Skip to content

fix: surface model extractor fallback warnings#38

Open
Sanjays2402 wants to merge 1 commit into
KRLabsOrg:mainfrom
Sanjays2402:fix/model-extractor-warnings
Open

fix: surface model extractor fallback warnings#38
Sanjays2402 wants to merge 1 commit into
KRLabsOrg:mainfrom
Sanjays2402:fix/model-extractor-warnings

Conversation

@Sanjays2402

Copy link
Copy Markdown

What

Format-detection errors and legacy token-budget truncation now emit warnings instead of silently falling back or dropping sentences. Adds regression coverage for both paths.

Fixes #33

Checklist

  • pytest tests/ -v passes
  • ruff check packages/core/verbatim_core/ verbatim_rag/ api/ tests/ and
    ruff format packages/core/verbatim_core/ verbatim_rag/ api/ tests/ pass
  • Tests added/updated for behavior changes
  • Docs updated if user-facing (README / docs/ / docstrings)

Rights & sign-off (required)

  • I certify that I have the right to submit this code and that it may be
    distributed under the repository's MIT license
    (see CONTRIBUTING).

Warn when highlighter format detection falls back to the legacy model and when legacy token packing drops remaining sentences. Add regression coverage for both silent failure paths.

Signed-off-by: Sanjay Santhanam <51058514+Sanjays2402@users.noreply.github.com>
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.

ModelSpanExtractor: silent fallback to legacy qa_model path can turn env issues into zero-span results

1 participant