Skip to content

justin-thakral/DullyPDF

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

148 Commits
.github/workflows
.github/workflows
 
 
backend
backend
 
 
config
config
 
 
frontend
frontend
 
 
internal_stats
internal_stats
 
 
mcp
mcp
 
 
quickTestFiles
quickTestFiles
 
 
runs
runs
 
 
scripts
scripts
 
 
test-results
test-results
 
 
test
test
 
 
tmp
tmp
 
 
.coveragerc
.coveragerc
 
 
.dockerignore
.dockerignore
 
 
.firebaserc
.firebaserc
 
 
.gcloudignore
.gcloudignore
 
 
.gitignore
.gitignore
 
 
AGENTS.md
AGENTS.md
 
 
Dockerfile
Dockerfile
 
 
Dockerfile.ai-rename-remap
Dockerfile.ai-rename-remap
 
 
Dockerfile.detector
Dockerfile.detector
 
 
Dockerfile.detector.gpu
Dockerfile.detector.gpu
 
 
GIT_WORKFLOW.md
GIT_WORKFLOW.md
 
 
GSC.md
GSC.md
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 
clean.py
clean.py
 
 
debug_checkbox_labels.py
debug_checkbox_labels.py
 
 
docker-compose.yml
docker-compose.yml
 
 
dump_openai_rename_payload.py
dump_openai_rename_payload.py
 
 
file_search_and_fill_crediting.md
file_search_and_fill_crediting.md
 
 
firebase.json
firebase.json
 
 
firestore.indexes.json
firestore.indexes.json
 
 
package-lock.json
package-lock.json
 
 
package.json
package.json
 
 
pytest.ini
pytest.ini
 
 

Repository files navigation

DullyPDF

dullypdf.com — Free PDF form automation platform that converts existing PDFs into reusable fillable templates with AI field detection, then fills them from spreadsheets, web form responses, API calls, or scanned documents.

Upload any PDF, let the AI detection pipeline find every text field, checkbox group, radio button, date, and signature region, then clean the template in the visual editor and reuse it across every fill workflow the platform supports. No Acrobat license needed.

Platform Workflows

Workflow What it does
PDF to Fillable Form Detect fields automatically and build a reusable fillable template from any flat, scanned, or native PDF
Fill PDF from Spreadsheet Map template fields to CSV, Excel, JSON, or TXT headers and bulk fill from matching rows — data stays in-browser
Collect Answers by Web Form Publish a hosted respondent form so clients submit answers that fill the PDF when you generate it
JSON-to-PDF API Endpoint Expose a template-scoped REST endpoint for programmatic document generation from structured payloads
E-Signature Workflow Route filled PDFs into email-based signing requests with immutable record freezing under U.S. E-SIGN and UETA
Batch Fill PDF Forms Automate high-volume filling across document packets and template groups from a single data source
Fill from Photos and Scanned Documents Extract data from IDs, invoices, pay stubs, and insurance cards using AI vision and populate template fields
PDF Field Detection Tool Review AI-detected field candidates with confidence scoring before committing the template layout
Field Name Normalization Standardize field names across messy or inconsistent PDF templates for reliable downstream mapping

Industry-Specific Solutions

Industry Supported document types
Healthcare Intake Automation Patient registration, dental intake, HIPAA consent, and medical history forms
Insurance Certificate Processing ACORD certificates, carrier-specific policy forms, and COI packet workflows
ACORD Form Filler ACORD 25, 24, 27, 28, 126, and 140 certificate of insurance templates
HR Onboarding Document Automation New hire packets, benefits enrollment, W-4, I-9, and compliance forms
Shipping and Freight Paperwork Bills of lading, delivery receipts, customs declarations, and transport logs
Loan Application PDF Automation Mortgage applications, financial disclosures, and lending document packets
Rental and Lease Packet Filler Rental applications, lease agreements, property disclosure, and tenant screening forms
Legal Document Workflow Contracts, court filings, retainer agreements, and case intake packets
Government Form Processing Permit applications, tax forms, licensing packets, and public records requests
Student Application Automation Enrollment forms, transcript requests, financial aid applications, and registration packets
Nonprofit Intake Automation Grant applications, volunteer onboarding, donor pledge forms, and program intake

