Skip to content

Fix: Non-ASCII characters silently ignored in content gap analysis #122

Description

@himanshu231204

Description

In \openagent_eval/diagnosis/chunking.py, the content gap detection uses naive substring matching that silently ignores non-ASCII characters (Unicode, accented characters, CJK, emoji, etc.).

Current Behavior

If the retrieved context contains non-ASCII characters, the gap detection might miss them or produce false positives because string matching doesn't account for Unicode normalization.

Expected Behavior

Content gap detection should handle non-ASCII characters correctly using Unicode-aware string comparison (\unicodedata.normalize\ or similar).

Files to Look At

  • \openagent_eval/diagnosis/chunking.py\ — the chunking quality analyzer
  • \ ests/unit/test_diagnosis/test_chunking.py\ — existing tests

Good First Issue ✅

A contained fix with clear before/after behavior. Good introduction to Unicode handling in Python.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workinggood first issueGood for newcomers

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions