_bytes{mountpoint=~\"(/|/home|/var|/zfs.*)\",fstype!=\"\"} / node_filesystem_size_bytes{mountpoint=~\"(/|/home|/var|/zfs.*)\",fstype!=\"\"}))",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{mountpoint}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Filesystem Usage",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 0,
+        "y": 17
+      },
+      "id": 7,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, rate(namedprocess_namegroup_cpu_seconds_total[5m]))",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top Grouped CPU",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 12,
+        "y": 17
+      },
+      "id": 8,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, namedprocess_namegroup_memory_bytes{memtype=\"resident\"})",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top Grouped Memory",
+      "type": "table"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "monitoring"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "timepicker": {},
+  "timezone": "",
+  "title": "Overview",
+  "uid": "monitor-overview",
+  "version": 1,
+  "weekStart": ""
+}

systems/jeeves/monitoring/dashboards/process-history-grouped.json

@@ -0,0 +1,216 @@
+{
+  "annotations": {
+    "list": [
+      {
+        "builtIn": 1,
+        "datasource": {
+          "type": "grafana",
+          "uid": "-- Grafana --"
+        },
+        "enable": true,
+        "hide": true,
+        "iconColor": "rgba(0, 211, 255, 1)",
+        "name": "Annotations & Alerts",
+        "type": "dashboard"
+      }
+    ]
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "links": [],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, rate(namedprocess_namegroup_cpu_seconds_total[5m]))",
+          "legendFormat": "{{instance}} {{groupname}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Grouped CPU",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 12,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, namedprocess_namegroup_memory_bytes{memtype=\"resident\"})",
+          "legendFormat": "{{instance}} {{groupname}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Grouped Resident Memory",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 0,
+        "y": 10
+      },
+      "id": 3,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, rate(namedprocess_namegroup_read_bytes_total[5m]))",
+          "legendFormat": "{{instance}} {{groupname}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Grouped Read I/O",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 12,
+        "y": 10
+      },
+      "id": 4,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(10, rate(namedprocess_namegroup_write_bytes_total[5m]))",
+          "legendFormat": "{{instance}} {{groupname}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Grouped Write I/O",
+      "type": "timeseries"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "monitoring",
+    "process"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-7d",
+    "to": "now"
+  },
+  "timepicker": {},
+  "timezone": "",
+  "title": "Process History Grouped",
+  "uid": "monitor-process-history",
+  "version": 1,
+  "weekStart": ""
+}

systems/jeeves/monitoring/dashboards/process-live-pid.json

@@ -0,0 +1,224 @@
+{
+  "annotations": {
+    "list": [
+      {
+        "builtIn": 1,
+        "datasource": {
+          "type": "grafana",
+          "uid": "-- Grafana --"
+        },
+        "enable": true,
+        "hide": true,
+        "iconColor": "rgba(0, 211, 255, 1)",
+        "name": "Annotations & Alerts",
+        "type": "dashboard"
+      }
+    ]
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "links": [],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-pid-short"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-pid-short"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, rate(namedprocess_namegroup_cpu_seconds_total[2m]))",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top PID CPU",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-pid-short"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 12,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-pid-short"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, namedprocess_namegroup_memory_bytes{memtype=\"resident\"})",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top PID RSS",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-pid-short"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 0,
+        "y": 10
+      },
+      "id": 3,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-pid-short"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, rate(namedprocess_namegroup_read_bytes_total[2m]))",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top PID Read I/O",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-pid-short"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "Bps"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 10,
+        "w": 12,
+        "x": 12,
+        "y": 10
+      },
+      "id": 4,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-pid-short"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, rate(namedprocess_namegroup_write_bytes_total[2m]))",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{groupname}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top PID Write I/O",
+      "type": "table"
+    }
+  ],
+  "refresh": "15s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "monitoring",
+    "process"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-10m",
+    "to": "now"
+  },
+  "timepicker": {},
+  "timezone": "",
+  "title": "Process Live PID",
+  "uid": "monitor-process-pid",
+  "version": 1,
+  "weekStart": ""
+}

