Repository navigation
馃敶 [DATA] save_chunks is insert-only - causes duplicates on re-ingest聽#13
Description
Problem
save_chunks() does INSERT without checking for existing chunks:
self.client.table('chunks').insert(chunk_dicts).execute()On re-ingestion:
- Duplicate chunks created
- Search returns garbage
Solution
Use UPSERT with unique constraint:
# Option 1: Unique on (document_id, chunk_index)
# Option 2: Unique on (document_id, content_hash)
self.client.table('chunks').upsert(
chunk_dicts,
on_conflict='document_id,chunk_index'
).execute()Requires DB migration to add unique constraint.
Files
src/infrastructure/database/supabase_repository.py- Migration for unique constraint
Acceptance Criteria
- Unique constraint on chunks table
-
save_chunksuses upsert - Re-ingestion updates instead of duplicating
Activity
FullFran commentedon Jan 26, 2026
Resuelto. ahora usa upsert con en y existe el 铆ndice 煤nico en . Cierro la issue.
FullFran commentedon Jan 26, 2026
Corrijo comentario anterior (el backtick se interpret贸 en shell).
Resuelto.
save_chunksahora usa upsert conon_conflict=document_id,chunk_indexensrc/infrastructure/database/supabase_repository.pyy existe el 铆ndice 煤nico enmigrations/001_add_chunk_unique_constraint.sql.