AI Form Assistant

AI intermediary that fills forms through conversation. Speak or type naturally, AI handles the fields.

Why

Forms suck. Users skip them, make mistakes, or struggle with validation. This puts AI in the middle—users talk naturally, AI fills everything correctly.

Use cases:

• Easy data structurization (dump a resume → AI fills fields → correct what you want)
• Complex multi-step forms
• Mobile users who hate typing
• Voice form filling

What it does

• Conversational form filling: Chat with AI, it fills fields as you answer
• Voice input: Browser speech recognition (works offline)
• Voice output: Text-to-speech with auto-playback
• Dynamic forms: Build forms in UI, no code
• Smart parsing: AI extracts info from natural language ("My name is John Doe" → firstName + lastName)
• Validation: Real-time per-field validation with retry logic
• Multi-field updates: Fill several fields at once

Demo

Form Builder

AI Conversation

What it can do

Currently focuses on form filling. With custom system prompts and tool extensions, you can:

• Sales nurturing: Guide users toward purchases, highlight benefits as they fill
• Onboarding assistance: Explain options/fields before users choose, answer questions in context
• Smart defaults: Suggest values based on previous answers or user context
• Multi-step workflows: Chain forms, conditionally show fields, collect data across sessions

Extend by modifying the system prompt in src/app/api/chat/route.ts and adding custom tools.

Field types

• Text - Single-line text input
• Textarea - Multi-line text input
• Email - Email address with validation
• Phone - Phone number with pattern validation
• URL - Website URL with protocol validation
• Number - Numeric input with min/max validation
• Date - Date picker
• Time - Time picker
• Checkbox - Boolean true/false values
• Single choice (Select) - Choose one option from a list
• Multiple choice (Multi-select) - Choose multiple options from a list
• Range/Slider - Numeric range selector
• File upload - File selection and upload

---

Quick start

# Clone
git clone <repo-url>
cd ai-form

# Install
bun install

# Setup env (see below)
cp .env.example .env

# Run
bun dev

Open http://localhost:3000

Environment variables

Copy .env.example to .env and fill:

# form filling chat (cheaper, fast)
GOOGLE_GENERATIVE_AI_API_KEY="your-google-api-key"

# text-to-speech
OPENAI_API_KEY="your-openai-api-key"

How it works

1. Build form: Use form builder (gear icon) → add fields, set types/validation
2. Start chat: Click "Start Conversation"
3. Talk to AI: Answer questions naturally (voice or text)
4. AI fills form: Fields update automatically as you talk
5. Submit: AI validates and submits when complete

Form definition lives in localStorage. Export/import coming soon.

Note

This is a demo/educational repo. Tool execution (form field updates) happens client-side. In production, data saves would typically happen server-side before passing to the client for security and data integrity.

Tech

• Next.js 16 (App Router)
• AI SDK
• Google Gemini 2.5 Flash (chat)
• OpenAI TTS-1 (speech)
• React Hook Form + Zod (validation)
• shadcn/ui
• Browser SpeechRecognition API

License

MIT