systems/jeeves/monitoring/dashboards/storage-zfs.json

@@ -0,0 +1,351 @@
+{
+  "annotations": {
+    "list": [
+      {
+        "builtIn": 1,
+        "datasource": {
+          "type": "grafana",
+          "uid": "-- Grafana --"
+        },
+        "enable": true,
+        "hide": true,
+        "iconColor": "rgba(0, 211, 255, 1)",
+        "name": "Annotations & Alerts",
+        "type": "dashboard"
+      }
+    ]
+  },
+  "editable": true,
+  "fiscalYearStartMonth": 0,
+  "graphTooltip": 0,
+  "links": [],
+  "panels": [
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 8,
+        "x": 0,
+        "y": 0
+      },
+      "id": 1,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "100 * (zfs_pool_allocated_bytes / zfs_pool_size_bytes)",
+          "legendFormat": "{{instance}} {{pool}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Pool Usage",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 8,
+        "x": 8,
+        "y": 0
+      },
+      "id": 2,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "zfs_pool_free_bytes",
+          "legendFormat": "{{instance}} {{pool}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Pool Free Bytes",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "bytes"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 8,
+        "x": 16,
+        "y": 0
+      },
+      "id": 3,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, zfs_dataset_used_bytes{type=\"filesystem\"})",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{name}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Top Filesystems by Used Bytes",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "ns"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 0,
+        "y": 8
+      },
+      "id": 4,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, zpool_iostat_total_wait_read_ns{vdev!=\"_pool\"})",
+          "legendFormat": "{{host}} {{pool}} {{vdev}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "ZFS Read Wait",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "ns"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 12,
+        "y": 8
+      },
+      "id": 5,
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "topk(20, zpool_iostat_total_wait_write_ns{vdev!=\"_pool\"})",
+          "legendFormat": "{{host}} {{pool}} {{vdev}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "ZFS Write Wait",
+      "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "celsius"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 0,
+        "y": 17
+      },
+      "id": 6,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": true,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "smartctl_device_temperature{temperature_type=\"current\"}",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{device}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Disk Temperature",
+      "type": "table"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "prom-main"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 12,
+        "y": 17
+      },
+      "id": 7,
+      "options": {
+        "cellHeight": "sm",
+        "showHeader": true,
+        "sortBy": [
+          {
+            "desc": false,
+            "displayName": "Value"
+          }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "prom-main"
+          },
+          "editorMode": "code",
+          "expr": "smartctl_device_smart_status",
+          "format": "table",
+          "instant": true,
+          "legendFormat": "{{instance}} {{device}}",
+          "refId": "A"
+        }
+      ],
+      "title": "SMART Health",
+      "type": "table"
+    }
+  ],
+  "refresh": "30s",
+  "schemaVersion": 39,
+  "style": "dark",
+  "tags": [
+    "monitoring",
+    "zfs"
+  ],
+  "templating": {
+    "list": []
+  },
+  "time": {
+    "from": "now-24h",
+    "to": "now"
+  },
+  "timepicker": {},
+  "timezone": "",
+  "title": "Storage and ZFS",
+  "uid": "monitor-storage",
+  "version": 1,
+  "weekStart": ""
+}

