anthonypdawson / vector-inspector / 27625198247

80%

master: 80%

Pull #39

github

refactor: implement code review suggestions for robustness

Improvements based on code review feedback:

1. **Logging for fallback detection**: Add log_info when content column
   detection falls back to first text column for debugging visibility

2. **Metadata padding warnings**: Log warning when LanceDB pads sparse
   metadata arrays to indicate potential data quality issues

3. **Thread-safety**: Add threading.Lock to content column cache for
   safe concurrent access across threads

4. **Performance documentation**: Add detailed performance section to
   content_column_detection.md documenting caching behavior and
   characteristics

5. **Configurable Ollama timeout**: Add OLLAMA_TIMEOUT environment
   variable to allow override of default 30s timeout, with updated
   documentation and troubleshooting guide

Also improved error handling for empty Ollama embeddings (now raises
error instead of returning empty list) and updated tests accordingly.

All 46 tests passing with 80% coverage maintained.

150 of 309 new or added lines in 14 files covered. (48.54%)

15296 of 19056 relevant lines covered (80.27%)

0.8 hits per line