yangfei4913438

yangfei4913438 / codex-deepseek

Public

Protocol translation proxy — converts OpenAI Responses API to Chat Completions API, enabling Codex to use DeepSeek or any OpenAI-compatible model via cc-switch.

cc-switch codex deepseek
14
1
85% credibility
Found May 24, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This is an open-source translation tool that acts as a bridge between Codex (an AI coding assistant) and DeepSeek or any other AI model that uses a different communication format. It runs locally on your computer, converting requests and responses in real-time so you can use your preferred AI model with Codex. The project is well-documented, uses only standard Python tools, and works alongside cc-switch to manage the connection.

How It Works

1
💡 You want to use a different AI model with Codex

You've been using Codex for coding help, but you want to try DeepSeek or another AI model instead of the default.

2
🔌 You download the translation tool

You grab this small program that acts like a bridge between Codex and your preferred AI model.

3
🔑 You connect your AI service account

You enter your API credentials for DeepSeek or another compatible AI provider so the tool can talk to it.

4
🚀 You start the bridge

With one click, the translation service starts running on your computer, ready to convert messages back and forth.

5
You set up how Codex finds your AI model
📝
Automatic setup

cc-switch handles the configuration automatically

✋
Manual setup

You fill in the connection details yourself

6
🔄 Everything translates automatically

Every time you ask Codex a question, the bridge quietly converts it into your AI model's language and translates the answer back.

🎉 Codex now works with your favorite AI model

You get coding help from DeepSeek or any compatible model, as if it were built into Codex from the start.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is codex-deepseek?

codex-deepseek is a Python proxy that bridges the gap between Codex and DeepSeek (or any OpenAI-compatible model). Codex speaks the OpenAI Responses API, while most model providers speak the Chat Completions API -- this proxy translates between the two in real time. It runs locally, listens on port 11435, and requires no external dependencies beyond Python's standard library.

The setup involves starting the proxy, then configuring cc-switch to route Codex requests through it. You point Codex at the local endpoint, and the proxy handles converting the request format, forwarding it to your chosen model provider, and translating the response back.

Why is it gaining traction?

Developers want to run Codex with cheaper or locally-hosted models, but the API mismatch has been a blocker. This proxy removes that friction entirely. The thinking mode support is a major draw -- it preserves DeepSeek's reasoning chains across multi-turn conversations with tool calls, which isn't handled correctly by most translation layers.

The zero-dependency approach also appeals to folks who want something lightweight and auditable. You can read the entire codebase in an afternoon and understand exactly what it's doing to your requests.

Who should use this?

Developers running Codex who want to swap in DeepSeek V4 Pro, DeepSeek Coder, or similar models without leaving the Codex ecosystem. Self-hosted AI enthusiasts who prefer DeepSeek pricing over OpenAI's will find this essential. It's also useful for teams experimenting with different model providers while keeping Codex as their primary coding interface.

Verdict

This fills a real gap for the DeepSeek + Codex combination, and the code is clean and well-documented. At 14 stars it's early-stage -- treat it as a solid v1 rather than battle-tested infrastructure. The 0.850% credibility score reflects a new but technically sound project with good documentation and test coverage. Worth trying if you fit the use case, but keep expectations calibrated for something actively being developed.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.