systems/jeeves/monitoring/default.nix

@@ -0,0 +1,186 @@
+{
+  lib,
+  pkgs,
+  ...
+}:
+let
+  vars = import ../vars.nix;
+
+  prometheusDataRoot = "${vars.database}/prometheus";
+  mainPrometheusDataDir = "${prometheusDataRoot}/main";
+  pidPrometheusDataDir = "${prometheusDataRoot}/pid-short";
+
+  prometheusYaml = pkgs.formats.yaml { };
+
+  mkPrometheusConfig =
+    name: cfg:
+    let
+      configFile = prometheusYaml.generate "${name}.yaml" cfg;
+    in
+    pkgs.runCommand "${name}-checked.yaml"
+      {
+        nativeBuildInputs = [ pkgs.prometheus.cli ];
+      }
+      ''
+        promtool check config ${configFile}
+        cp ${configFile} $out
+      '';
+
+  mkTarget = host: address: {
+    targets = [ address ];
+    labels.instance = host;
+  };
+
+  mainPrometheusConfig = mkPrometheusConfig "prometheus-main" {
+    global = {
+      scrape_interval = "30s";
+      scrape_timeout = "10s";
+      evaluation_interval = "30s";
+    };
+    scrape_configs = [
+      {
+        job_name = "node";
+        static_configs = [
+          (mkTarget "jeeves" "192.168.90.40:9100")
+          (mkTarget "bob" "192.168.90.25:9100")
+        ];
+      }
+      {
+        job_name = "process_grouped";
+        static_configs = [
+          (mkTarget "jeeves" "192.168.90.40:9256")
+          (mkTarget "bob" "192.168.90.25:9256")
+        ];
+      }
+      {
+        job_name = "smartctl";
+        static_configs = [
+          (mkTarget "jeeves" "192.168.90.40:9633")
+          (mkTarget "bob" "192.168.90.25:9633")
+        ];
+      }
+      {
+        job_name = "zfs";
+        static_configs = [
+          (mkTarget "jeeves" "192.168.90.40:9134")
+          (mkTarget "bob" "192.168.90.25:9134")
+        ];
+      }
+    ];
+  };
+
+  pidPrometheusConfig = mkPrometheusConfig "prometheus-pid-short" {
+    global = {
+      scrape_interval = "15s";
+      scrape_timeout = "10s";
+      evaluation_interval = "15s";
+    };
+    scrape_configs = [
+      {
+        job_name = "process_pid";
+        static_configs = [
+          (mkTarget "jeeves" "192.168.90.40:9257")
+          (mkTarget "bob" "192.168.90.25:9257")
+        ];
+      }
+    ];
+  };
+
+  mkPrometheusService =
+    {
+      dataDir,
+      configFile,
+      port,
+      retention,
+    }:
+    {
+      after = [
+        "zfs-media-database-prometheus.mount"
+        "network.target"
+      ];
+      requires = [ "zfs-media-database-prometheus.mount" ];
+      wantedBy = [ "multi-user.target" ];
+      unitConfig.RequiresMountsFor = [ dataDir ];
+      serviceConfig = {
+        ExecStart = "${lib.getExe pkgs.prometheus} ${
+          lib.escapeShellArgs [
+            "--config.file=${configFile}"
+            "--storage.tsdb.path=${dataDir}"
+            "--storage.tsdb.retention.time=${retention}"
+            "--web.listen-address=127.0.0.1:${toString port}"
+          ]
+        }";
+        User = "prometheus";
+        Group = "prometheus";
+        Restart = "always";
+        RestartSec = "5s";
+        WorkingDirectory = dataDir;
+        ReadWritePaths = [ dataDir ];
+        CapabilityBoundingSet = [ "" ];
+        DeviceAllow = [ "/dev/null rw" ];
+        DevicePolicy = "strict";
+        LockPersonality = true;
+        MemoryDenyWriteExecute = true;
+        NoNewPrivileges = true;
+        PrivateDevices = true;
+        PrivateTmp = true;
+        ProtectClock = true;
+        ProtectControlGroups = true;
+        ProtectHome = true;
+        ProtectHostname = true;
+        ProtectKernelLogs = true;
+        ProtectKernelModules = true;
+        ProtectKernelTunables = true;
+        ProtectProc = "invisible";
+        ProtectSystem = "strict";
+        RemoveIPC = true;
+        RestrictAddressFamilies = [
+          "AF_INET"
+          "AF_INET6"
+          "AF_UNIX"
+        ];
+        RestrictNamespaces = true;
+        RestrictRealtime = true;
+        RestrictSUIDSGID = true;
+        SystemCallArchitectures = "native";
+        SystemCallFilter = [
+          "@system-service"
+          "~@privileged"
+        ];
+      };
+    };
+in
+{
+  users = {
+    groups.prometheus = { };
+    users.prometheus = {
+      isSystemUser = true;
+      group = "prometheus";
+      description = "Prometheus daemon user";
+    };
+  };
+
+  systemd = {
+    services = {
+      prometheus-main = mkPrometheusService {
+        configFile = mainPrometheusConfig;
+        dataDir = mainPrometheusDataDir;
+        port = 9090;
+        retention = "90d";
+      };
+
+      prometheus-pid-short = mkPrometheusService {
+        configFile = pidPrometheusConfig;
+        dataDir = pidPrometheusDataDir;
+        port = 9092;
+        retention = "10m";
+      };
+    };
+
+    tmpfiles.rules = [
+      "d ${prometheusDataRoot} 0755 root root - -"
+      "d ${mainPrometheusDataDir} 0750 prometheus prometheus - -"
+      "d ${pidPrometheusDataDir} 0750 prometheus prometheus - -"
+    ];
+  };
+}

