# IA Vein

Status: **live**

Connects soleprint to AI/LLM providers. Uses an OpenAI-compatible API interface, so it works with OpenAI, Anthropic (via proxy), local models, or any compatible endpoint.

---

## What It Does

- Generic chat completion endpoint
- Health check against the configured provider
- Use-case-specific routers mounted as sub-routes
- JSON extraction from AI responses

## Configuration

Create a `.env` file in the vein directory:

```env
AI_API_URL=https://api.openai.com/v1
AI_API_KEY=your-api-key
AI_MODEL=gpt-4o
API_PORT=8005
```

`AI_API_URL` defaults to OpenAI. Point it at any OpenAI-compatible endpoint.

## Endpoints

| Method | Path | Description |
|--------|------|-------------|
| GET | `/ia/health` | Test API connection, returns provider and model info |
| POST | `/ia/chat` | Generic chat completion |
| | `/ia/practice/*` | Practice use-case routes |

## Authentication

API key resolves in order:

1. `X-AI-Token` HTTP header
2. `.env` file value

This lets soleprint tools pass their own keys per-request.

## Chat Request

```json
{
  "messages": [
    {"role": "system", "content": "You are helpful."},
    {"role": "user", "content": "Hello"}
  ],
  "temperature": 0.7,
  "max_tokens": 1024
}
```

The response includes `content` (raw text) and `parsed` (first valid JSON object extracted from the response, if any).

## Use Cases

The IA vein supports use-case-specific routers mounted under `/ia/{usecase}/`. Each use case has its own prompts, models, and routes.

Currently implemented:

- **Practice** -- AI-powered practice sessions with structured prompts and formatted output

## Running Standalone

```bash
cd soleprint/artery/veins/ia
python run.py
# Runs on port 8005
```

Or through soleprint, where it is mounted under `/artery/ia/`.