Documentation

Blog

Browse

Development

FastAPI + React app for detecting PDF form fields, renaming candidates with OpenAI, editing fields in a PDF viewer, and publishing native Fill By Link respondent forms from saved templates. The main pipeline is CommonForms (by jbarrow) detection, optional OpenAI rename, schema-only mapping, and Search & Fill or Fill By Link respondent selection at generation time.

Getting Started

This guide covers a quick local setup for the main pipeline and points to small, tracked fixtures for manual testing.

Prereqs

  • Node.js (for frontend tooling)
  • Python 3.10+ (for backend)

Run the full stack (dev)

From the repo root:

npm install
npm run dev

This starts the FastAPI backend and Vite frontend together, and also starts Stripe CLI webhook forwarding to the local billing webhook when STRIPE_SECRET_KEY is configured.

Notes for Stripe local billing:

  • npm run dev injects the Stripe CLI session whsec_... into the backend process for that run.
  • Checkout health enforcement is forced off for local CLI forwarding (STRIPE_ENFORCE_WEBHOOK_HEALTH=false) because Stripe CLI forwarding does not create a dashboard webhook endpoint.
  • Set STRIPE_DEV_LISTEN_ENABLED=false to skip automatic Stripe forwarding.

You can still run frontend/backend separately with npm run backend:dev and npm run frontend:dev.

Open the UI at http://localhost:5173.

Run the internal stats dashboard

From the repo root:

npm run stats

This starts a standalone local dashboard on http://127.0.0.1:5174. It is intentionally not part of the deployed frontend or backend app, and it reads the production Firestore project directly with your local Google credentials instead of using DullyPDF app sign-in.

Run the prod-like stack

From the repo root:

npm run dev:stack

This runs the backend in Docker on host port 8010, starts the frontend dev server on 5173, targets the dullypdf-dev Cloud Tasks / Cloud Run services, and also auto-starts Stripe CLI webhook forwarding to http://localhost:8010/api/billing/webhook when STRIPE_SECRET_KEY is configured.

Notes for Stripe stack billing:

  • npm run dev:stack injects the Stripe CLI session whsec_... into the backend container for that run.
  • Checkout health enforcement is forced off for local CLI forwarding (STRIPE_ENFORCE_WEBHOOK_HEALTH=false) because Stripe CLI forwarding does not create a dashboard webhook endpoint.
  • Set STRIPE_DEV_LISTEN_ENABLED=false to skip automatic Stripe forwarding.

Environment setup

Backend dev env is local-only and created on first run:

  • Backend: env/backend.dev.env (from config/backend.dev.env.example)

Frontend uses committed public env files:

  • config/public/frontend.dev.env
  • config/public/frontend.stack.env
  • config/public/frontend.prod.env

Optional local frontend overrides can be added in ignored files:

  • env/frontend.dev.local.env
  • env/frontend.stack.local.env
  • env/frontend.prod.local.env

npm run backend:dev loads env/backend.dev.env, then pulls Firebase Admin credentials via Secret Manager if configured. npm run frontend:dev builds frontend/.env.local from config/public/frontend.dev.env and appends local override files when present.

OpenAI (optional)

Rename and schema mapping require OPENAI_API_KEY. If the key is missing, those actions fail while CommonForms (by jbarrow) detection still works.

