MEDIUM SEVERITYCONFIRMED

_suggest_similar_files char-set heuristic returns semantically unrelated files

Package
hermes-agent
Version
latest
Verified
2026-03-28

Description

hermes-agent's `_suggest_similar_files` uses a character-set intersection heuristic. `'unrelated_file.py'` scores 0.86 against `'main.py'` (above the 0.50 threshold) because the character sets overlap heavily. The heuristic returns semantically irrelevant files as suggestions.

Reproduction

# _suggest_similar_files('main.py')
# 'unrelated_file.py' scores 0.86 via character-set intersection
# Threshold is 0.50 — semantically unrelated file returned
← All bugsScan your code →