README_CONFIGURATION.md

# Ollama Chat Configuration Guide

## Overview
This application now supports conversation memory and environment-based configuration for switching between local Ollama (testing) and production Ollama servers.

## What Was Fixed

### 1. **Conversation Memory Issue**
The model wasn't remembering previous messages because:
- The app was using Ollama's `/api/generate` endpoint (stateless, single prompt)
- No conversation history was being passed to the model

**Solution:**
- Switched to Ollama's `/api/chat` endpoint
- Now reads chat history from `chat-history.json` and passes all previous messages to the model
- The model now maintains full conversation context

### 2. **Environment-Based Configuration**
Added support for automatic switching between local and production Ollama servers.

## Configuration

### Environment Variables

Create a `.env.local` file in the root directory:

```bash
# For local development (default)
OLLAMA_BASE_URL=http://localhost:11434
NODE_ENV=development

# For production deployment
# OLLAMA_BASE_URL=https://your-ollama-server.com
# NODE_ENV=production
```

### Automatic Switching

The application will automatically use:
- **Local Ollama** (`http://localhost:11434`) when `OLLAMA_BASE_URL` is not set
- **Production Ollama** when you set `OLLAMA_BASE_URL` to your deployed server

### Deployment Setup

#### Option 1: Environment Variables (Recommended)
Set the `OLLAMA_BASE_URL` environment variable in your deployment platform:

**Vercel:**
```bash
vercel env add OLLAMA_BASE_URL
# Enter: https://your-ollama-server.com
```

**Docker:**
```bash
docker run -e OLLAMA_BASE_URL=https://your-ollama-server.com ...
```

**Railway/Render/Heroku:**
Add `OLLAMA_BASE_URL` in the environment variables section of your dashboard.

#### Option 2: Git-Based Switching
Create different `.env` files for different environments:

```bash
# .env.development (for local)
OLLAMA_BASE_URL=http://localhost:11434

# .env.production (for deployment)
OLLAMA_BASE_URL=https://your-ollama-server.com
```

Add to `.gitignore`:
```
.env.local
.env*.local
```

Commit only `.env.example` to git, and set up your CI/CD to copy the appropriate file during deployment.

## Testing

### Local Testing
1. Make sure Ollama is running locally:
   ```bash
   ollama serve
   ```

2. Start the development server:
   ```bash
   npm run dev
   ```

3. The app will connect to `http://localhost:11434`

### Production Testing
1. Set `OLLAMA_BASE_URL` to your production server
2. Build and deploy:
   ```bash
   npm run build
   npm start
   ```

## How It Works

### Conversation Flow
1. User sends a message
2. Message is saved to `chat-history.json`
3. API reads all previous messages from history
4. Sends complete conversation to Ollama's `/api/chat` endpoint
5. Model responds with context from entire conversation
6. Response is saved to history and displayed

### API Endpoints Updated
- `pages/api/generate.ts` - Now uses chat API with history
- `pages/api/models.ts` - Uses environment-based URL
- `pages/api/pull.ts` - Uses environment-based URL

### Configuration Files
- `next.config.js` - Exposes `OLLAMA_BASE_URL` to the app
- `.env.local` - Local environment configuration (not committed)
- `.env.example` - Template for environment variables (committed)

## Troubleshooting

### Model not remembering conversations
- Check that `chat-history.json` exists and contains messages
- Verify the API is using `/api/chat` endpoint (not `/api/generate`)
- Clear chat history and start a new conversation

### Connection errors
- Verify `OLLAMA_BASE_URL` is set correctly
- Check that Ollama server is running and accessible
- Test the URL directly: `curl http://localhost:11434/api/tags`

### Environment variable not working
- Restart the development server after changing `.env.local`
- Verify the variable is set: `console.log(process.env.OLLAMA_BASE_URL)`
- For production, ensure the platform has the environment variable set

## Additional Notes

- The chat history is stored in `chat-history.json` in the project root
- Clearing chat history (🗑️ button) will reset the conversation context
- Each message includes metadata (timestamp, model used)
- The system automatically filters and formats messages for Ollama's API