Spaces:
Sleeping
Sleeping
Commit History
-
Stop tracking .coverage, and ignore local tool output
Claude +1: Claude Opus 5Unverified signatureA `git add -A` in the previous commit swept up .coverage, the binary SQLite file `pytest --cov` writes. It is machine-specific build output and has no business in the tree. Plain `pytest` - the command CONTRIBUTING.md documents - doesn't produce it, which is why the gap went unnoticed: pytest and ruff both write self-ignoring cache directories, and coverage.py does not. Ignore .coverage, htmlcov and .ruff_cache so running any of the documented commands leaves the tree clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MSNSYnceqvdVz7Csis4K9e72a7212
-
Mirror SERPent from Hugging Face Space with bug fixes
ClaudeUnverified signatureMirrors OrganizedProgrammers/SERPent (https://huggingface.co/spaces/OrganizedProgrammers/SERPent) with the following fixes applied on top: - app.py: make /serp/search try all queries concurrently instead of sequentially, and surface per-query backend failures in the response's error field instead of silently dropping them when at least one other query in the batch succeeded. - app.py: add opt-in SERPENT_API_KEY protection (X-API-Key or Authorization: Bearer) gating REST and MCP alike via HTTP middleware, closing the unauthenticated open-scraping-proxy surface. Unset by default, so behavior is unchanged until explicitly enabled. - app.py: load docs/docs.md safely (context-managed read, fallback description) instead of an unguarded open().read() at import time. - serp.py: raise a clear BrowserUnavailableError instead of an AttributeError when Playwright failed to start and pw_browser is None. - serp.py: clamp n_results to [1, 100] in SerpQuery so a negative or huge val…cb1e234