LLM Model Configuration By CLI UI

Carry Code supports 17+ AI model providers. You can add, switch, or manage different models at any time.

---

Supported Providers

Carry Code supports these AI model providers out of the box:

Provider	Example Models	Protocol	Features
OpenAI	GPT-4o, GPT-5.2	OpenAI	Comprehensive capabilities
Anthropic	Claude Opus 4.5, Claude Sonnet	Anthropic	Strong programming
Google	Gemini 3 Pro, Gemini 3 Flash	Gemini	Strong multimodal
DeepSeek	DeepSeek R1, DeepSeek V3	OpenAI Compatible	High cost-performance
Moonshot / Kimi	Kimi K2	OpenAI Compatible	Strong Chinese
Zhipu AI	GLM-4.7	OpenAI Compatible	Domestic service
MiniMax	MiniMax M2.1	Anthropic	Multimodal
Alibaba Cloud	Tongyi Qianwen Qwen3 Max	OpenAI Compatible	Stable domestic
xAI	Grok 4	OpenAI Compatible	Elon Musk’s AI
SiliconFlow	DeepSeek V3.2	OpenAI Compatible	API aggregation
Ollama	Any local model	Ollama	Free local
vLLM	Any local model	vLLM	Local deployment
OpenAI Compatible	Any compatible endpoint	OpenAI Compatible	Self-hosted

---

Adding a New Model

Method 1: Interactive Add (Recommended)

1. Type /model and press Space
2. Select add
3. Follow the prompts to enter:
   • Provider type (e.g., openai, deepseek, anthropic)
   • Model name (e.g., gpt-4o, deepseek-chat)
   • API Key (your key)
   • API URL (optional, for custom endpoints)

Method 2: Manual Config Edit

The config file is located at ~/.carry/carrycode.json:

{
  "providers": [
    {
      "providerId": "deepseek",
      "modelName": "deepseek-chat",
      "apiKey": "your-api-key-here",
      "baseUrl": "https://api.deepseek.com"
    }
  ]
}

---

Switching Models

Interactive Selection (Recommended)

1. Type /model and press Space
2. Select the model you want to use from the list
3. Press Enter to confirm

---

Editing Existing Models

Interactive Edit

1. Type /model edit
2. Select the model to edit from the list
3. Modify the configuration in the editor

Direct Edit

/model edit deepseek:deepseek-chat

Modify Model Parameters

Edit the config file at ~/.carry/carrycode.json:

{
  "providers": [
    {
      "providerId": "openai",
      "modelName": "gpt-4o",
      "apiKey": "your-key",
      "temperature": 0.7,
      "maxTokens": 4000
    }
  ]
}

---

Deleting Models

1. Type /model edit
2. Select the model to edit
3. In the editor, delete the model or clear its API key

Alternatively, edit the config file directly at ~/.carry/carrycode.json and remove the provider entry.

---

Setting Default Model

The default model is automatically set to the first configured model. To change it:

1. Edit the config file at ~/.carry/carrycode.json
2. Reorder the providers array - the first model will be the default

Or use the environment variable:

export CARRYCODE_MODEL="deepseek:deepseek-chat"

See Environment Variable Model Configuration for more details.

---

Model Parameter Tuning

You can adjust model parameters in the config file:

{
  "providers": [
    {
      "providerId": "openai",
      "modelName": "gpt-4o",
      "apiKey": "your-key",
      "temperature": 0.7,
      "maxTokens": 4000,
      "topP": 0.9
    }
  ]
}

Parameter Description

Parameter	Description	Recommended
temperature	Randomness control (0-2)	0.7
maxTokens	Max output tokens	4000
topP	Nucleus sampling probability	0.9

---

Free Model Recommendations

Ollama (Local Free)

If you want to use AI for free, you can install Ollama:

1. Install Ollama: https://ollama.ai
2. Download models: ollama pull llama2
3. Add Ollama config in Carry Code

Free API Credits

Provider	Free Credits
DeepSeek	¥10 on sign-up
Alibaba Cloud	¥60 on sign-up
SiliconFlow	¥14 on sign-up

---

FAQ

Q: Which model should I choose?

• Programming tasks: Claude Sonnet or GPT-4o recommended
• Chinese tasks: DeepSeek or Kimi recommended
• Free usage: Ollama local deployment recommended

Q: Will switching models lose context?

No. The conversation context is preserved when switching models.

Q: How to view the current model?

The current model name is displayed at the top of the interface.

---

Environment Variable Configuration

Carry Code also supports configuring models via environment variables, which is useful for:

• CI/CD pipelines
• Container deployments
• Quick setup without editing config files

For detailed information, see Environment Variable Model Configuration.

---

Next Steps

• MCP Servers - Extend AI capabilities
• Skills - Use preset skills
• Themes - Customize the interface