badrisnarayanan/antigravity-claude-proxy v1.2.4

on GitHub

What's New

Smarter Rate Limiting

• Model-specific rate limits: Rate limits now apply per-model instead of globally. If one model is exhausted, you can still use other models on the same account.
• Better status visibility: /account-limits?format=table now shows (x/y) limited indicating exactly how many models are rate-limited per account.

Improved Reliability

• Network error handling: Transient network errors (timeouts, connection resets) now trigger automatic retry with the next account instead of failing immediately.
• Thinking budget validation: Automatically adjusts max_tokens when it's less than thinking_budget to prevent API errors.

Enhanced Monitoring

• Detailed health endpoint: /health now returns per-account model quotas with remaining percentages and reset times.
• Clearer error messages: Rate limit errors now include the affected model name for easier debugging.

CLI Improvements

• Exit option: Added (e)xit option to the accounts CLI menu.
• Single account flow: Account addition now handles one account at a time to avoid OAuth state conflicts.
• No more hanging: CLI properly exits after completion.

Other

• Startup banner now shows active modes (e.g., debug mode)
• Only Claude and Gemini models are returned from /v1/models

Full Changelog: v1.2.3...v1.2.4