lbjlaq/Antigravity-Manager v4.9.8-beta.9

Antigravity Tools v4.9.8-beta.9

on GitHub

• [Eliminate Ineffective Prod Endpoint & Extend Streaming First-Chunk Watchdog to 120s] (PR #3691, Fixes #3506, Ref #3525, Thanks to @EricZhou05):
  • Eliminate 100% 429-Yielding Prod Public Endpoint: Empirical multi-scale benchmarks demonstrated that the public Prod endpoint (cloudcode-pa.googleapis.com) unconditionally returns 429 RESOURCE_EXHAUSTED within ~1s across all context window sizes, acting as an invalid fallback that triggers false-positive account rate-limiting and 503 error loops. Completely removed Prod from V1_INTERNAL_BASE_URL_FALLBACKS, retaining only official native primary Daily and verified backup Sandbox endpoints to sever the fatal failover loop. (PR #3691, Ref #3525, Thanks to @EricZhou05)
  • Recalibrate Streaming Watchdog First-Chunk Timeout to 120s Baseline: Resolved a flaw where the previous adaptive watchdog formula max(15, body_len / 256KB + 15) accounted only for network transmission bandwidth while overlooking long-context and Thinking reasoning time-to-first-token (TTFT ~35s), prematurely aborting requests at 16s. Streaming generation (streamGenerateContent) first-chunk timeout is now relaxed to a 120-second baseline (max(120, body_len / 64KB + 120)), fully accommodating model reasoning and cold-cache preheating while preventing false-positive 429/503 cutoffs; lightweight non-streaming requests retain the 15-second fast-probing timeout. (PR #3691, Fixes #3506, Thanks to @EricZhou05)
• [Preserve Multi-Turn Message Roles for Responses-Style Chat Completions] (PR #3672, Fixes #3671, Thanks to @blackdm666):
  • Preserve System/User/Assistant Roles in Array Inputs: Resolved an issue where /v1/chat/completions receiving Responses-style payloads (instructions / input without messages) collapsed array input items into a single user text via to_string(), destroying role boundaries and causing Gemini to echo conversation history rather than answering the latest turn. Reused convert_codex_to_openai_request to accurately restore structured message roles per turn while preserving string input compatibility. (PR #3672, Fixes #3671, Thanks to @blackdm666)
• [Intelligent 403 State Identification, Upstream Action URL Extraction & Closed-Loop Self-Healing] (Ref #3630, Ref #3684):
  • Full 403 Body Capture & Action URL Extraction (validation_url / appeal_url): Eliminated premature discarding of 403 HTTP response bodies in modules/quota.rs. Full Google RPC error payload details are now parsed via extract_action_url to accurately extract validation_url (human/captcha verification links) and appeal_url (Terms of Service appeal links) from details.metadata and persist them into the account model. Integrated with frontend AccountErrorDialog to immediately present actionable buttons ("Verify Now", "Copy Verification URL", "Submit Appeal"), completely replacing vague and opaque 403 error states.
  • Decouple Temporary Verification (VALIDATION_REQUIRED) from Permanent Ban: Refactored gateway handlers across Gemini, Claude, and OpenAI (handlers/gemini.rs, handlers/claude.rs, handlers/openai.rs) and warmup probes to make temporary validation and permanent ban (set_forbidden) mutually exclusive. Encountering VALIDATION_REQUIRED now places the account into a 10-minute temporary observation window while retaining the verification URL, eliminating accidental permanent proxy_disabled = true deadlocks for temporary verification prompts.
  • Automatic Deadlock Clearing & Closed-Loop Recovery on Successful Probe: Redesigned update_account_quota and clear_validation_blocked so that when an account completes verification externally (e.g. via the official IDE or web browser) and subsequent quota probing succeeds (is_forbidden == false), the system automatically clears the proxy_disabled flag and resets stale 403 error reasons. Dispatches trigger_account_reload and broadcasts updated state to the frontend, restoring recovered accounts back into the active proxy pool seamlessly without manual user intervention.
• [Classic Auto Hot-Switch on Depleted Quota, Dual-Pool Monitoring & Legacy Sync Cleanup] (Ref #3509):
