README.md

---
title: Haiku API
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
---

# Haiku API

OpenAI-compatible API proxy for Claude Haiku 4.5 via chatgpt.org.

Supports **tool/function calling**, auto-continue for the 1K token limit, rotating proxy, and SSE keep-alive.

## Features

- **Tool/Function Calling**: Full OpenAI-compatible tool calling support. Converts `tools` definitions to system prompts, parses Claude's output for `<tool_call_>` blocks, and returns properly formatted `tool_calls` responses.
- **Auto-Continue**: When the upstream 1K token limit is hit, automatically continues the response with "Continue" messages.
- **SSE Keep-Alive**: Sends keep-alive comments during continuation gaps to prevent socket timeouts.
- **Rotating Proxy**: Supports unstable rotating proxies with automatic retries on connection failures.
- **Message Normalization**: Handles Orchids.app's content array format and converts it to plain text.

## Environment Variables

| Variable | Description | Default |
|---|---|---|
| `PROXY_URL` | Rotating proxy URL (e.g. `http://user:pass@proxy.op.wtf:32424`) | `""` (direct) |

Set these in HF Spaces > Settings > Variables and Secrets.

## Usage

### Chat Completions (non-streaming)

```bash
curl https://YOUR_SPACE.hf.space/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-4-5",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

### Chat Completions (streaming)

```bash
curl https://YOUR_SPACE.hf.space/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-4-5",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'
```

### With Tool Calling

```bash
curl https://YOUR_SPACE.hf.space/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-4-5",
    "messages": [{"role": "user", "content": "Create a file called hello.txt with hello world"}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "Write",
        "description": "Write content to a file",
        "parameters": {
          "type": "object",
          "properties": {
            "file_path": {"type": "string", "description": "Path to the file"},
            "content": {"type": "string", "description": "Content to write"}
          },
          "required": ["file_path", "content"]
        }
      }
    }]
  }'
```

### With OpenAI Python SDK

```python
from openai import OpenAI

client = OpenAI(
    base_url="https://YOUR_SPACE.hf.space/v1",
    api_key="not-needed",
)

response = client.chat.completions.create(
    model="anthropic/claude-haiku-4-5",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(response.choices[0].message.content)
```

### List Models

```bash
curl https://YOUR_SPACE.hf.space/v1/models
```

## Endpoints

| Endpoint | Description |
|---|---|
| `POST /v1/chat/completions` | OpenAI-compatible chat completions (with tool calling) |
| `POST /chat/completions` | Same, without /v1 prefix |
| `GET /v1/models` | List available models |
| `GET /health` | Health check |
| `GET /debug/session` | Session debug info |
| `GET /debug/refresh` | Force session refresh |