danvics/haiku.rag
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Set repeat_table_header=False to match docling serve

Browse source
This commit is contained in:
Yiorgis Gozadinos 2026-03-05 13:02:18 +02:00
parent 30ebd6694b
commit d460d553f3
No known key found for this signature in database
2 changed files with 16166 additions and 16344 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
haiku_rag_slim/haiku/rag/chunkers/docling_local.py
View file

@ -83,6 +83,9 @@ class DoclingLocalChunker(DocumentChunker):
tokenizer=tokenizer, tokenizer=tokenizer,
merge_peers=config.processing.chunking_merge_peers, merge_peers=config.processing.chunking_merge_peers,
serializer_provider=serializer_provider, serializer_provider=serializer_provider,
# Context expansion reconstructs full tables, so repeating
# headers in each chunk is redundant and inflates chunk count.
repeat_table_header=False,
) )
elif self.chunker_type == "hierarchical": elif self.chunker_type == "hierarchical":
serializer_provider = _create_markdown_serializer_provider( serializer_provider = _create_markdown_serializer_provider(

32507
tests/cassettes/test_chunker/test_local_and_serve_chunkers_produce_same_output.yaml
View file

File diff suppressed because one or more lines are too long