highperfocused/danswer
1
0
Fork 0
mirror of https://github.com/danswer-ai/danswer.git synced 2025-07-01 18:20:49 +02:00
Code Issues Packages Projects Releases Wiki Activity

removal of keyword 1st phase

Browse Source
This commit is contained in:
joachim-danswer
2025-04-22 20:18:57 -07:00
committed by Chris Weaver
parent c76dc2ea2c
commit d8b050026d
1 changed files with 4 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
backend/onyx/context/search/retrieval/search_runner.py
View File

@ -164,9 +164,7 @@ def doc_index_retrieval(
top_base_chunks_standard_ranking_thread: ( top_base_chunks_standard_ranking_thread: (
TimeoutThread[list[InferenceChunkUncleaned]] | None TimeoutThread[list[InferenceChunkUncleaned]] | None
) = None ) = None
top_base_chunks_keyword_ranking_thread: (
TimeoutThread[list[InferenceChunkUncleaned]] | None
) = None
top_semantic_chunks_thread: TimeoutThread[list[InferenceChunkUncleaned]] | None = ( top_semantic_chunks_thread: TimeoutThread[list[InferenceChunkUncleaned]] | None = (
None None
) )
@ -190,20 +188,6 @@ def doc_index_retrieval(
query.offset, query.offset,
) )
# same query but with 1st vespa phase as keyword retrieval
top_base_chunks_keyword_ranking_thread = run_in_background(
document_index.hybrid_retrieval,
query.query,
query_embedding,
query.processed_keywords,
query.filters,
query.hybrid_alpha,
query.recency_bias_multiplier,
query.num_hits,
QueryExpansionType.KEYWORD,
query.offset,
)
if ( if (
query.expanded_queries query.expanded_queries
and query.expanded_queries.keywords_expansions and query.expanded_queries.keywords_expansions
@ -264,9 +248,6 @@ def doc_index_retrieval(
top_base_chunks_standard_ranking = wait_on_background( top_base_chunks_standard_ranking = wait_on_background(
top_base_chunks_standard_ranking_thread top_base_chunks_standard_ranking_thread
) )
top_base_chunks_keyword_ranking = wait_on_background(
top_base_chunks_keyword_ranking_thread
)
top_keyword_chunks = wait_on_background(top_keyword_chunks_thread) top_keyword_chunks = wait_on_background(top_keyword_chunks_thread)
@ -274,11 +255,7 @@ def doc_index_retrieval(
assert top_semantic_chunks_thread is not None assert top_semantic_chunks_thread is not None
top_semantic_chunks = wait_on_background(top_semantic_chunks_thread) top_semantic_chunks = wait_on_background(top_semantic_chunks_thread)
all_top_chunks = ( all_top_chunks = top_base_chunks_standard_ranking + top_keyword_chunks
top_base_chunks_standard_ranking
+ top_base_chunks_keyword_ranking
+ top_keyword_chunks
)
# use all three retrieval methods to retrieve top chunks # use all three retrieval methods to retrieve top chunks
@ -293,12 +270,8 @@ def doc_index_retrieval(
top_base_chunks_standard_ranking = wait_on_background( top_base_chunks_standard_ranking = wait_on_background(
top_base_chunks_standard_ranking_thread top_base_chunks_standard_ranking_thread
) )
top_base_chunks_keyword_ranking = wait_on_background(
top_base_chunks_keyword_ranking_thread top_chunks = _dedupe_chunks(top_base_chunks_standard_ranking)
)
top_chunks = _dedupe_chunks(
top_base_chunks_standard_ranking + top_base_chunks_keyword_ranking
)
logger.info(f"Overall number of top initial retrieval chunks: {len(top_chunks)}") logger.info(f"Overall number of top initial retrieval chunks: {len(top_chunks)}")