  • Automatic Seamless Hot-Switching for Antigravity Classic: Integrated an auto-switch engine directly into the background quota inspection pipeline (refresh_all_quotas_internal). When Antigravity Classic is running and the active account's quota falls below the configured threshold, the engine automatically selects the healthiest available account from the pool based on threshold satisfaction, tier priority (Pro/Ultra > Free), highest remaining quota, and LRU anti-oscillation. Seamlessly restarts the backend language_server via supervisor within 2 seconds without restarting the main IDE window, guarded by a 120-second cooldown lock and desktop notifications. (Ref #3509)
  • Streamlined Dual Quota Pool Monitoring & Threshold Control: Eliminated cluttered per-model selection tags in favor of intuitive dual-pool cards aligned with the dashboard — "Gemini Pool (Pro / Flash / Image)" and "Third-Party Models (Claude 3.7 / 3.5 / GPT)", with customizable trigger thresholds from 0% to 99%.
  • Removal of Redundant Legacy Account Sync & UI Streamlining: Removed the confusing legacy "Auto Sync Current Account" card from Account Settings and removed its unused background interval timer, keeping the settings interface clean and focused.
  • Full 12-Language Localization: Synchronized complete translations across all 12 supported languages including Chinese (Simplified & Traditional), English, Korean, Japanese, Russian, Spanish, Portuguese, Turkish, Vietnamese, Arabic, and Malay.
• [Local Account Scanning Alignment & Legacy Backup Removal]:
  • Streamlined Removal of Legacy V1 Backup & UI Simplification: Completely eliminated the legacy V1 (~/.antigravity-agent) import path and divider in the Add Account dialog, focusing the interface cleanly into "Auto-Scan Local Accounts" and "Import Custom DB" with synchronized multi-language typography and improved spatial efficiency.
  • Aggregated Classic & IDE Database Scanning (state.vscdb): Introduced db::get_all_antigravity_db_paths() to remove the hardcoded restriction that previously limited scanning to Classic Antigravity. Now concurrently aggregates all candidate databases across both Classic Antigravity and Antigravity IDE (including antigravity-ide variants), resolving the issue where accounts logged in via the IDE were silently overlooked.
  • Independent Concurrent Antigravity CLI (agy) Credential Detection: Introduced read_antigravity_cli_credentials() to bypass Keychain short-circuiting that previously masked CLI tokens. Independently scans Antigravity CLI credentials and backups (~/.gemini/antigravity-cli/antigravity-oauth-token*), strictly targeting Antigravity tokens for automated multi-source deduplication and onboarding.
• [Independent Active Account Tracking & Contextual Multi-Client Badges for Classic, IDE & CLI (agy)] (Ref #3630, Ref #3686):
  • Independent Multi-Client Active Account Ground-Truth Detection: Added detect_active_clients_status and get_active_clients_status Tauri commands to accurately determine the live logged-in account for each client environment independently — Antigravity Classic (Keychain / DB), Antigravity IDE (state.vscdb), and Antigravity CLI (agy credentials & Keyring). Resolves status collision and ambiguity caused by a single global current_account_id.
  • Persistent Multi-Target Switch Tracking: Extended AccountIndex with current_classic_account_id, current_ide_account_id, and current_agy_account_id fields, persisting exact client-specific active assignments whenever an account is switched for a targeted client.
  • Contextual Multi-Client Badges & Row Highlighting: Implemented distinct client badges ([Classic], [IDE], [CLI]) across both Card (AccountCard) and Table (AccountTable) views. Supports multi-badge combinations when an account is active in multiple clients, and highlights borders/rows with the active theme tint for superior readability across multi-client workflows.
  • Complete Multi-Language Localization: Synchronized badges and explanatory hover tooltips across English, Simplified Chinese, and Traditional Chinese.