systems/jeeves/networking.nix

@@ -1,4 +1,13 @@
 {
+  # Docker loads br_netfilter on jeeves. Disable bridge netfilter so
+  # br-nix-builder behaves like a pure L2 bridge and bridged traffic
+  # does not hit the host firewall/rpfilter path.
+  boot.kernel.sysctl = {
+    "net.bridge.bridge-nf-call-arptables" = 0;
+    "net.bridge.bridge-nf-call-ip6tables" = 0;
+    "net.bridge.bridge-nf-call-iptables" = 0;
+  };
+
   networking = {
     hostName = "jeeves";
     hostId = "0e15ce35";
@@ -34,11 +43,18 @@
       };
     };
     networks = {
-      "10-1GB_Primary" = {
-        matchConfig.Name = "enp97s0f1";
+      "10-Primary" = {
+        matchConfig.Name = "enp97s0";
         address = [ "192.168.99.14/24" ];
+        dns = [
+          "192.168.99.1"
+          "2600:4040:abfb:d700::1"
+        ];
         routes = [ { Gateway = "192.168.99.1"; } ];
         vlan = [ "internet-vlan" ];
+        dhcpV4Config.UseDNS = false;
+        dhcpV6Config.UseDNS = false;
+        ipv6AcceptRAConfig.UseDNS = false;
         linkConfig.RequiredForOnline = "routable";
       };
       "50-internet-vlan" = {
@@ -49,23 +65,10 @@
       "60-br-nix-builder" = {
         matchConfig.Name = "br-nix-builder";
         bridgeConfig = { };
-        address = [ "192.168.3.10/24" ];
-        routingPolicyRules = [
-          {
-            From = "192.168.3.0/24";
-            Table = 100;
-            Priority = 100;
-          }
-        ];
-        routes = [
-          {
-            Gateway = "192.168.3.1";
-            Table = 100;
-            GatewayOnLink = false;
-            Metric = 2048;
-            PreferredSource = "192.168.3.10";
-          }
-        ];
+        networkConfig = {
+          IPv6AcceptRA = false;
+          LinkLocalAddressing = "no";
+        };
         linkConfig.RequiredForOnline = "no";
       };
     };

systems/jeeves/programs.nix

@@ -3,5 +3,6 @@
   environment.systemPackages = with pkgs; [
     filebot
     docker-compose
+    ffmpeg
   ];
 }

systems/jeeves/runners/default.nix

@@ -1,20 +1,7 @@
-{ pkgs, ... }:
+{ ... }:
 {
   imports = [ ./nix_builder.nix ];
 
-  users = {
-    users.github-runners = {
-      shell = pkgs.bash;
-      isSystemUser = true;
-      group = "github-runners";
-      uid = 601;
-      openssh.authorizedKeys.keys = [
-        "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIA/S8i+BNX/12JNKg+5EKGX7Aqimt5KM+ve3wt/SyWuO github-runners" # cspell:disable-line
-      ];
-    };
-    groups.github-runners.gid = 601;
-  };
-
   services.nix_builder.containers = {
     nix-builder-00.enable = true;
     nix-builder-01.enable = true;

systems/jeeves/runners/nix_builder.nix

@@ -2,6 +2,7 @@
   config,
   lib,
   outputs,
+  utils,
   ...
 }:
 
@@ -9,6 +10,8 @@ with lib;
 let
   vars = import ../vars.nix;
   cfg = config.services.nix_builder;
+  runnerUsername = "gitea-runner";
+  runnerUserid = 601;
 in
 {
   options.services.nix_builder = {
@@ -23,37 +26,40 @@ in
         types.submodule (
           { name, ... }:
           {
-            options.enable = mkEnableOption "GitHub runner container";
+            options.enable = mkEnableOption "Gitea runner container";
           }
         )
       );
       default = { };
-      description = "GitHub runner container configurations";
+      description = "Gitea runner container configurations";
     };
   };
 
