A deployable web UI for DeepRead — AI-native document processing: two engines, Extract and Deep Extract, measured against other vendors on deepread.tech/benchmarks.
After deploying, set DEEPREAD_API_KEY as an environment variable. Get a free key (2,000 pages/month, no credit card) at deepread.tech/dashboard.
DeepRead is a document AI platform with one API key:
- OCR & Structured Extraction — Extract Markdown text or typed JSON fields from PDFs and images. Each field carries a review flag (
needs_review) and the reason, so you know exactly which values to check. Two engines: Extract ($15 per 1,000 pages) and Deep Extract ($35), chosen withpipeline=extractorpipeline=deep-extract - PDF Form Filling (Enterprise plans) — Fill blank PDF forms with AI vision (works on scanned, non-editable forms — no AcroForm required)
- PII Redaction (Enterprise plans) — Detect and redact 14 types of PII (names, SSNs, credit cards, medical records, etc.) with irreversible black bars, built for HIPAA/GDPR redaction workflows
- Human review — name a reviewer with
reviewer_emailand DeepRead runs the review: they get a link, correct values, and your webhook receives the result
Enterprise plans also include searchable PDF, incognito processing, retention controls and bring-your-own LLM key.
Four modes via simple drag-and-drop UI:
- OCR Extract — Upload PDF/image, get extracted text
- Structured JSON — Define a schema, get typed fields with review flags
- Form Fill — Upload a blank form and the values to fill (Enterprise plans; other plans get a
402) - PII Redact — Remove 14 PII types with irreversible black bars (Enterprise plans; other plans get a
402)
A free key covers OCR Extract and Structured JSON (2,000 pages a month, 50 pages per document).
git clone https://github.com/deepread-tech/deepread-webapp.git
cd deepread-webapp
pip install -r requirements.txt
export DEEPREAD_API_KEY=sk_live_your_key
python app.py
# Open http://localhost:8080| Variable | Required | Description |
|---|---|---|
DEEPREAD_API_KEY |
Yes | Get free at deepread.tech/dashboard |
PORT |
No | Defaults to 8080 (Railway sets this automatically) |
- Python 3.10+
- Flask (web framework)
- Gunicorn (production server)
- Requests (HTTP client for DeepRead API)
GET /— Upload UIPOST /extract— Process document (returns DeepRead API response)GET /health— Health check (used by Railway)
- Examples: github.com/deepread-tech/deepread-demo — Python, Node.js, cURL examples for OCR, structured extraction, form fill, PII redaction
- n8n Node: n8n-nodes-deepread — drag-and-drop in n8n workflows
- Claude Code:
/plugin install deepread@claude-community - Universal (50+ AI agents):
npx skills add deepread-tech/skills
MIT