Fill any AcroForm PDF from JSON — no manual field mapping.
Open-source FastAPI service · browser playground · Python SDK · Docker image
Try in Codespaces · Install · API · Recipes · Docs
Keywords: pdf · acroform · form-filling · fastapi · python · automation · api · docker · document-processing · govtech · w9 · self-hosted
Turn {"firstname":"Jane","lastname":"Doe"} into a filled PDF — even when the form uses txtFName, givenName, or field_12.
| Input | Output |
|---|---|
| Any fillable AcroForm PDF | Completed PDF with fields written |
| JSON user profile | Mapped automatically via aliases + normalization |
| Optional AI (off by default) | Semantic inference for opaque field names |
Popular uses: W-9 / tax forms · HR onboarding packets · government PDFs · insurance intake · workflow automation (n8n, Zapier, curl)
Upload a PDF, paste JSON, download the result — no Postman required.
Try free: Open in GitHub Codespaces → auto-opens /playground
# Docker (fastest local)
docker run --rm -p 8000:8000 -e API_AUTH_ENABLED=false \
ghcr.io/lindseystead/ai-pdf-autofiller:latest
# → http://localhost:8000/playground
# Or fill from curl
curl -s -X POST http://localhost:8000/fill \
-F "pdf_file=@samples/sample_form.pdf;type=application/pdf" \
-F 'user_data={"firstname":"Jane","lastname":"Doe","dob":"1990-01-01"}' \
-F "strict=true" -o filled.pdf| Method | Command |
|---|---|
| Docker | docker run -p 8000:8000 -e API_AUTH_ENABLED=false ghcr.io/lindseystead/ai-pdf-autofiller:latest |
| GitHub Release | curl -fsSL .../scripts/install-from-release.sh | bash |
| From source | git clone https://github.com/lindseystead/ai-pdf-autofiller.git && cd ai-pdf-autofiller && pip install -r requirements-dev.txt && make run-api |
from pdf_autofiller import fill
fill("form.pdf", {"firstname": "Jane", "lastname": "Doe"}, "filled.pdf")| Method | Path | Description |
|---|---|---|
GET |
/playground |
Browser UI |
GET |
/health |
Health + dependency checks |
POST |
/fill |
PDF in, filled PDF out |
Full contract: docs/API.md
Manual PDF field mapping does not scale. AI-only fillers are hard to audit. PDF Autofiller is deterministic-first: most fields match via normalization and aliases with no API key required.
- FastAPI HTTP API with structured error codes
- Browser playground at
/playground - Python SDK (
fill()+PDFAutofillerClient) - W-9 and HR alias packs + recipes
- Docker on GHCR · Render blueprint · GitHub Release wheels
- Auth, rate limits, and upload guards on by default
flowchart LR
A[PDF + JSON] --> B[pipeline]
B --> C[pdf_reader]
C --> D{AI?}
D -->|optional| E[field_semantics]
D -->|skip| F[mapping]
E --> F --> G[pdf_writer] --> H[Filled PDF]
| Recipe | Form |
|---|---|
| w9.md | IRS W-9 |
| hr-onboarding.md | Employee intake |
| sample-form.sh | One-line demo |
| Doc | Contents |
|---|---|
| docs/API.md | Endpoints and errors |
| docs/OPERATIONS.md | Config and deployment |
| docs/TESTING.md | Tests and CI |
| docs/integrations/ | n8n, Zapier, LangChain |
| CONTRIBUTING.md | How to contribute |
| CHANGELOG.md | Release history |
make test && make lint && make smoke-check105 tests · 85%+ coverage · Python 3.11 & 3.12
MIT — LICENSE
Star the repo if this saves you from mapping txtFirstName by hand.