   config = {
+    users = {
+      users.${runnerUsername} = {
+        isSystemUser = true;
+        group = runnerUsername;
+        uid = runnerUserid;
+      };
+      groups.${runnerUsername}.gid = runnerUserid;
+    };
+
     containers = mapAttrs (
       name: containerCfg:
       mkIf containerCfg.enable {
         autoStart = true;
         privateNetwork = true;
         hostBridge = cfg.bridgeName;
-        ephemeral = true;
         bindMounts = {
-          storage = {
-            hostPath = "/zfs/media/github-runners/${name}";
-            mountPoint = "/zfs/media/github-runners/${name}";
-            isReadOnly = false;
-          };
           host-nix = {
             mountPoint = "/host-nix/var/nix/daemon-socket";
             hostPath = "/nix/var/nix/daemon-socket";
             isReadOnly = false;
           };
-          pat = {
-            hostPath = "${vars.secrets}/services/github-runners/runner_pat";
-            mountPoint = "${vars.secrets}/services/github-runners/runner_pat";
+          token = {
+            hostPath = "${vars.secrets}/services/gitea-runners";
+            mountPoint = "/run/secrets/gitea-runners";
             isReadOnly = true;
           };
         };
@@ -92,46 +98,69 @@ in
                 "nix-command"
               ];
               sandbox = true;
-              allowed-users = [ "github-runners" ];
+              allowed-users = [ "gitea-runner" ];
               trusted-users = [
                 "root"
-                "github-runners"
+                "gitea-runner"
               ];
             };
             nixpkgs = {
               overlays = builtins.attrValues outputs.overlays;
               config.allowUnfree = true;
             };
-            services.github-runners.${name} = {
+            users = {
+              users.${runnerUsername} = {
+                isSystemUser = true;
+                group = runnerUsername;
+                uid = runnerUserid;
+              };
+              groups.${runnerUsername}.gid = runnerUserid;
+            };
+            services.gitea-actions-runner.instances.${name} = {
               enable = true;
-              replace = true;
-              workDir = "/zfs/media/github-runners/${name}";
-              url = "https://github.com/RichieCahill/dotfiles";
-              extraLabels = [ "nixos" ];
-              tokenFile = "${vars.secrets}/services/github-runners/runner_pat";
-              user = "github-runners";
-              group = "github-runners";
-              extraPackages = with pkgs; [
+              name = "jeeves-${name}";
+              url = "http://192.168.99.14:6443/";
+              labels = [
+                "self-hosted:host"
+                "nixos:host"
+              ];
+              tokenFile = "/run/secrets/gitea-runners/registration-token";
+              hostPackages = with pkgs; [
+                bash
+                coreutils
+                curl
+                gawk
                 gitMinimal
-                gh
+                gnused
+                my_python
+                nix
                 nixfmt
                 nixos-rebuild
+                nodejs
                 treefmt
-                my_python
+                wget
               ];
             };
-            users = {
-              users.github-runners = {
-                shell = pkgs.bash;
-                isSystemUser = true;
-                group = "github-runners";
-                uid = 601;
+            systemd.services."gitea-runner-${utils.escapeSystemdPath name}" = {
+              serviceConfig = {
+                DynamicUser = mkForce false;
+                User = mkForce runnerUsername;
+                Group = mkForce runnerUsername;
               };
-              groups.github-runners.gid = 601;
             };
             system.stateVersion = "24.05";
           };
       }
     ) cfg.containers;
+
+    systemd.services = builtins.listToAttrs (
+      map (name: {
+        name = "container@${name}";
+        value = {
+          requires = [ "gitea.service" ];
+          after = [ "gitea.service" ];
+        };
+      }) (builtins.attrNames (filterAttrs (_: c: c.enable) cfg.containers))
+    );
   };
 }

systems/jeeves/scripts/zfs.sh

@@ -21,7 +21,9 @@ sudo zfs create media/secure/docker -o compression=zstd-9
 sudo zfs create media/secure/github-runners -o compression=zstd-9 -o sync=disabled
 sudo zfs create media/secure/home_assistant -o compression=zstd-19
 sudo zfs create media/secure/notes -o copies=2
-sudo zfs create media/secure/postgres -o recordsize=16k -o primarycache=metadata
+sudo zfs create media/secure/postgres -o mountpoint=/zfs/media/database/postgres -o recordsize=16k -o primarycache=metadata
+sudo zfs create media/secure/postgres-wal -o mountpoint=/zfs/media/database/postgres-wal -o recordsize=32k -o primarycache=metadata -o special_small_blocks=32K -o compression=lz4 -o secondarycache=none -o logbias=latency
+sudo zfs create media/secure/prometheus -o mountpoint=/zfs/media/database/prometheus -o compression=lz4
 sudo zfs create media/secure/services -o compression=zstd-9
 sudo zfs create media/secure/share -o mountpoint=/zfs/media/share -o exec=off
 
@@ -40,3 +42,4 @@ sudo zfs create storage/secure/plex -o recordsize=1M -o compression=zstd-19
 sudo zfs create storage/secure/secrets -o compression=zstd-19 -o copies=3
 sudo zfs create storage/secure/syncthing -o compression=zstd-19
 sudo zfs create storage/secure/transmission -o recordsize=1M -o compression=zstd-9 -o exec=off -o sync=disabled
+sudo zfs create storage/secure/important -o compression=zstd-19 -o copies=2 -o mountpoint=/zfs/storage/important

systems/jeeves/services/audiobookshelf.nix

@@ -3,7 +3,10 @@ let
   vars = import ../vars.nix;
 in
 {
-  services.audiobookshelf.enable = true;
+  services.audiobookshelf = {
+    enable = true;
+    port = 8000;
+  };
   systemd.services.audiobookshelf.serviceConfig.WorkingDirectory =
     lib.mkForce "${vars.docker_configs}/audiobookshelf";
   users.users.audiobookshelf.home = lib.mkForce "${vars.docker_configs}/audiobookshelf";