Plan limit messaging

  • Free defaults are 5 saved forms, 5 detect pages, 50 fillable pages, no active Fill By Link cap with 25 accepted responses/month across the account, 1 API Fill endpoint with 250 successful fills/month and 50 pages/request, 50 Fill by File credits/month, 25 sent signing requests/month, 25 generated PDF downloads/month, and 10 starter credits.
  • Premium defaults are 100 saved forms, 100 detect pages, 1,000 fillable pages, no active Fill By Link cap with 10,000 accepted responses/month across the account, 20 API Fill endpoints with 10,000 successful fills/month and 500 pages/request, 10,000 Fill by File credits/month, 10,000 sent signing requests/month, unlimited generated PDF downloads, and a 500-credit monthly pool before refill packs.
  • Workspace generated PDF downloads reset on the backend by UTC month. Saving templates, API Fill outputs, respondent downloads, and signing artifacts use their own workflow limits and are not charged against that workspace download quota.
  • Respondents fill a DullyPDF-hosted HTML form. The final PDF is generated later when the owner selects a respondent inside the workspace.
  • Public plan summaries are available at /free-features and /premium-features.

Quick test files

Use the tracked fixtures in quickTestFiles/:

  • quickTestFiles/new_patient_forms_1915ccb015.pdf
  • quickTestFiles/new_patient_forms_1915ccb015_mock.csv (Search & Fill rows)
  • quickTestFiles/healthdb_vw_form_fields.csv (schema headers)

Notes:

  • CSV/Excel/JSON rows stay in the browser; only headers/types are sent to the server.
  • Do not add PHI/PII to tracked files.

API Fill QA

Targeted API Fill checks are available at the repo root:

npm run test:backend:template-api
npm run coverage:backend:template-api
npm run test:frontend:template-api
npm run test:playwright:template-api
npm run test:qa:template-api
  • test:backend:template-api runs the focused backend unit and integration suite for publish, public fill, service normalization, and Firestore bookkeeping.
  • coverage:backend:template-api reports coverage only for the core API Fill backend modules.
  • test:frontend:template-api runs the focused dialog and hook unit tests.
  • test:playwright:template-api runs both the harnessed browser regression and the authenticated /ui owner flow.
  • test:qa:template-api chains the focused backend, frontend, and Playwright checks together.

Cleanup

Use the repo cleanup entrypoint to clear generated artifacts:

python3 clean.py --mcp --mcp-logs --mcp-screenshots
python3 clean.py --runs --tmp --test-results
python3 clean.py --field-detect-logs --mcp-bug-logs --frontend-tmp
python3 clean.py --outbound-leads
python3 clean.py --bug-reports --mcp-security-logs
python3 clean.py --coverage --pytest-cache --python-cache --frontend-dist --output --repo-logs --pipeline-improve
python3 clean.py --all --dry-run

Each directory also ships its own cleanOutput.py script (see mcp/, runs/, test-results/, tmp/, backend/fieldDetecting/logs/, mcp/codexBugs/logs/, and frontend/). Root-level cleanup also supports bug-report folders and local cache/build artifacts.

Fullstack Read Audit

Scope: main pipeline only (backend + detector + frontend), with supporting tooling called out separately. Legacy OpenCV pipeline in legacy/ is excluded from the runtime path.

1) Languages in this stack

  • Python (backend API + detector services)
  • TypeScript (frontend app + build config)
  • JavaScript (Node scripts/tooling + MCP server)
  • HTML (frontend entry)
  • CSS (frontend styling)
  • Bash/Shell (dev/deploy/runtime scripts)
  • YAML (container/service config)
  • JSON (app/tooling config and manifests)
  • Markdown (project docs)

Also present (secondary tooling): PowerShell scripts (.ps1).

2) Backend API libraries (backend/requirements.txt)

  • fastapi==0.128.2
  • uvicorn==0.30.6
  • pdfplumber==0.11.9
  • pymupdf==1.24.9 (fitz)
  • opencv-python-headless==4.10.0.84
  • numpy==1.26.4
  • pillow==12.0.0
  • openai==2.11.0
  • python-multipart==0.0.22
  • httpx==0.28.1
  • pypdf==6.6.2
  • firebase-admin==7.1.0
  • google-cloud-storage==3.9.0
  • google-cloud-tasks==2.21.0
  • protobuf==5.29.6
  • PyJWT==2.10.1

