Skip to content

Releases: cellwebb/gac

v3.11.0

11 Dec 21:23

Choose a tag to compare

@sogoiii contributed PR #63 to add custom system prompt management, which you can use with the subcommand uvx gac prompt set --edit or uvx gac prompt set --file <filepath>. You can also show/clear the custom system prompt with uvx gac prompt show and uvx gac prompt clear, respectively.

Thank you @sogoiii!

Full Changelog: v3.10.15...v3.11.0

v3.10.15

09 Dec 19:16

Choose a tag to compare

I just updated the Mistral default to devstral 2! Devstral 2 is free through the Mistral API right now, so go ahead and take advantage!

Additionally, I've been doing a lot of under-the-hood updates to make gac run smoother, and increasing test coverage!

Full Changelog: v3.8.2...v3.10.15

v3.8.2

05 Dec 17:26

Choose a tag to compare

This update by @mpfaffenberger allows you to skip SSL verification, which is useful if you're trying to use gac behind certain firewalls.

You can either use the --no-verify-ssl flag or make it permanent with gac config set GAC_NO_VERIFY_SSL true.

Full Changelog: v3.7.1...v3.8.2

v3.7.1

04 Dec 21:19

Choose a tag to compare

This update adds Qwen.ai as a provider. You can sign up for a free account at qwen.ai and use gac for free with one of the best open-source coding models! Authentication for Qwen.ai is handled via OAuth rather than an API key.

Full Changelog: v3.6.2...v3.7.1

v3.6.2

01 Dec 18:38

Choose a tag to compare

Interactive Mode

If you haven't tried it yet, the new --interactive/-i flag is a game-changer for complex changes. gac analyzes your staged changes and asks smart questions about context, intent, and impact - giving you much more accurate commit messages for non-trivial work.

Try it with gac -i or combine it with verbose mode: gac -iv

CleanShot 2025-12-01 at 10 06 28

Under the hood

  • Replaced click.prompt with prompt_toolkit.prompt for rich input handling
  • Implemented adaptive question generation based on change complexity (file count + line modifications)
  • Enhanced exception handling for better error reporting
  • Removed the unused Halo dependency to streamline the package
  • Added comprehensive test coverage for all interactive functionality

Full Changelog: v3.5.0...v3.6.2

v3.5.0

30 Nov 21:14

Choose a tag to compare

This release adds Azure OpenAI, Kimi Coding, and Moonshot AI providers! 🎉 I've also restructured the CLI architecture to make adding new providers much cleaner - the model configuration is now in its own dedicated module, and the
init flow has been simplified to orchestrate model + language setup much more elegantly.

The provider ecosystem is growing fast - we now support 26+ providers including the new additions:

  • Azure OpenAI - Full enterprise Azure support with proper endpoint construction and API version handling
  • Kimi Coding - OpenAI-compatible API for Kimi's coding-optimized models
  • Moonshot AI - Direct Moonshot AI integration

Under the hood, I've:

  • Extracted model configuration to a dedicated model_cli.py module
  • Simplified the init workflow from 600+ lines to ~100 lines by leveraging the new model and language modules
  • Enhanced test infrastructure with better isolation and comprehensive test suites for new providers

Full Changelog: v3.3.0...v3.5.0

v3.3.0

15 Nov 23:43

Choose a tag to compare

This update adds a --message-only flag, which returns the commit message only (no commit is performed). This can be useful for incorporating into other tools.

I also fixed a bug when using the --group flag with "git mv" changes

Full Changelog: v3.2.0...v3.3.0

v3.2.0

14 Nov 23:08

Choose a tag to compare

This release adds support for Replicate!

Full Changelog: v3.1.0...v3.2.0

v3.1.0

12 Nov 19:40

Choose a tag to compare

This update just does stuff behind the scenes that makes adding new providers a snap.

If there are any providers you think I should add, drop them below!

Full Changelog: v3.0.0...v3.1.0

v3.0.0

10 Nov 19:56

Choose a tag to compare

This release tightens up configuration handling and polishes the onboarding flows.

  • Breaking change: gac no longer reads plain .env files. Move any project-level .env variables to project-level .gac.env (or copy the contents into $HOME/.gac.env) before upgrading so your overrides keep working.
  • gac config show now prints both the user-level and project-level .gac.env when they exist, and explicitly reminds you that the project file wins for any duplicated variables.
  • The init/model wizard reflects the real provider names (MiniMax.io, Synthetic.new), normalizes them to the right internal keys, and stores the correct GAC_MODEL/API key combinations. MiniMax.io also warns if you skip the API
    key.
  • Claude Code setup no longer forces you through OAuth again if you cancel the follow-up action prompt, and RTL language selection now includes a one-time confirmation that’s remembered via GAC_RTL_CONFIRMED.

Full Changelog: v2.7.3...v3.0.0