• [剔除不可用 Prod 端点并放宽流式首包看门狗超时至 120s] (PR #3691, Fixes #3506, Ref #3525, Thanks to @EricZhou05):
  • 剔除 100% 报 429 的 Prod 生产公共端点: 经全规模对照实测，Prod 生产端点（cloudcode-pa.googleapis.com）对该接口全上下文规模均在 1 秒左右必然返回 429 RESOURCE_EXHAUSTED，无法作为有效 fallback，切过去只会引发假阳性限流并触发账号禁用死循环。在 V1_INTERNAL_BASE_URL_FALLBACKS 中彻底剔除 Prod 端点，仅保留官方原生主力 Daily 与实测可用的备用 Sandbox 双端点容灾，切断流向必死 Prod 的自杀通道。(PR #3691, Ref #3525, Thanks to @EricZhou05)
  • 校准流式接口自适应防假死首包超时门槛至 120 秒: 修复早期看门狗超时公式 max(15, body_len / 256KB + 15) 仅考虑网络带宽而忽略「大模型长上下文 + Thinking 深度思考首字生成 (TTFT)」导致 16 秒误判假死掐断的问题。将流式接口（streamGenerateContent）首包超时保底放宽至 120 秒（max(120, body_len / 64KB + 120)），充分保障大模型冷缓存预热与思考链推理时间，杜绝长会话被反代误杀切入 429 与 503 假阳性；非流式普通请求维持 15 秒快速探活回退。(PR #3691, Fixes #3506, Thanks to @EricZhou05)
• [Chat Completions 对齐 Responses 格式多轮角色保留与结构化转换] (PR #3672, Fixes #3671, Thanks to @blackdm666):
  • Responses 格式数组 input 完整保留多轮对话角色: 解决 /v1/chat/completions 接收仅带 instructions / input 而无 messages 的 Responses 格式请求时，原实现将数组形式的 input 强行 to_string() 为单条 user 文本导致丢失 system / user / assistant 角色边界、引发 Gemini 上游把整段历史当新问题重答的缺陷。改为无缝复用 convert_codex_to_openai_request 结构化转换器，按条目精确还原各轮次角色，并兼容字符串格式 input 输入。(PR #3672, Fixes #3671, Thanks to @blackdm666)
• [403 状态智能识别、上游操作指引提取与全自动闭环自愈] (Ref #3630, Ref #3684):
  • 403 报文体全量捕获与动作链接提取 (validation_url / appeal_url): 在配额探测模块（modules/quota.rs）中根除 403 立即丢弃响应体的缺陷；全量读取 Google RPC 错误详情报文，通过新增的 extract_action_url 精准解析 details.metadata 中的 validation_url（人机/安全验证链接）与 appeal_url（服务条款申诉链接）并持久化至账号模型。配合前端 AccountErrorDialog，即时点亮「点击去验证」、「复制验证链接」或「前往申诉」交互按钮，彻底告别空洞抽象的单一 403 报错状态。
  • 临时人机验证 (VALIDATION_REQUIRED) 与永久封禁严格解耦: 在 Gemini、Claude 与 OpenAI 协议网关（handlers/gemini.rs、handlers/claude.rs、handlers/openai.rs）及预热检测中，将临时人机验证与永久封禁（set_forbidden）改造为严格互斥逻辑。发生 VALIDATION_REQUIRED 时仅进入 10 分钟观察窗口并记录验证链接，不再强行写入永久 proxy_disabled = true，杜绝临时验证被误判为永久死刑。
  • 配额探测成功自动解除死锁与闭环自愈: 彻底重塑 update_account_quota 与 clear_validation_blocked，当账号在外部（如官方 IDE 或网页端）通过验证、配额探测恢复正常（is_forbidden == false）时，自动解除 proxy_disabled 禁用状态并清空历史 403 错误信息；同时触发 trigger_account_reload 与前端状态推送，使恢复的账号无需人工干预即可全自动重归现役代理号池。
• [经典款配额自动无感热切换、双核算力池监控与冗余同步清理] (Ref #3509):
