Richie/dotfiles
1
0
Fork 0
Code Issues 1 Pull Requests 3 Actions Packages Projects Releases Wiki Activity

perf(ebook-search): run phrase detection in parallel with retrieval

Browse Source 
Move protected phrase detection into the retrieval gather so it runs
concurrently with vector and BM25 candidates instead of sequentially
before them. Make the search API accept real bool form fields for
rerank/phrase_matching, gate phrase matching on both the request and
config kill switch, and reflow log f-strings for readability.
This commit is contained in:
Richie
2026-07-24 11:38:51 -04:00
parent 8f1a69529c
commit 0028237579
5 changed files with 113 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ebook_search/protected_phrases/extraction.py
+7 -6
View File
@@ -688,14 +688,14 @@ def extract_phrase_candidates_for_book(
raw_started_at = perf_counter()
raw = extract_raw_ngrams_by_chapter(chapters, config)
logger.info(
f"ebook_phrase_candidate_extract_raw_complete candidates={len(raw)} duration_ms={(perf_counter() - "
f"raw_started_at) * 1000:.1f}"
f"ebook_phrase_candidate_extract_raw_complete candidates={len(raw)} "
f"duration_ms={(perf_counter() - raw_started_at) * 1000:.1f}"
)
yake_started_at = perf_counter()
yake_candidates = extract_yake_candidates(book_text, config)
logger.info(
f"ebook_phrase_candidate_extract_yake_complete candidates={len(yake_candidates)} duration_ms={(perf_counter() - "
f"yake_started_at) * 1000:.1f}"
f"ebook_phrase_candidate_extract_yake_complete candidates={len(yake_candidates)} "
f"duration_ms={(perf_counter() - yake_started_at) * 1000:.1f}"
)
capitalized_started_at = perf_counter()
capitalized = extract_capitalized_phrases(book_text, config)
@@ -728,7 +728,8 @@ def extract_phrase_candidates_for_book(
f"ebook_phrase_candidate_extract_complete raw={len(raw)} yake={len(yake_candidates)} "
f"capitalized={len(capitalized)} metadata={len(metadata_candidates)} {pre_filter_count=} {filtered_too_short=} "
f"{filtered_too_rare=} {filtered_too_common=} {filtered_junk=} min_uses={minimum_candidate_raw_count(config)} "
f"storable={len(candidates)} limited={len(limited)} enrich_score_ms={(perf_counter() - enriched_started_at) * "
f"1000:.1f} duration_ms={(perf_counter() - started_at) * 1000:.1f}"
f"storable={len(candidates)} limited={len(limited)} "
f"enrich_score_ms={(perf_counter() - enriched_started_at) * 1000:.1f} "
f"duration_ms={(perf_counter() - started_at) * 1000:.1f}"
)
return limited
python/ebook_search/protected_phrases/generate_ngrams.py
+6 -6
View File
@@ -68,8 +68,8 @@ async def generate_candidate_phrases_for_books(
source_ids = (await session.scalars(source_query)).all()
books_seen = len(source_ids)
logger.info(
f"ebook_candidate_phrase_generation_start {books_seen=} {config.phrase_min_tokens=} {config.phrase_max_tokens=} "
f"{config.protected_phrase_max_candidates_per_book=}"
f"ebook_candidate_phrase_generation_start {books_seen=} {config.phrase_min_tokens=} "
f"{config.phrase_max_tokens=} {config.protected_phrase_max_candidates_per_book=}"
)
pool = get_extraction_pool(config.protected_phrase_extraction_workers)
@@ -244,8 +244,8 @@ async def generate_candidate_phrases_for_book(
await session.rollback()
raise
logger.info(
f"ebook_candidate_phrase_generation_book_duration {book_id=} {saved_count=} duration_ms={(perf_counter() - "
f"started_at) * 1000:.1f}"
f"ebook_candidate_phrase_generation_book_duration {book_id=} {saved_count=} "
f"duration_ms={(perf_counter() - started_at) * 1000:.1f}"
)
return saved_count
@@ -289,7 +289,7 @@ async def store_candidate_phrases_for_book(
)
saved_count = await bulk_upsert_unjudged_candidates(session, book_id, series_id, limited_candidates)
logger.info(
f"ebook_candidate_phrase_save_complete {book_id=} {saved_count=} save_ms={(perf_counter() - save_started_at) * "
f"1000:.1f}"
f"ebook_candidate_phrase_save_complete {book_id=} {saved_count=} "
f"save_ms={(perf_counter() - save_started_at) * 1000:.1f}"
)
return saved_count