This module owns built-in document formats, parser routing, and PDF engines.
| Type | Scope |
|---|---|
unit/ |
Format extraction, routing, metadata, shared utilities, and PDF backend behavior |
failure/ |
Corrupt, unsupported, and malformed input normalization |
security/ |
Archive bombs, safe paths, input coercion, and hardened XML parsing |
performance/ |
Input limits, concurrent parsing, bulk scale, and expensive backend reuse |
smoke/ |
Real local documents and explicitly selected PDF engines/profiles |
Run deterministic parser coverage with:
python -m pytest packages/harborrag-adapters/tests/parsers
Real documents and optional PDF/OCR engines may need native libraries, model downloads, network access, and substantial CPU/GPU resources. Follow the complete parser smoke setup before running them.