GPT-5.1 Series, GPT-5.1-Codex-Max

Pull requests merged

• Fix tool call argument serialization issue by @phawrylak in #39
• Document docker env switches for reasoning and web search by @alexx-ftw in #64
• Add CONTRIBUTING guide by @alexx-ftw in #48
• fix: Add graceful error handling for ChunkedEncodingError during streaming by @alexx-ftw in #57
• feat: improve compatiblity with certain apps by @Magniquick in #72
• GPT-5.1 models "minimal" removed, add gpt-5.1-codex-max by @RayBytes in #80

Full Changelog: v1.3...v1.35

GPT-5-Codex Release!

This release adds the new GPT-5-Codex model! In my limited testing, this performs significantly faster then regular GPT-5 in most thinking modes, especially in places of less complexity. Enjoy!

Since the last release, we have also added a few other changes which are notable.

Changelog

Native OpenAI Web search

This works the same as the regular web search in the ChatGPT app. You can opt-in to enable web search by appending the parameter --enable-web-search to your serve command. Currently, this is opt-in, but if you believe it should be different, leave feedback with a github issue!
If you programmatically use the API, you can check an example for how to send a request with web search enabled in the README.

Reasoning effort as models

If you ever have had issues changing the reasoning effort in your chat apps for a model, then this feature should help you out. By appending the parameter --expose-reasoning-models to the end of a serve command, models with reasoning effort (gpt-5, gpt-5-codex) will have each of their reasoning efforts be available as different models. This means that you would have slugs like gpt-5-minimal, gpt-5-low, gpt-5-medium, gpt-5-high, etc, which means you can easily switch between them in a model switcher.

Note

Currently, the GUI is a bit left behind with these updates in terms of configuration. Would also request help & PRs regarding the GUI, and on a proper build pipeline for the application, especially for windows!

If you wish to upgrade your brew installation (which is the intended installation method of this project), you should do brew upgrade chatmock.

Thank you very much to these lovely contributors for their help in these updates:

• Native openai web_search tool by @alexx-ftw in #34
• Add GPT-5-Codex support by @mLupine in #37

Full Changelog: v1.2...v1.3

QoL Improvements

Quite a few changes and features have been added over the past few days, and it's now stable enough that I can call it a full release.

Changelog

Token caching
The server smartly creates IDs to be used for caching tokens now, and sends it to OpenAI, which should increase usage significantly.

Docker:
There's now an official docker container which can be used to allow this project to run across your network, and on multiple machines.

Homebrew
Can now install the CLI python flask server as a package through homebrew, which should make it easier to use.

Token counting
Support for token counting has been added, which should fix the fact that the amount of tokens wasn't showing up in applications like RooCode, and the other issue where the api request never seemed to stop.

Other smaller functionally (but important) fixes include:

• added the "minimal" reasoning option (might be better for code, and would get you more usage)
• refactor of the entire project to hopefully make it more maintainable onwards

Next release will focus mainly on a proper CI/CD pipeline to build and ensure the ChatMock GUI applications are always available and built, for all platforms
Also will work on the rest of the GitHub issues.

upd: cool people at codex also up'd the plus plan's usage limits, should help too

Pull requests from some lovely people

• Update requirements.txt by @ishanray in #9
• fix: stream(legacy): include delta.reasoning alongside reasoning_summary by @FammasMaz in #16

Full Changelog: 1.15...v1.2

System prompt support!

I've finally come to adding system prompt support! This allows this project to very easily work with many of your favourite chat apps more seamlessly, and especially with coding applications. More importantly, the Codex CLI system prompt, which cannot be removed, doesn't seem to cause any visible degradation in my small testing with code quality or failing tool calls, etc.

Currently, only a Mac build will be available to ensure everything works as required, and soon I will release a full release including this along with a few other QoL features. You are still free to run the flask server yourself on windows or linux while waiting, as that contains the new changes.

Ollama support

ChatMock now officially supports being also used as a Ollama-compatible server!
This allows it to be used in apps which restrict their usage to an ollama api only, for e.g raycast.
You only need to put the current api url (127.0.0.1:8000 by default) where you would put the ollama host, and GPT-5 should automatically show up as a model.
Tool call support and the ability to change the reasoning effort natively has also been added for ollama.

If you encounter any problems please make a issue.

Also a thank you to this work by @XInTheDark which allowed me to figure this out quickly :)

GUI Release

Initial release

This is a complete GUI version of this project, where you can modify anything as needed in the GUI, and serve the local server, then keep it stashed in the tray until need be. If you find any bugs feel free to report them in issues.

Currently only a MacOS version will be released, a windows version will soon follow.