systems/jeeves/services/bluesky_firehose.nix

@@ -1,96 +0,0 @@
-{
-  pkgs,
-  inputs,
-  ...
-}:
-let
-  commonEnv = {
-    PYTHONPATH = "${inputs.self}";
-    KAFKA_BOOTSTRAP_SERVERS = "localhost:9092";
-    BLUESKY_FIREHOSE_TOPIC = "bluesky.firehose.posts";
-  };
-  commonServiceConfig = {
-    Type = "simple";
-    WorkingDirectory = "${inputs.self}";
-    User = "richie";
-    Group = "users";
-    Restart = "on-failure";
-    RestartSec = "10s";
-    StandardOutput = "journal";
-    StandardError = "journal";
-    NoNewPrivileges = true;
-    ProtectSystem = "strict";
-    ProtectHome = "read-only";
-    PrivateTmp = true;
-    ReadOnlyPaths = [ "${inputs.self}" ];
-  };
-in
-{
-  systemd.services.bluesky-firehose-topic-init = {
-    description = "Create Kafka topic for Bluesky firehose";
-    after = [ "apache-kafka.service" ];
-    requires = [ "apache-kafka.service" ];
-    wantedBy = [ "multi-user.target" ];
-
-    serviceConfig = {
-      Type = "oneshot";
-      RemainAfterExit = true;
-      ExecStart = pkgs.writeShellScript "create-bluesky-topic" ''
-        ${pkgs.apacheKafka}/bin/kafka-topics.sh \
-          --bootstrap-server localhost:9092 \
-          --create \
-          --if-not-exists \
-          --topic bluesky.firehose.posts \
-          --partitions 6 \
-          --replication-factor 1
-      '';
-    };
-  };
-
-  systemd.services.bluesky-firehose-producer = {
-    description = "Bluesky Jetstream to Kafka producer";
-    after = [
-      "network.target"
-      "apache-kafka.service"
-      "bluesky-firehose-topic-init.service"
-    ];
-    requires = [
-      "apache-kafka.service"
-      "bluesky-firehose-topic-init.service"
-    ];
-    wantedBy = [ "multi-user.target" ];
-
-    environment = commonEnv;
-
-    serviceConfig = commonServiceConfig // {
-      ExecStart = "${pkgs.my_python}/bin/python -m python.data_science.firehose_producer";
-    };
-  };
-
-  systemd.services.bluesky-firehose-consumer = {
-    description = "Bluesky Kafka to PostgreSQL consumer";
-    after = [
-      "network.target"
-      "apache-kafka.service"
-      "bluesky-firehose-topic-init.service"
-      "postgresql.service"
-    ];
-    requires = [
-      "apache-kafka.service"
-      "bluesky-firehose-topic-init.service"
-      "postgresql.service"
-    ];
-    wantedBy = [ "multi-user.target" ];
-
-    environment = commonEnv // {
-      DATA_SCIENCE_DEV_DB = "data_science_dev";
-      DATA_SCIENCE_DEV_USER = "richie";
-      DATA_SCIENCE_DEV_HOST = "/run/postgresql";
-      DATA_SCIENCE_DEV_PORT = "5432";
-    };
-
-    serviceConfig = commonServiceConfig // {
-      ExecStart = "${pkgs.my_python}/bin/python -m python.data_science.firehose_consumer";
-    };
-  };
-}

systems/jeeves/services/camofox-browser.nix

@@ -0,0 +1,80 @@
+{
+  ...
+}:
+let
+  vars = import ../vars.nix;
+in
+{
+  systemd.tmpfiles.rules = [
+    "d ${vars.docker_configs}/camofox-browser 0750 root root - -"
+  ];
+
+  containers.camofox-browser = {
+    autoStart = true;
+    privateNetwork = false;
+    bindMounts = {
+      camofox-browser = {
+        hostPath = "${vars.docker_configs}/camofox-browser";
+        mountPoint = "/var/lib/camofox-browser";
+        isReadOnly = false;
+      };
+    };
+    config =
+      {
+        pkgs,
+        lib,
+        ...
+      }:
+      {
+        networking.hostName = "camofox-browser";
+
+        environment.systemPackages = with pkgs; [
+          ffmpeg
+          git
+          nodejs
+          python3Packages.yt-dlp
+        ];
+
+        systemd.services.camofox-browser = {
+          description = "Camofox browser server";
+          wantedBy = [ "multi-user.target" ];
+          after = [ "network.target" ];
+          environment = {
+            CAMOFOX_HOST = "127.0.0.1";
+            CAMOFOX_PORT = "9377";
+            HOME = "/var/lib/camofox-browser";
+          };
+          path = with pkgs; [
+            bash
+            coreutils
+            git
+            nodejs
+          ];
+          serviceConfig = {
+            Restart = "always";
+            RestartSec = "5s";
+            WorkingDirectory = "/var/lib/camofox-browser";
+          };
+          script = ''
+            set -eu
+
+            app_dir=/var/lib/camofox-browser/app
+
+            if [ ! -d "$app_dir/.git" ]; then
+              git clone --depth 1 https://github.com/jo-inc/camofox-browser "$app_dir"
+            fi
+
+            cd "$app_dir"
+
+            if [ ! -d node_modules ]; then
+              npm install
+            fi
+
+            exec npm start
+          '';
+        };
+
+        system.stateVersion = lib.mkDefault "24.05";
+      };
+  };
+}

