# oc-probe

**Repository Path**: gabrielslls/oc-probe

## Basic Information

- **Project Name**: oc-probe
- **Description**: A zero-dependency CLI tool to probe LLM models from opencode.json for accessibility and performance.
- **Primary Language**: Python
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-03-24
- **Last Updated**: 2026-04-05

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# oc-probe 🚀

**[opencode](https://opencode.ai/) 模型探针** — 一键测试你配置的所有 LLM 模型是否可用、响应速度如何。

A zero-dependency CLI tool that probes all LLM models configured in your [opencode](https://opencode.ai/) `opencode.json`, checking availability and measuring response time.

---

## ✨ Features

- **零依赖** — 仅需 Python 3.6+，无需 `pip install` | Zero dependencies, Python 3.6+ only
- **自动发现** — 读取 opencode 配置，自动提取所有 provider 和模型 | Auto-discovers providers & models from config
- **环境变量解析** — 支持 `{env:VAR_NAME}` 格式的 API Key | Resolves `{env:VAR_NAME}` style env vars
- **Thinking 模式** — 可测试模型的 thinking/reasoning 能力 | Test thinking/reasoning mode
- **速度排行** — 自动按响应时间排序 | Auto-ranked by response time
- **彩色输出** — 清晰的终端彩色展示 | Colorful terminal output

## 📦 Installation

```bash
# Just download the script — no installation needed
curl -O https://raw.githubusercontent.com/gabrielslls/oc-probe/main/oc-probe.py
chmod +x oc-probe.py
```

Or clone the repo:

```bash
git clone https://github.com/gabrielslls/oc-probe.git
cd oc-probe
```

## 🚀 Usage

```bash
# Default mode — test all models without thinking parameter
python3 oc-probe.py

# Enable thinking mode
python3 oc-probe.py think

# Disable thinking mode
python3 oc-probe.py nothink

# Show help
python3 oc-probe.py help
```

## ⚙️ Configuration

oc-probe reads your opencode configuration file from:

```
~/.config/opencode/opencode.json
```

You can override this path via the `OC_PROBE_CONFIG` environment variable:

```bash
OC_PROBE_CONFIG=/path/to/my/opencode.json python3 oc-probe.py
```

### Config Format

The tool reads the `provider` section of `opencode.json`. Each provider needs:

- `options.apiKey` — API key (supports `{env:VAR_NAME}` syntax)
- `options.baseURL` — API base URL (OpenAI-compatible `/chat/completions` endpoint)
- `models` — Map of model identifiers to model info

See [opencode.example.json](opencode.example.json) for a complete example.

```jsonc
{
  "provider": {
    "my-provider": {
      "options": {
        "apiKey": "{env:MY_API_KEY}",
        "baseURL": "https://api.example.com/v1"
      },
      "models": {
        "gpt-4o": { "name": "GPT-4o" },
        "claude-3-opus": { "name": "Claude 3 Opus" }
      }
    }
  }
}
```

## 📊 Output Example

```
======================================================================
  🚀 多平台模型测速工具
======================================================================

📋 加载配置...
   ✅ [provider-a] 3 个模型
   ✅ [provider-b] 2 个模型
   ✅ 总计: 5 个模型

⚙️  测试配置:
   • 超时: 180秒
   • 模型间隔: 10秒
   • 题目: 简单自我介绍

======================================================================
  🌐 测试 [provider-a]
======================================================================

[1/3] GPT-4o
----------------------------------------------------------------------
   🧠 Thinking模式: 使用服务端默认
   ⏳ 发送请求...
   ✅ 成功 | ⏱️  1.23s | 📝 15→85 (100) | 无Thinking
   📝 输出: 我是一个AI助手...

...

======================================================================
  📊 测试总结
======================================================================

⏱️  总耗时: 45.2秒
📈 结果: 5/5 成功

🏆 速度排行:
   1. [provider-a] GPT-4o: 1.23s
   2. [provider-b] Claude 3 Opus: 2.45s
   ...

======================================================================
  🎉 所有模型测试成功！
======================================================================
```

## ❓ FAQ

**Q: 支持哪些 API 格式？**
A: 任何兼容 OpenAI `/chat/completions` 接口的 API 均可，包括 OpenAI、Anthropic（通过兼容层）、各种国产大模型 API 等。

**Q: 为什么模型之间要等 10 秒？**
A: 避免触发 API 速率限制（Rate Limit）。

**Q: 支持流式（streaming）测试吗？**
A: 目前不支持，使用非流式请求以获得准确的总响应时间。

## 📄 License

[MIT](LICENSE)