fix(analyzer): prioritize matching source and preserve Unicode queries by harshitethic · Pull Request #49 · harshitethic/RepoPilot

harshitethic · 2026-10-25T13:07:52Z

Reproduced issues

search_code(..., limit=0) and negative limits returned a result instead of respecting the requested budget.
build_context used .lower() and missed Unicode casefold equivalents, even though source search supports them.
Symbol names inside backticks and parentheses were not extracted from questions reliably.
README filename bonuses could outrank a file containing the requested symbol, causing irrelevant AI context.

Fix these behaviors without network access or changing repository ingestion. Four focused regressions each accompany their respective correction.

Reproduced zero-limit emission and two irrelevant README selections before correction.
After each fix, the focused analyzer regression suite passed.
Python 3.12.3 in isolated virtual environment with locked backend and test requirements: 111/111 backend unit and HTTP-level tests passed.
python3 -m compileall -q backend/services backend/tests passed.
No live Ollama model, external clone, browser UI, or deployed service was exercised; GitHub Actions remains a separate check.

harshitethic added 8 commits October 25, 2026 09:07

fix(search): honor nonpositive result bounds

17707f6

test(search): assert zero and negative budgets

d90e3ed

fix(context): Unicode-casefold source matches

b239bed

test(context): verify non-ASCII symbol retrieval

dd21872

fix(context): extract identifiers from code punctuation

c32a332

test(context): check backtick-wrapped symbols

6424b81

fix(context): prefer true source relevance over filename priors

c01e4d2

test(context): assert matching source outranks README

ce706e5

harshitethic merged commit 293acab into main Oct 25, 2026
9 checks passed