Model Command `/model`¶

Overview¶

The /model slash command lets you switch the active LLM provider and model at runtime without editing configuration files. It provides an interactive TUI picker and direct shortcut syntax for quick switching.

All selections are persisted to the project-level .datus/config.yml so they survive session restarts and are scoped to the current working directory.

Basic Usage¶

Interactive Picker¶

Type /model (or /models) with no arguments to open the full TUI picker:

/model

The picker has three tabs, navigable with Tab / Shift+Tab:

Tab	Description
Providers	General-purpose LLM providers (OpenAI, DeepSeek, Claude, etc.)
Plans	Coding-oriented plan providers (Alibaba Coding, GLM Coding, etc.)
Custom	Self-hosted or private-deployment models from `agent.models`

Use Up/Down arrows to navigate, Enter to select, and Esc / q to cancel.

Direct Shortcuts¶

Switch without opening the picker:

# Switch to a specific provider + model
/model openai/gpt-4.1
/model deepseek/deepseek-chat
/model claude/claude-sonnet-4-5

# Switch to a custom model entry
/model custom:my-internal-model

# Open the picker pre-drilled into a provider
/model openai

Provider Authentication¶

API Key Providers¶

For providers that require an API key (OpenAI, DeepSeek, Claude API, Kimi, Qwen, Gemini, etc.):

Select the provider in the picker
If no credentials are configured, a credential form appears
Enter your API key (masked input)
The key is persisted to agent.yml under agent.providers.<name>

API keys can also be pre-configured via environment variables. Each provider has a default env var name defined in conf/providers.yml (e.g., OPENAI_API_KEY, DEEPSEEK_API_KEY).

Claude Subscription¶

For Claude Pro/Max subscription users:

Select claude_subscription in the picker
The system auto-detects locally stored subscription tokens
If auto-detection fails, you are prompted to enter the token manually (sk-ant-oat01-...)

Codex (ChatGPT Plus/Pro) OAuth¶

For ChatGPT Plus/Pro users via Codex:

Select codex in the picker
A browser window opens for OAuth authentication
After completing the browser flow, connectivity is verified automatically
On success, you return to the picker with Codex models available

Custom Models¶

Adding a Custom Model¶

Navigate to the Custom tab in the picker
Select + Add custom model
Fill in the form fields:
Name — Unique identifier (used as agent.models.<name> key)
Type — Interface type (openai, claude, deepseek, kimi, gemini, etc.)
Base URL — API endpoint URL
API Key — Authentication credential (masked)
Model — Model name / SKU
Press Enter to save

Custom models are stored in agent.yml under agent.models and appear in the Custom tab.

Deleting a Custom Model¶

Navigate to the Custom tab
Select the model to delete
Press d twice (two-press confirmation) to delete

Configuration Files¶

Project Override (`.datus/config.yml`)¶

When you select a model via /model, the choice is saved to .datus/config.yml in the current working directory:

# Provider-level selection
target:
  provider: openai
  model: gpt-4.1

# Or, custom model selection
target:
  custom: my-internal-model

This file supports three keys only:

Key	Description
`target`	Active LLM selection (provider+model or custom name)
`default_datasource`	Default datasource for this project
`project_name`	Override for session/data shard name

Provider Credentials (`agent.yml`)¶

Provider credentials live in agent.yml under agent.providers:

agent:
  providers:
    openai:
      api_key: ${OPENAI_API_KEY}
    deepseek:
      api_key: ${DEEPSEEK_API_KEY}
    claude_subscription:
      auth_type: subscription
    codex:
      auth_type: oauth

Provider Catalog (`conf/providers.yml`)¶

The provider catalog defines available providers, their models, base URLs, and model specs. It is read-only and shipped with Datus. Adding a new provider to this file makes it available in the /model picker without code changes.

How It Works¶

Model resolution follows this precedence:

Project override — .datus/config.yml in the current working directory
Base config — agent.target in agent.yml

The /model command updates the project override. It does not rebuild the agent — the switch takes effect on the next LLM call within the same session.

Model Command /model¶