AI Image Chat

This is a full-stack chat application that uses AI to analyze and describe images. Users can upload an image or send text prompts, and the application returns a detailed description along with relevant meta tags.

Features

• Image Analysis: Upload images to receive AI-generated descriptions and keyword tags.
• Persistent Chat: Conversation history, including images, is stored in a PostgreSQL database and persists across browser sessions using a unique session ID.
• Full Stack Architecture: Built with a Next.js backend that handles database operations and communication with the AI service.

Tech Stack

• Framework: Next.js 15 (App Router)
• Language: TypeScript
• Styling: Tailwind CSS
• Database: Supabase (PostgreSQL)
• ORM: Prisma
• AI Service: Pollinations AI (Proxy for OpenAI/Vision models)

Setup

1. Clone the repository.

2. Install dependencies:

   npm install
   

3. Set up your environment variables in .env (requires a DATABASE_URL from Supabase and the hugging face api token, you can also replace it with some other api token of your prefered model). You can also check the .env.example file for reference

4. Sync the database schema: npx prisma db push.

5. Run the development server: npm run dev.

Testing via CLI (Curl)

You can verify the API is working without using the frontend by running these commands in your terminal.

1. Test Image Analysis

Create a file named test-payload.json containing a test session ID and a tiny Base64 image (a red dot):

{
  "sessionId": "curl-test-session-1",
  "message": "What color is this?",
  "imageBase64": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAUAAAAFCAYAAACNbyblAAAAHElEQVQI12P4//8/w38GIAXDIBKE0DHxgljNBAAO9TXL0Y4OHwAAAABJRU5ErkJggg=="
}

Run the curl command to send this file to your local API:

curl -X POST http://localhost:3000/api/chat \
  -H "Content-Type: application/json" \
  -d @test-payload.json

Expected Output: The AI should reply identifying the image as a "red dot" or "red square".

2. Verify Chat History (GET)

Check if the database successfully saved your conversation by retrieving the history for that session:

curl "http://localhost:3000/api/chat?sessionId=curl-test-session-1"

Expected Output: A JSON array [...] containing the user message and the AI's response from the previous test.