systems/jeeves/services/cloud_flare_tunnel.nix

@@ -1,17 +0,0 @@
-{ pkgs, ... }:
-let
-  vars = import ../vars.nix;
-in
-{
-  systemd.services.cloud_flare_tunnel = {
-    description = "cloud_flare_tunnel proxy's traffic through cloudflare";
-    after = [ "network.target" ];
-    wantedBy = [ "multi-user.target" ];
-    serviceConfig = {
-      Type = "simple";
-      EnvironmentFile = "${vars.secrets}/docker/cloud_flare_tunnel";
-      ExecStart = "${pkgs.cloudflared}/bin/cloudflared --no-autoupdate tunnel run";
-      Restart = "on-failure";
-    };
-  };
-}

systems/jeeves/services/gitea.nix

@@ -2,7 +2,10 @@ let
   vars = import ../vars.nix;
 in
 {
-  networking.firewall.allowedTCPPorts = [ 6443 ];
+  networking.firewall.allowedTCPPorts = [
+    6443
+    2223
+  ];
 
   services.gitea = {
     enable = true;
@@ -18,13 +21,17 @@ in
       createDatabase = false;
     };
     settings = {
+      actions = {
+        ENABLED = true;
+        DEFAULT_ACTIONS_URL = "github";
+      };
       service.DISABLE_REGISTRATION = true;
       server = {
         DOMAIN = "tmmworkshop.com";
         ROOT_URL = "https://gitea.tmmworkshop.com/";
         HTTP_PORT = 6443;
         SSH_PORT = 2223;
-        SSH_LISTEN_PORT = 2224;
+        SSH_LISTEN_PORT = 2223;
         START_SSH_SERVER = true;
         PUBLIC_URL_DETECTION = "auto";
       };

systems/jeeves/services/grafana.nix

@@ -0,0 +1,80 @@
+{
+  ...
+}:
+let
+  vars = import ../vars.nix;
+  grafanaDataDir = "${vars.services}/grafana";
+in
+{
+  networking.firewall.allowedTCPPorts = [ 3000 ];
+
+  services.grafana = {
+    enable = true;
+    dataDir = grafanaDataDir;
+    settings = {
+      database.type = "sqlite3";
+      security = {
+        admin_password = "$__file{${vars.secrets}/services/grafana/admin_password}";
+        admin_user = "admin";
+        secret_key = "$__file{${vars.secrets}/services/grafana/secret_key}";
+      };
+      server = {
+        http_addr = "192.168.90.40";
+        http_port = 3000;
+        root_url = "http://192.168.90.40:3000/";
+      };
+    };
+    provision = {
+      enable = true;
+      dashboards.settings = {
+        apiVersion = 1;
+        providers = [
+          {
+            name = "monitoring";
+            folder = "Monitoring";
+            type = "file";
+            disableDeletion = false;
+            editable = false;
+            allowUiUpdates = false;
+            updateIntervalSeconds = 30;
+            options.path = ../monitoring/dashboards;
+          }
+        ];
+      };
+      datasources.settings = {
+        apiVersion = 1;
+        prune = true;
+        datasources = [
+          {
+            access = "proxy";
+            editable = false;
+            isDefault = true;
+            name = "prom-main";
+            type = "prometheus";
+            uid = "prom-main";
+            url = "http://127.0.0.1:9090";
+          }
+          {
+            access = "proxy";
+            editable = false;
+            name = "prom-pid-short";
+            type = "prometheus";
+            uid = "prom-pid-short";
+            url = "http://127.0.0.1:9092";
+          }
+        ];
+      };
+    };
+  };
+
+  systemd = {
+    services.grafana.after = [
+      "prometheus-main.service"
+      "prometheus-pid-short.service"
+    ];
+
+    tmpfiles.rules = [
+      "d ${grafanaDataDir} 0750 grafana grafana - -"
+    ];
+  };
+}

