thomwebb/gac

Improvements

• 🚀 Fireworks provider now automatically caps token count at 4,096 for non‑streaming requests, so your calls won’t fail
• 🧪 Added extensive automated tests for many parts of the app, leading to more reliable performance
• 🧹 Cleaned up code style and added clearer type hints, helping keep the app stable and ready for future features

What's New

• 🤖 Add AI‑agent server that lets tools inspect your repository and create commits automatically
• 🛠️ New “gac_status” and “gac_commit” commands let agents show repo status and generate commits
• ⚡ “gac serve” subcommand starts the server directly from the main CLI

Improvements

• ✂️ Text is now trimmed at word boundaries, so messages stay tidy and readable
• 📈 Model list updated with the latest Claude Code, MiniMax, and Moonshot versions for better results
• 📚 Documentation now has consistent spacing and clearer notes across all languages

Bug Fixes

• 🔗 Fixed broken links in the MCP documentation for all languages
• 🧹 Removed old wrapper script and unused files, making the tool lighter
• 🛡️ Improved status output formatting and limited diff size to avoid overwhelming information

What's New

• 🚀 New --50-72 flag and GAC_USE_50_72_RULE setting to keep commit messages short and tidy (subject ≤50 chars, lines ≤72 chars).

Improvements

• ⚡ Updated model selections to the latest GLM‑5 versions for smarter suggestions.
• 🔢 Generation spinner now shows an estimated token count so you know how much data is being sent.

Improvements

• 🧹 Simplified token counting by removing the old tiktoken setting, so you don't need to configure anything extra.
• 🇰🇷 Korean docs now have corrected terms and clearer explanations, making it easier to follow.
• 📂 The -g grouping flag is now listed in the common options tables, so you can quickly see how to group changes.
• 🤖 Model names for Cerebras and Z.AI were updated, ensuring the latest models work smoothly.

Improvements

• ⚡ Token counting is now faster and works consistently across all AI providers, with old settings removed for smoother use.
• 🚀 Updated AI models to the latest versions, giving you more accurate and powerful results.
• 🔧 Improved the release process to avoid accidental updates, keeping your experience stable.