# LLM Proxy

HTTP proxy for LLM APIs with streaming support and chunk processing.

## Usage

```bash
./llm-proxy          Start the proxy
./llm-proxy -h      Show help
```

## Configuration

Configuration file (optional): `llm-proxy.toml`

Environment variables take priority over config file.

| Variable | Alternative | Description | Default |
|----------|-------------|-------------|---------|
| `UPSTREAM_URL` | `OPENAI_API_BASE` | Upstream LLM API URL | `https://api.openai.com/v1/chat/completions` |
| `LISTEN_ADDR` | - | Listen address | `127.0.0.1:8080` |
| `API_KEY` | `OPENAI_API_KEY` | Upstream API key | - |
| `INSECURE` | - | Skip TLS verification | `false` |

## Example

```bash
# Via environment variables
UPSTREAM_URL=https://api.openai.com/v1/chat/completions \
API_KEY=sk-... \
LISTEN_ADDR=127.0.0.1:8080 \
./llm-proxy
```

```toml
# Via config file (llm-proxy.toml)
upstream_url = "https://api.openai.com/v1/chat/completions"
listen_addr = "127.0.0.1:8080"
api_key = "sk-..."
insecure = false
```

## Endpoints

- `GET /health` - Health check
- `/*` - Proxies all requests to upstream

## Streaming

Supports SSE (`text/event-stream`) and NDJSON (`application/x-ndjson`) streaming. Each chunk is processed via `processChunk()` before forwarding.