apunkt/SurfSense
1
0
Fork 0
mirror of https://github.com/MODSetter/SurfSense.git synced 2026-06-28 21:49:40 +02:00
Code Issues Projects Releases Packages Wiki Activity

feat: return chunk char spans from hybrid search

Browse source
This commit is contained in:
CREDO23 2026-06-19 14:53:49 +02:00
parent 90502d21d3
commit 04b679e2bf
1 changed files with 16 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

18
surfsense_backend/app/retriever/chunks_hybrid_search.py
View file

@ -440,8 +440,15 @@ class ChucksHybridSearchRetriever:
chunk_filter = numbered.c.rn <= _MAX_FETCH_CHUNKS_PER_DOC chunk_filter = numbered.c.rn <= _MAX_FETCH_CHUNKS_PER_DOC
# Select only the columns we need (skip Chunk.embedding ~12KB/row). # Select only the columns we need (skip Chunk.embedding ~12KB/row).
# start_char/end_char carry the citation span; None for legacy rows.
chunk_query = ( chunk_query = (
select(Chunk.id, Chunk.content, Chunk.document_id) select(
Chunk.id,
Chunk.content,
Chunk.document_id,
Chunk.start_char,
Chunk.end_char,
)
.join(numbered, Chunk.id == numbered.c.chunk_id) .join(numbered, Chunk.id == numbered.c.chunk_id)
.where(chunk_filter) .where(chunk_filter)
.order_by(Chunk.document_id, Chunk.position, Chunk.id) .order_by(Chunk.document_id, Chunk.position, Chunk.id)
@ -476,7 +483,14 @@ class ChucksHybridSearchRetriever:
if doc_id not in doc_map: if doc_id not in doc_map:
continue continue
doc_entry = doc_map[doc_id] doc_entry = doc_map[doc_id]
doc_entry["chunks"].append({"chunk_id": row.id, "content": row.content}) doc_entry["chunks"].append(
{
"chunk_id": row.id,
"content": row.content,
"start_char": row.start_char,
"end_char": row.end_char,
}
)
if row.id in matched_chunk_ids: if row.id in matched_chunk_ids:
doc_entry["matched_chunk_ids"].append(row.id) doc_entry["matched_chunk_ids"].append(row.id)