github lbjlaq/Antigravity-Manager v5.0.4-beta.20
Antigravity Tools v5.0.4-beta.20

pre-release55 minutes ago
  • [Token Cache Hit Rate Overflow Defense & Financial Inversion Auto-Heal] (PR #4015, Thanks to @cubelikeplayDaniel):
    • Cross-Protocol Prompt Cache Accounting Normalization: Resolved an issue under the Anthropic Claude protocol where uncached incremental prompt tokens and cached tokens are reported separately (input_tokens vs cache_read_input_tokens). When computing hit rates using raw incremental tokens as total input, the rate exceeded 100% (e.g. 108.8%). Implemented non-destructive dynamic healing across TokenStats, AppleBentoView, and CyberCockpitView to restore effective full prompt context (total_input + cached), strictly clamping hit rate to [0, 100%]. (PR #4015, Thanks to @cubelikeplayDaniel)
    • Eliminate Cost Zeroing & Savings Inversion: Fixed Math.max(0, total_input - cached) in cost computation erroneously clamping to 0 when incremental input was less than cached tokens, which severely deflated total expenditure and produced inverted savings exceeding total costs. Added canonical saturating additions on the backend token_stats.rs insertion path to prevent future database pollution. (PR #4015, Thanks to @cubelikeplayDaniel)
  • [Gemini Inbound Dynamic Thinking Budget Alignment] (Fixes #4012, PR #4014, Thanks to @vecyang1):
    • Restore Dynamic Thinking for thinkingBudget: -1: Fixed an issue where client requests passing thinkingBudget: -1 (Gemini SDK standard for adaptive/dynamic thinking) via the Gemini protocol were erroneously mapped to 0 in wrap_request_v2, causing InboundThinkingPipeline to disable thinking and strip thought chains. Accurately identifies -1 as enabled dynamic thinking while strictly respecting explicit thinkingBudget: 0 and includeThoughts: false opt-outs. (Fixes #4012, PR #4014, Thanks to @vecyang1)

  • [Token 缓存命中率防溢出与核算成本倒挂自愈] (PR #4015, Thanks to @cubelikeplayDaniel):
    • 跨协议 Prompt 缓存记账口径归一化: 修复在 Anthropic Claude 协议下,上游响应将未命中缓存增量与缓存读取分离(input_tokens 与 cache_read_input_tokens),导致前端将增量输入直接作为总输入计算时,缓存命中率异常溢出至 100% 以上(如 108.8%)的问题。在 TokenStats、AppleBentoView 与 CyberCockpitView 中引入非破坏性自愈机制,自动恢复有效完整 Prompt 上下文(total_input + cached),将命中率严格收敛于 [0, 100%]。(PR #4015, Thanks to @cubelikeplayDaniel)
    • 根除成本截断归零与节省金额倒挂: 修复成本计算中 Math.max(0, total_input - cached) 因增量输入小于缓存 token 数量而被强行截断归零、导致总支出严重低估且估算节省金额反超总支出数倍的财务核算倒挂缺陷;并在后端 token_stats.rs 写入层引入饱和加法规范化,杜绝未来增量数据污染 SQLite 数据库。(PR #4015, Thanks to @cubelikeplayDaniel)
  • [Gemini 入站动态思考开启判定与协议对齐] (Fixes #4012, PR #4014, Thanks to @vecyang1):
    • 修复 thinkingBudget: -1 被误转为 0 导致思考静默关闭: 解决客户端通过 Gemini 协议原生传递 thinkingBudget: -1(自适应/动态思考开启)时,wrap_request_v2 误将其映射为 0 触发关闭思考逻辑、导致上游请求静默丢失思考链与工具规划能力退化的问题。将 -1 正确识别为动态思考开启并交由入站管线按目标模型通道下发,同时严格保持 thinkingBudget: 0 与 includeThoughts: false 的显式关闭一票否决权。(Fixes #4012, PR #4014, Thanks to @vecyang1)

What's Changed

  • fix(proxy): merge object-only anyOf/oneOf unions instead of keeping one branch by @snailb1007 in #3878

New Contributors

Full Changelog: v5.0.4-beta.19...v5.0.4-beta.20

Don't miss a new Antigravity-Manager release

NewReleases is sending notifications on new releases.