systems/jeeves/services/kafka.nix

@@ -7,6 +7,13 @@ in
     settings = {
       listeners = [ "PLAINTEXT://localhost:9092" ];
       "log.dirs" = [ vars.kafka ];
+      "num.partitions" = 6;
+      "default.replication.factor" = 1;
+      "log.retention.hours" = 168;
+      "log.retention.bytes" = 10737418240;
+      "log.segment.bytes" = 1073741824;
+      "log.cleanup.policy" = "delete";
+      "auto.create.topics.enable" = false;
     };
   };
 }

systems/jeeves/services/nornsight.nix

@@ -0,0 +1,107 @@
+{ pkgs, ... }:
+let
+  vars = import ../vars.nix;
+  stateDir = "${vars.services}/nornsight";
+  appDir = "${stateDir}/app";
+  binPath = pkgs.lib.makeBinPath [
+    pkgs.binutils
+    pkgs.libpq
+    pkgs.postgresql
+    pkgs.stdenv.cc
+  ];
+  libraryPath = pkgs.lib.makeLibraryPath [
+    pkgs.libpq
+    pkgs.postgresql.lib
+  ];
+in
+{
+  systemd.tmpfiles.rules = [
+    "d ${stateDir} 0750 nornsight nornsight - -"
+  ];
+
+  users.users.nornsight = {
+    isSystemUser = true;
+    group = "nornsight";
+    home = stateDir;
+  };
+
+  systemd.services.nornsight = {
+    description = "Norn Sight";
+    after = [ "network-online.target" ];
+    wants = [ "network-online.target" ];
+    wantedBy = [ "multi-user.target" ];
+
+    environment = {
+      HOME = stateDir;
+      UV_CACHE_DIR = "${stateDir}/.cache/uv";
+      UV_PROJECT_ENVIRONMENT = "${appDir}/.venv";
+      UV_PYTHON = "${pkgs.python313}/bin/python3.13";
+      UV_PYTHON_DOWNLOADS = "never";
+      LD_LIBRARY_PATH = libraryPath;
+      LIBRARY_PATH = libraryPath;
+      PSYCOPG_IMPL = "python";
+    };
+
+    path = with pkgs; [
+      bash
+      coreutils
+      git
+      uv
+    ];
+
+    serviceConfig = {
+      Type = "simple";
+      User = "nornsight";
+      Group = "nornsight";
+      EnvironmentFile = "-${vars.secrets}/services/nornsight";
+      WorkingDirectory = stateDir;
+      Restart = "on-failure";
+      RestartSec = "5s";
+      StandardOutput = "journal";
+      StandardError = "journal";
+      NoNewPrivileges = true;
+      PrivateTmp = true;
+      ProtectHome = true;
+      ProtectSystem = "strict";
+      ReadWritePaths = [ stateDir ];
+    };
+
+    script = ''
+      set -eu
+      export PATH="${binPath}:$PATH"
+      export LD_LIBRARY_PATH="${libraryPath}:''${LD_LIBRARY_PATH:-}"
+      export LIBRARY_PATH="${libraryPath}:''${LIBRARY_PATH:-}"
+
+      : "''${NORN_SIGHT_REPO_URL:?NORN_SIGHT_REPO_URL is required}"
+      branch="''${NORN_SIGHT_BRANCH:-main}"
+
+      if [ -d "${appDir}/.git" ]; then
+        current_origin="$(git -C "${appDir}" remote get-url origin)"
+        if [ "$current_origin" != "$NORN_SIGHT_REPO_URL" ]; then
+          rm -rf "${appDir}"
+        fi
+      fi
+
+      if [ ! -d "${appDir}/.git" ]; then
+        git clone --branch "$branch" "$NORN_SIGHT_REPO_URL" "${appDir}"
+      else
+        cd "${appDir}"
+        git fetch origin "$branch"
+        git checkout "$branch"
+        git pull --ff-only origin "$branch"
+      fi
+
+      cd "${appDir}"
+      uv sync --upgrade
+      uv run python - <<'PY'
+      import ctypes.util
+      import os
+
+      print(f"LD_LIBRARY_PATH={os.environ.get('LD_LIBRARY_PATH')}")
+      print(f"LIBRARY_PATH={os.environ.get('LIBRARY_PATH')}")
+      print(f"libpq={ctypes.util.find_library('pq')}")
+      PY
+      exec uv run uvicorn pipelines.web.main:app --host 0.0.0.0 --port 8001
+    '';
+  };
+}