DeepSeek

[official website]

DeepSeek models can be added to Tresor via API key authentication.

Web UI

Web UI is more intuitive, but require more manual operations.

In "Downstreams" tab

• If DeepSeek does not appear in the list of providers, click + Add Provider.
• Enter DeepSeek or anything you like as the provider name.
• Under "API Formats", select OpenAI, OpenAI Responses, and Anthropic.
• Under "Authentication", select API key, and enter the API key you obtained from DeepSeek official website (usually starts with sk-).
• Under "API Host", enter https://api.deepseek.com.
• Click Advanced. Under "Per-format URLs", enter https://api.deepseek.com/anthropic.
• Finally, under "Models", click ⟳ Fetch Models, and add the models you need.

config.yaml

Editing yaml is less intuitive, but require less manual operations.

Replace sk-YOUR-DEEPSEEK-API-KEY with the API key you obtained from DeepSeek official website.

~/.config/tresor/config.yaml

... # Omitted showing other settings here
downstreams:
  - id: deepseek
    name: DeepSeek
    base_url: https://api.deepseek.com
    api_formats:
        - openai
        - openai_responses
        - anthropic
    output_model_ids:
        - deepseek-flash
        - deepseek-v4-pro
    is_enabled: true
    auth:
        type: api_key
        api_key: sk-YOUR-DEEPSEEK-API-KEY
    format_urls:
        anthropic: https://api.deepseek.com/anthropic

Restart Tresor to apply the changes in yaml.

Done! Clients accessing Tresor can use the models from DeepSeek now!

📚 Next step

• Configure your LLM Apps