Chinese Model provider

[wing (Malowking)]

feat(cli): add Qwen and DeepSeek model support via ChatOpenAI

Adds Qwen and DeepSeek model support using langchain_community.chat_models.ChatOpenAI base implementation. Extends model providers while maintaining existing functionality.

Fixes #ISSUE_NUMBER

Changes:

• Added Qwen and DeepSeek to model providers
• Implemented ChatOpenAI-based model creation
• Added API key validation
• Preserved backward compatibility

Verification:

• make format and make lint passed
• Tested model instantiation
• Confirmed error handling works

No breaking changes.

[Eugene Yurtsev (eyurtsev)]

We want to support this by allowing folks to configure something like

--completions-api --base-url --api-key etc

rather than using a lot of specific env variables per provider -- since a lot of providers rely on the exact same conventions

Using the completions API will be basically unlock all models at once that rely on completions API compatibility

[wing (Malowking)]

我们希望通过允许用户配置类似的功能来支持这一点

--完成 --基础-URL --api-key等

而不是每个供应商使用大量特定的环境变量——因为很多供应商依赖完全相同的约定

使用完备API基本上可以一次性解锁所有依赖完备API兼容性的模型

We want to support this by allowing folks to configure something like

--completions-api --base-url --api-key etc

rather than using a lot of specific env variables per provider -- since a lot of providers rely on the exact same conventions

Using the completions API will be basically unlock all models at once that rely on completions API compatibility

Even if we add this kind of configuration when starting the service, Google and Anthropic still have their own API invocation patterns. This means we would still need to add a lot of conditional logic to distinguish between the Completions API and other API styles.

If we introduce a .env file and switch to configuring the model through environment variables instead, the create_model method would only need to read parameters from the file, without requiring changes in many other places.

I’m not sure whether this approach is feasible, so I would greatly appreciate your advice. Thank you.

[Mason Daugherty (mdrxy)]

Closing in favor of #1127