  • 经典款配额耗尽/低于阈值全自动无感热切: 在后台配额巡检流水线（refresh_all_quotas_internal）中接入自动热切引擎；当检测到 Antigravity 经典款当前正在运行且活动账号配额低于预设阈值时，自动在账号池中按「满足阈值门槛 + 付费等级优先 (Pro/Ultra > Free) + 可用配额更多 + LRU 防震荡」挑选最佳可用账号。通过精准终止后台 language_server 子进程并自动更新系统 Keyring，由内置 supervisor 在 2 秒内原地无感重启生效，并附带 120 秒冷却锁与系统桌面通知。 (Ref #3509)
  • 极简双核算力池监控与阈值调节: 彻底告别繁复密集的细分子模型标签，与仪表盘配额体系深度对齐，提供「Gemini 算力池 (Pro / Flash / 绘图)」与「第三方模型池 (Claude 3.7 / 3.5 / GPT)」两大核心池多选卡片；支持用户自定义 0% ~ 99% 换号触发阈值。
  • 清理冗余的账号同步轮询与界面精简: 彻底移除设置页中容易引起与“自动刷新配额”混淆的历史老旧功能「自动获取当前账号」配置卡片，同步清理前端无效定时器，保持账号设置页聚焦整洁。
  • 全量 12 种多国语言覆盖: 经典款自动热切相关文案完整覆盖中文（简/繁）、英文、韩文、日文、俄文、西文、葡文、土文、越文、阿文、马文等全部 12 种语言。
• [本地账号扫描全链路对齐与旧版备份移除]:
  • 精简移除旧版 V1 备份与界面轻量化: 在新增账号弹窗中彻底移除早期 V1 版本（~/.antigravity-agent）冗余导入入口与分割线，将界面聚焦并精简为「自动扫描本地已登录账号」与「从自定义 DB 导入」，同步优化多语言文案与布局空间。
  • 经典款与 IDE 款数据库候选全量聚合 (state.vscdb): 新增 db::get_all_antigravity_db_paths()，破除先前导入扫描仅检索经典款 Antigravity 的限制；支持无缝并发聚合经典款与 Antigravity IDE（含 antigravity-ide 变体）所有活动数据库实例，彻底解决 IDE 款已登录账号被静默漏检的问题。
  • Antigravity CLI (agy) 原生凭据独立并发探测: 新增 read_antigravity_cli_credentials()，解决系统 Keyring 短路拦截导致无法检索 CLI 令牌的缺陷；独立扫描 Antigravity CLI 原生凭据与备份文件（~/.gemini/antigravity-cli/antigravity-oauth-token*），严格限定仅处理 Antigravity 生态凭据，实现多源账号自动去重全量入池。
• [经典款、IDE 款与 CLI (agy) 三端当前账号独立追踪与情境化多端徽章] (Ref #3630, Ref #3686):
  • 三端活动账号独立探测与事实对齐: 新增 detect_active_clients_status 与 get_active_clients_status Tauri 命令，分别精准探查 Antigravity 经典款（Keychain / DB）、Antigravity IDE (state.vscdb) 以及 Antigravity CLI (agy 原生凭据与 Keyring) 三端真实的当前生效账号；告别以往单全局 current_account_id 互相覆盖、状态混乱与归属不明的问题。
  • 持久化索引结构与多端切换记忆: 在 AccountIndex 中扩展 current_classic_account_id、current_ide_account_id 与 current_agy_account_id 字段；在针对特定目标端（Classic / IDE / CLI）切换生效账号时，精准更新对应端的活动账号记录并持久化索引。
  • 情境化多端徽章呈现与列表聚焦: 在卡片视图（AccountCard）与表格视图（AccountTable）中引入专属情境化徽章（[经典款]、[IDE]、[CLI]）；支持同一账号被多端同时登录时直观复合展示，并对任一活动端账号赋予专属高亮边框与行背景，大幅提升多环境并行开发时的辨识度与体验。
  • 全量多语言国际化支持: 覆盖简体中文、繁体中文、英文客户端徽章文案与悬浮提示说明。

What's Changed

• fix(upstream): 剔除不可用 Prod 端点并放宽流式首包 Watchdog 超时至 120s (Fixes #3506, Ref #3525) by @EricZhou05 in #3691

Full Changelog: v4.9.8-beta.8...v4.9.8-beta.9