Using FishXCode with CodeBuddy

First time using FishXCode? Complete Account Setup first, then get your API Key from the FishXCode console.

1. Install and Start CodeBuddy Code

This tutorial configures Tencent Cloud CodeBuddy Code, the command-line tool. Make sure Node.js 18.0 or later is installed locally, then install it:

npm install -g @tencent-ai/codebuddy-code

Check the version:

codebuddy --version

Start CodeBuddy Code:

codebuddy

On first launch, follow the prompts to complete login. Later, you can enter /model in the command line to check whether custom models were loaded successfully.

2. Open the CodeBuddy Configuration File

CodeBuddy Code supports custom models through a local models.json file. The configuration file is located at:

• macOS / Linux: ~/.codebuddy/models.json
• Windows: C:\Users\<用户名>\.codebuddy\models.json

Create the directory or file first if it does not exist.

3. Write the FishXCode Model Configuration

CodeBuddy Code uses the OpenAI Chat Completions compatible protocol for third-party LLM configuration. Here, vendor: "OpenAI" means protocol compatibility. It does not mean only OpenAI models can be used; Claude models are also called through the FishXCode OpenAI compatible endpoint.

Write the following content to ~/.codebuddy/models.json:

{
  "models": [
    {
      "id": "gpt-5.5",
      "name": "FishXCode GPT-5.5",
      "vendor": "OpenAI",
      "apiKey": "sk-你的FishXCode-Token",
      "url": "https://api.fishxcode.com/v1/chat/completions"
    },
    {
      "id": "claude-opus-4-7",
      "name": "FishXCode Claude Opus 4.7",
      "vendor": "OpenAI",
      "apiKey": "sk-你的FishXCode-Token",
      "url": "https://api.fishxcode.com/v1/chat/completions"
    }
  ],
  "availableModels": [
    "gpt-5.5",
    "claude-opus-4-7"
  ]
}

Important

• Replace sk-你的FishXCode-Token with the actual API Key you created in the FishXCode console.
• availableModels must include the model IDs from models[].id; otherwise, the new models may not appear in /model.
• Third-party relay services require the full Chat Completions endpoint: https://api.fishxcode.com/v1/chat/completions.

4. Keep Tencent Coding Plan Alongside FishXCode

If you also need to keep Tencent's official Coding Plan, append the FishXCode models in the same models.json file:

{
  "models": [
    {
      "id": "tc-code-latest",
      "name": "Auto",
      "vendor": "Tencent Cloud",
      "apiKey": "YOUR_TENCENT_CODING_PLAN_KEY",
      "url": "https://api.lkeap.cloud.tencent.com/coding/v3"
    },
    {
      "id": "gpt-5.5",
      "name": "FishXCode GPT-5.5",
      "vendor": "OpenAI",
      "apiKey": "sk-你的FishXCode-Token",
      "url": "https://api.fishxcode.com/v1/chat/completions"
    },
    {
      "id": "claude-opus-4-7",
      "name": "FishXCode Claude Opus 4.7",
      "vendor": "OpenAI",
      "apiKey": "sk-你的FishXCode-Token",
      "url": "https://api.fishxcode.com/v1/chat/completions"
    }
  ],
  "availableModels": [
    "tc-code-latest",
    "gpt-5.5",
    "claude-opus-4-7"
  ]
}

5. Restart and Select a Model

1. Save ~/.codebuddy/models.json
2. Fully quit any running CodeBuddy Code process
3. Run codebuddy again
4. Enter /model
5. Select gpt-5.5 or claude-opus-4-7
6. Send a test prompt, such as 解释当前项目结构

FAQ

/model does not show the models

Check availableModels first. The model ID must appear in both models[].id and availableModels.

The model appears but calls fail

Check url first. FishXCode third-party model configuration should use the full endpoint https://api.fishxcode.com/v1/chat/completions.

Authentication fails

Copy the API Key from the console again and make sure there are no extra spaces or line breaks.