VerifyDoc¶
The trust layer for document → structured-JSON extraction. Wrap any extractor and get back JSON where every field carries a calibrated confidence, a source grounding (page + bbox / char-span), and an accept/review decision tuned to your error budget.
Every other parser tells you what it read; VerifyDoc tells you which values to trust.
Install¶
pip install verifydoc # core
pip install 'verifydoc[pdf]' # + PDF/image ingestion
30-second example¶
from verifydoc import verify
result = verify("invoice.pdf", schema="invoice_schema.json", threshold=0.8)
for f in result.fields:
print(f.path, f.value, round(f.confidence, 2), f.decision) # accept / review
Where to go next¶
- How it works — the ingest → adapter → confidence → calibration → grounding → policy pipeline.
- Integrations — drop VerifyDoc into Instructor, LangChain, LlamaIndex, Pydantic-AI, or any MCP agent (Claude Code, Cursor, Codex).
- Self-host & bots — the FastAPI server, web review app, and Telegram/WhatsApp bots; runs on your own infra.
- Why VerifyDoc — how it compares to Docling/MinerU/Marker and the commercial APIs, and what "calibrated confidence" actually buys you.
Privacy-first: every extractor can run fully local and offline; hosted API models are opt-in and comparison-only.