AI Form Assistant
AI intermediary that fills forms through conversation. Speak or type naturally, AI handles the fields with intelligent parsing, validation, and multi-field updates.
Forms are often a barrier to user engagement. Users skip them, make mistakes, or struggle with validation. This open-source project puts AI in the middle—users talk naturally, AI fills everything correctly.
Built with Next.js 16 and the AI SDK, AI Form Assistant demonstrates how conversational interfaces can transform form filling from a chore into a natural dialogue. The system intelligently extracts information from natural language, validates inputs in real-time, and handles multi-field updates seamlessly.
Demos
Form Builder
Build forms dynamically in the UI without writing code. Add fields, set types and validation rules.
AI Conversation
Chat with AI to fill forms naturally. Answer questions in conversation, AI automatically fills fields as you respond.
Key Features
- Conversational Form Filling: Chat with AI, it fills fields as you answer questions naturally
- Voice Input: Browser speech recognition (works offline) for hands-free form completion
- Voice Output: Text-to-speech with auto-playback for accessible, guided form completion
- Dynamic Forms: Build forms in UI, no code required—add fields, set types, configure validation
- Smart Parsing: AI extracts information from natural language ("My name is John Doe" → firstName + lastName)
- Real-Time Validation: Per-field validation with retry logic and helpful error messages
- Multi-Field Updates: Fill several fields at once from a single response
Use Cases
Data Structurization
Dump unstructured data (like a resume) and let AI fill structured form fields, then correct what you want
Complex Multi-Step Forms
Navigate complex forms through conversation instead of clicking through multiple steps
Mobile Users
Voice input and conversational interface make form filling on mobile devices much easier
Voice Form Filling
Complete entire forms hands-free using voice commands and responses
Supported Field Types
AI Form Assistant supports a comprehensive range of form field types, each with appropriate validation:
- 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
Technical Architecture
AI Integration
- AI SDK: Vercel AI SDK for seamless integration with multiple LLM providers
- Google Gemini 2.5 Flash: Fast, efficient model for form filling conversations (cheaper, fast)
- OpenAI TTS-1: High-quality text-to-speech for voice output
- Tool Calling: AI uses structured tools to update form fields with type-safe operations
- Smart Prompting: System prompts engineered to extract structured data from natural language
Form Management
- React Hook Form: Performant form state management with minimal re-renders
- Zod Validation: Runtime type validation with schema-driven validation rules
- Dynamic Field Generation: Forms built dynamically from JSON schema stored in localStorage
- Real-Time Updates: Form fields update automatically as AI processes responses
Voice Capabilities
- Browser SpeechRecognition API: Native browser speech-to-text (works offline in Chromium browsers)
- Text-to-Speech: OpenAI TTS API integration for voice output
- Auto-Playback: Automatic voice response playback for guided form completion
Extensibility
Currently focuses on form filling, but with custom system prompts and tool extensions, you can extend functionality for:
- Sales Nurturing: Guide users toward purchases, highlight benefits as they fill forms
- 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.
Technology Stack


Frontend
- Next.js 16 - App Router with React Server Components
- TypeScript - Complete type safety across the application
- shadcn/ui - Accessible UI component library
- Tailwind CSS - Utility-first styling
AI & Processing
- Vercel AI SDK - Unified interface for AI providers
- Google Gemini 2.5 Flash - Fast, efficient LLM for form filling conversations
- OpenAI TTS-1 - High-quality text-to-speech synthesis
- Browser SpeechRecognition API - Native speech-to-text (offline capable)
Form Management
- React Hook Form - Performant form state management
- Zod - Schema validation and type inference
- LocalStorage - Client-side form schema persistence
Open Source
AI Form Assistant is an open-source project released under the MIT License.
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.