3) Detector-specific libraries

From backend/requirements-detector.txt and Dockerfile.detector:

  • commonforms==0.2.1
  • torch==2.9.1+cpu
  • torchvision==0.24.1+cpu

4) Backend test libraries (backend/requirements-dev.txt)

  • pytest>=8.0,<9
  • pytest-mock>=3.14,<4
  • pytest-cov>=5,<7

5) Frontend runtime libraries (frontend/package.json)

  • react@^19.2.0
  • react-dom@^19.2.0
  • pdfjs-dist@^4.5.136
  • firebase@^10.11.0
  • firebaseui@^6.1.0
  • read-excel-file@^6.0.3

6) Frontend build/test/tooling libraries

  • typescript@~5.9.3
  • vite@^7.2.4
  • @vitejs/plugin-react@^5.1.1
  • vitest@^4.0.18
  • @testing-library/react@^16.2.0
  • @testing-library/user-event@^14.6.1
  • jsdom@^26.1.0
  • eslint@^9.39.1
  • @eslint/js@^9.39.1
  • typescript-eslint@^8.46.4
  • eslint-plugin-react-hooks@^7.0.1
  • eslint-plugin-react-refresh@^0.4.24
  • @types/node@^24.10.1
  • @types/react@^19.2.5
  • @types/react-dom@^19.2.3
  • globals@^16.5.0
  • undici override: 6.23.0

7) Root workspace tooling (package.json)

  • concurrently@^8.2.2 (run frontend/backend together)
  • @playwright/test@^1.57.0 (E2E tooling)

8) MCP server libraries (mcp/server/package.json)

  • @modelcontextprotocol/sdk@^1.25.1
  • axios@^1.6.7
  • dotenv@^16.4.5
  • form-data@^4.0.0

9) Platform and service dependencies in the live pipeline

  • FastAPI backend service on Cloud Run
  • Dedicated detector service on Cloud Run
  • Firebase Hosting (SPA + selected /api rewrites)
  • Firebase Auth / Identity Platform
  • Firestore (session/schema/request metadata)
  • Google Cloud Storage (forms/templates/artifacts)
  • Google Cloud Tasks (detector job queue)
  • OpenAI API (rename + schema mapping)
  • reCAPTCHA Enterprise (contact/signup risk checks)
  • Gmail API (contact form email delivery)

10) Not on the main runtime path

  • docker-compose.yml defines SQL Server (mcr.microsoft.com/mssql/server:2022-latest) for local/support scenarios.
  • Backend docs explicitly state SQL/Postgres integrations are not part of the current main runtime path (moved to legacy/).

Audit inputs reviewed

  • backend/requirements.txt
  • backend/requirements-detector.txt
  • backend/requirements-dev.txt
  • frontend/package.json
  • package.json
  • mcp/server/package.json
  • Dockerfile
  • Dockerfile.detector
  • docker-compose.yml
  • backend/README.md
  • backend/fieldDetecting/README.md
  • backend/fieldDetecting/docs/commonforms.md
  • backend/fieldDetecting/docs/rename-flow.md
  • frontend/README.md
  • frontend/docs/overview.md
  • frontend/docs/api-routing.md

More docs

  • backend/README.md
  • frontend/README.md
  • backend/fieldDetecting/docs/README.md
  • GIT_WORKFLOW.md

About

DullyPDF allows you to automatically convert PDFs to fillable forms for free. Search & Fill from API or Database with US esignature support. Backend is Python + FastAPI cloud run hosted. Frontend is TypeScript + React firebase hosted.

dullypdf.com

Topics

react pdf opencv firebase typescript signature python3 pytest firestore-database uvicorn vite fastapi fillable-forms playwright pdfplumber pdftemplate pdfai fillable-forms-and-signatures commonforms automatic-pdf-to-fillable-form

Resources

Readme

License

View license
Activity

Stars

5 stars

Watchers

2 watching

Forks

0 forks
Report repository

Contributors

Languages