added answer.py and config

python/ebook_search/answer.py

@@ -0,0 +1,57 @@
+"""Grounded answer generation."""
+
+from __future__ import annotations
+
+import logging
+from typing import TYPE_CHECKING
+
+from python.ebook_search.llm_interface import request_chat_completion
+
+if TYPE_CHECKING:
+    from python.ebook_search.config import EbookSearchConfig
+    from python.ebook_search.search import SearchResult
+
+logger = logging.getLogger(__name__)
+
+
+def answer_query(query: str, results: list[SearchResult], config: EbookSearchConfig) -> str:
+    """Answer a question using only retrieved chunks."""
+    if not config.answer_enabled:
+        logger.info("ebook_answer_skipped_disabled")
+        return "Answer generation is disabled. Source chunks are shown below."
+
+    if not results:
+        logger.info("ebook_answer_skipped_no_results")
+        return "No relevant sources were found."
+
+    logger.info(
+        "ebook_answer_request_start base_url=%s model=%s sources=%s query_length=%s",
+        config.vllm_base_url,
+        config.chat_model,
+        len(results),
+        len(query),
+    )
+    context = "\n\n".join(
+        f"[{index}] {result.source_title}{' - ' + result.chapter_title if result.chapter_title else ''}\n{result.text}"
+        for index, result in enumerate(results, start=1)
+    )
+    content = request_chat_completion(
+        config,
+        [
+            {
+                "role": "system",
+                "content": (
+                    "Answer only from the provided context. Cite sources with bracketed numbers like [1]. "
+                    "If the context is insufficient, say so."
+                ),
+            },
+            {"role": "user", "content": f"Question:\n{query}\n\nContext:\n{context}"},
+        ],
+    )
+
+    logger.info(
+        "ebook_answer_request_complete model=%s answer_length=%s",
+        config.chat_model,
+        len(content),
+    )
+    return content or "The model returned an empty answer."

python/ebook_search/config.py

@@ -0,0 +1,117 @@
+"""Configuration for the EPUB search app."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from os import getenv
+
+
+def getenv_bool(name: str, *, default: bool) -> bool:
+    """Read a boolean environment variable with a default fallback."""
+    value = getenv(name)
+    if value is None:
+        return default
+    return value.strip().lower() in {"1", "true", "yes", "on"}
+
+
+def getenv_int(name: str, *, default: int) -> int:
+    """Read an integer environment variable with a default fallback."""
+    value = getenv(name)
+    if value is None or not value.strip():
+        return default
+    return int(value)
+
+
+@dataclass(frozen=True)
+class RerankConfig:
+    """vLLM reranker settings."""
+
+    enabled: bool = False
+    base_url: str = "http://192.168.90.25:8001"
+    model: str = "qwen3-reranker-06b"
+    candidates: int = 24
+    timeout_seconds: float = 30.0
+
+
+@dataclass(frozen=True)
+class EbookSearchConfig:
+    """Runtime settings for EPUB search."""
+
+    rerank: RerankConfig
+    top_k: int = 12
+    library_paths: tuple[str, ...] = ()
+    vllm_base_url: str = "https://ollama.com/v1"
+    vllm_api_key: str = "not-needed"
+    chat_model: str = "deepseek-v4-flash"
+    answer_enabled: bool = True
+    embedding_base_url: str = "http://192.168.90.25:8000/v1"
+    embedding_api_key: str = "not-needed"
+    embedding_model: str = "qwen3-embedding-0.6b"
+    embedding_batch_size: int = 32
+    bm25_index_dir: str = ".ebook_search_bm25"
+    bm25_refresh_delay_seconds: int = 60
+
+
+def load_rerank_config() -> RerankConfig:
+    """Load reranker config from environment variables."""
+    return RerankConfig(
+        enabled=getenv_bool("EBOOK_SEARCH_RERANK_ENABLED", default=False),
+        base_url=getenv("EBOOK_SEARCH_RERANK_BASE_URL", "http://192.168.90.25:8001"),
+        model=getenv("EBOOK_SEARCH_RERANK_MODEL", "qwen3-reranker-06b"),
+        candidates=getenv_int("EBOOK_SEARCH_RERANK_CANDIDATES", default=24),
+        timeout_seconds=float(getenv_int("EBOOK_SEARCH_RERANK_TIMEOUT_SECONDS", default=30)),
+    )
+
+
+def load_config() -> EbookSearchConfig:
+    """Load EPUB search config from environment variables."""
+    return EbookSearchConfig(
+        rerank=load_rerank_config(),
+        top_k=getenv_int("EBOOK_SEARCH_TOP_K", default=12),
+        library_paths=library_paths_from_env(),
+        vllm_base_url=getenv("EBOOK_SEARCH_VLLM_BASE_URL", "https://ollama.com/v1"),
+        vllm_api_key=getenv("EBOOK_SEARCH_VLLM_API_KEY") or getenv("OLLAMA_API_KEY") or "not-needed",
+        chat_model=getenv("EBOOK_SEARCH_CHAT_MODEL", "deepseek-v4-flash"),
+        answer_enabled=getenv_bool("EBOOK_SEARCH_ANSWER_ENABLED", default=True),
+        embedding_base_url=getenv("EBOOK_SEARCH_EMBEDDING_BASE_URL", "http://192.168.90.25:8000/v1"),
+        embedding_api_key=getenv("EBOOK_SEARCH_EMBEDDING_API_KEY", "not-needed"),
+        embedding_model=normalize_embedding_model(),
+        embedding_batch_size=getenv_int("EBOOK_SEARCH_EMBEDDING_BATCH_SIZE", default=32),
+        bm25_index_dir=getenv("EBOOK_SEARCH_BM25_INDEX_DIR", ".ebook_search_bm25"),
+        bm25_refresh_delay_seconds=getenv_int("EBOOK_SEARCH_BM25_REFRESH_DELAY_SECONDS", default=60),
+    )
+
+
+def normalize_embedding_model(default: str = "qwen3-embedding-0.6b") -> str:
+    """Normalize supported embedding aliases to provider model names."""
+    aliases = {
+        "Qwen3-Embedding-0.6B": "qwen3-embedding-0.6b",
+        "Qwen3-Embedding-4B": "qwen3-embedding-4b",
+        "Qwen3-Embedding-8B": "qwen3-embedding-8b",
+        "Qwen/Qwen3-Embedding-0.6B": "qwen3-embedding-0.6b",
+        "Qwen/Qwen3-Embedding-4B": "qwen3-embedding-4b",
+        "Qwen/Qwen3-Embedding-8B": "qwen3-embedding-8b",
+        "qwen3-embedding:0.6b": "qwen3-embedding-0.6b",
+        "qwen3-embedding:4b": "qwen3-embedding-4b",
+        "qwen3-embedding:8b": "qwen3-embedding-8b",
+        "qwen3-embedding-0.6b": "qwen3-embedding-0.6b",
+        "qwen3-embedding-4b": "qwen3-embedding-4b",
+        "qwen3-embedding-8b": "qwen3-embedding-8b",
+    }
+
+    model = getenv("EBOOK_SEARCH_EMBEDDING_MODEL", default)
+    standard_model = aliases.get(model)
+
+    if standard_model is None:
+        error = f"Embedding model {model} is not supported. Supported models are {aliases.keys()}"
+        raise ValueError(error)
+
+    return standard_model
+
+
+def library_paths_from_env() -> tuple[str, ...]:
+    """Read configured EPUB library paths from the environment."""
+    value = getenv("EBOOK_SEARCH_LIBRARY_PATHS")
+    if value is None:
+        return ()
+    return tuple(path for path in value.split(":") if path)