Skip to content

Improve PDF extract cleanup and wire hybrid expand for agent search - #55

Open
prinskumar-tigergraph wants to merge 4 commits into
mainfrom
improve-pdf-extract-and-hybrid-expand
Open

Improve PDF extract cleanup and wire hybrid expand for agent search#55
prinskumar-tigergraph wants to merge 4 commits into
mainfrom
improve-pdf-extract-and-hybrid-expand

Conversation

@prinskumar-tigergraph

Copy link
Copy Markdown
Contributor

Description:

Summary

  • Cleaner PDF extract/chunking so chart labels and table numbers survive into the graph (picture-text normalization, unglue numbers, avoid fullwidth-digit collapse, mojibake page recovery).
  • Multimodal image prompt asks for categorical category: value transcription, not trend-only summaries.
  • Agent hybrid_search now honors hybrid_expand / hybrid_method from graph config, and keeps keyword seeds if vector seeding fails.

Why this helps

Better extract → more exact facts in chunks. Hybrid expand (keywords + vector) → agent is more likely to retrieve those chunks for table/chart questions. Together this raises grounded answer quality without question-specific hacks.

Normalize picture-text and numeric cleanup, recover mojibake pages, keep chart labels in chunks, and enable keyword+vector hybrid expand for agent search without eval-shaped age-band post-processing.
@prinskumar-tigergraph
prinskumar-tigergraph force-pushed the improve-pdf-extract-and-hybrid-expand branch from 23ec1c0 to 3a26df3 Compare July 30, 2026 13:40
Avoids an unnecessary rename that noisied the PR diff.
Restore the original OCR figure comment in the chunker to avoid noisy comment-only diff.
… docs.

No measured accuracy gain from these settings in Toppan runs; keep extract/chunk product fixes only.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants