fix(cli): auto-propagate LLM_API_KEY to provider-specific env vars via config-driven detection

[wutongyuonce]

No description provided.

[KylinMountain]

Thanks @wutongyuonce .

Since you're expanding the LLM_API_KEY propagation list, would you also consider adding a few other popular providers in the same vein?

• Moonshot (Kimi)
• Zhipu / GLM
• Qwen (Alibaba)

All three are supported by LiteLLM and widely used; the env var names follow the same {PROVIDER}_API_KEY convention.

[wutongyuonce]

Enhance LLM Provider Support and Streamline API Key Configuration Summary

This PR significantly expands OpenKB's support for major LLM providers and optimizes the logic for environment variable loading and distribution. By introducing a parallel "Unified Key" and "Provider-Specific Key" model, it greatly reduces configuration overhead for users while enhancing system robustness in complex multi-model scenarios.
Key Changes

1、Expanded Model Provider Support

• Expanded the _KNOWN_PROVIDER_KEYS list in cli.py .
• Added native support for several popular domestic and international providers:
  • Moonshot (Kimi) : MOONSHOT_API_KEY , KIMI_API_KEY
  • Zhipu (GLM) : ZHIPUAI_API_KEY , GLM_API_KEY
  • Alibaba (Qwen) : DASHSCOPE_API_KEY , QWEN_API_KEY
  • Also updated DeepSeek, Groq, Together AI, and others.

2、Optimized API Key Propagation Logic

• Improved the _setup_llm_key function with "intelligent identification" capabilities.
• Auto-Distribution : Setting a single LLM_API_KEY now automatically propagates it to all known provider variables, enabling "one-key configuration, full-range support."
• Priority Respect : Ensures that specific PROVIDER_API_KEY s take precedence over the unified LLM_API_KEY , protecting fine-grained user configurations from being overwritten.

3、Refactored .env.example Template

• Redesigned the structure of .env.example into clear sections:
  • Option 1 (Unified) : The recommended simple setup.
  • Option 2 (Specific) : Advanced management for multiple providers.
• Added a Configuration Architecture Note at the top of the file to clarify the relationship between .env (Authorization) and config.yaml (Model Routing), lowering the barrier for new users.

[KylinMountain]

why not deepseek-v4-pro?

[KylinMountain]

Please remove Together, GROQ and Cohere. I believe they are distributors, not model providers. Is that correct?

KIMI or MOONSHOT? ZHIPU or GLM, DASHSCOPE or QWEN ? Please help to confirm this confusion, thanks.

[KylinMountain]

I don’t think we need to modify this env example file. Adding DeepSeek should be sufficient.

[KylinMountain]

should remove this file, we have it in repo.

[wutongyuonce]

I have made the modifications as you said.

• remove Together, GROQ and Cohere
• remove "KIMI_API_KEY", "GLM_API_KEY" and "QWEN_API_KEY"
• change "deepseek/deepseek-v4-0417" to "deepseek/deepseek-v4-pro"