FormAI

demo.mp4

Hugging Face model repo: saksham0510/formai-tinyllama

FormAI was built to reduce the complexity of creating forms in tools like Google Forms and Microsoft Forms. Instead of building everything manually, you can describe the form in a prompt, generate the structure, and then edit the rest easily.

FormAI is an AI-powered form builder that turns natural-language prompts into editable, publishable forms. Describe the form you need, let the LLM draft the structure, fine-tune the fields, publish a public link, and review responses from your dashboard.

What It Does

FormAI is built around a simple workflow:

1. Prompt the assistant with a form idea, such as Customer feedback survey.
2. Generate a structured form with question types, required fields, and options.
3. Edit the form with the builder before publishing.
4. Share a public form link with respondents.
5. Collect responses and review them from the dashboard.

Features

• AI form generation from plain-language prompts.
• Editable form builder with reorder, add, remove, and field-type controls.
• Clerk authentication for protected workspaces.
• Public share links for published forms.
• Response dashboard for reviewing submissions.
• MongoDB persistence through Mongoose models.
• Resend-powered email support.
• Hugging Face LLM integration for form generation.
• Docker Compose support for local MongoDB development.

Tech Stack

Layer	Stack
Frontend	Next.js 14, React, TypeScript, Tailwind CSS
Backend	Express.js, TypeScript, Node.js
Database	MongoDB, Mongoose
Authentication	Clerk
Email	Resend
AI	Hugging Face model/API
Local Infrastructure	Docker Compose, MongoDB

Architecture

User
  -> Next.js frontend
  -> Express backend API
  -> MongoDB for forms and responses
  -> Hugging Face API or Space for AI form generation
  -> Clerk for authentication
  -> Resend for email

The frontend handles the workspace, editor, dashboard, auth UI, and public form pages. The backend owns API routes, MongoDB models, form persistence, response capture, email, and calls to the configured Hugging Face form-generation endpoint.

Project Structure

.
+-- backend/                 Express + TypeScript API
+-- frontend/                Next.js app
+-- docker/                  MongoDB initialization scripts
+-- docs/                    Deployment and troubleshooting notes
+-- docker-compose.yml       Local Docker services
+-- hugging_face.md          Hugging Face model and API contract
+-- next_step.md             Local setup walkthrough
+-- README.md                Project overview

Local Setup

Detailed local instructions are in next_step.md.

Prerequisites:

• Node.js 18+
• Docker, optional for local MongoDB
• Clerk account
• Resend account
• Hugging Face model/API or Space endpoint

Install dependencies:

cd frontend
npm install --legacy-peer-deps

cd ../backend
npm install

Create local env files:

backend/.env
frontend/.env.local

Use backend/.env.example and frontend/env.local.example as templates. Keep real secrets out of version control.

Run the app:

# terminal 1
cd backend
npm run dev

# terminal 2
cd frontend
npm run dev

Frontend runs on http://localhost:3000 and backend runs on http://localhost:8000.

Docker Mongo Setup

For local MongoDB:

docker compose up -d mongo

For the optional Mongo admin UI:

docker compose --profile debug up -d mongo-express

If Mongo becomes unhealthy after credential changes, the local Docker volume may be stale. See docs/MONGO_AUTH_SETUP.md for the reset and diagnosis flow.

Hugging Face Integration

The model repo for this project is:

https://huggingface.co/saksham0510/formai-tinyllama

That URL identifies the model, not necessarily the backend API endpoint. The backend expects HUGGINGFACE_API_URL to point to a deployed inference endpoint or Hugging Face Space API, with HUGGINGFACE_API_PATH=/generate when the endpoint uses a route path.

Read hugging_face.md for the model overview, request contract, response shape, and troubleshooting notes.

Deployment Notes

• AWS backend deployment should use Atlas through MONGODB_URL.
• Local Docker development should use DOCKER_MONGODB_URL.
• Do not set RUNNING_IN_DOCKER=true on AWS unless the backend is intentionally running inside the local-style Compose network.
• Set ALLOWED_ORIGINS and APP_BASE_URL to the deployed frontend URL.
• Use production Clerk, Resend, MongoDB, and Hugging Face credentials for production deployments.

Security Notes

• Never commit .env, .env.local, API keys, MongoDB passwords, Clerk keys, Resend keys, or Hugging Face tokens.
• Rotate any key that was pasted into chat, logs, screenshots, or committed files.
• Keep mongo-express for local debugging only.
• Prefer Atlas network allowlists or private networking for production MongoDB access.