Merge remote-tracking branch 'upstream/main' into feat/zws-v5

CHANGELOG.md

@@ -4,6 +4,56 @@
 
 ---
 
+## [2.8.7] — 2026-03-20
+
+> Sprint: Merge PR #495 (Bottleneck 429 drop), fix #496 (custom embedding providers), triage features.
+
+### Bug Fixes
+
+- **Bottleneck 429 infinite wait** (PR #495 by @xandr0s): On 429, `limiter.stop({ dropWaitingJobs: true })` immediately fails all queued requests so upstream callers can trigger fallback. Limiter is deleted from Map so next request creates a fresh instance.
+- **Custom embedding models unresolvable** (#496): `POST /v1/embeddings` now resolves custom embedding models from ALL provider_nodes (not just localhost). Enables models like `google/gemini-embedding-001` added via dashboard.
+
+### Issues Responded
+
+- **#452** — Per-API-key request-count limits (acknowledged, on roadmap)
+- **#464** — Auto-issue API keys with provider/account limits (needs more detail)
+- **#488** — Auto-update model lists (acknowledged, on roadmap)
+- **#496** — Custom embedding provider resolution (fixed)
+
+---
+
+## [2.8.6] — 2026-03-20
+
+> Sprint: Merge PR #494 (MiniMax role fix), fix KIRO MITM dashboard, triage 8 issues.
+
+### Features
+
+- **MiniMax developer→system role fix** (PR #494 by @zhangqiang8vip): Per-model `preserveDeveloperRole` toggle. Adds "Compatibility" UI in providers page. Fixes 422 "role param error" for MiniMax and similar gateways.
+- **roleNormalizer**: `normalizeDeveloperRole()` now accepts `preserveDeveloperRole` parameter with tri-state behavior (undefined=keep, true=keep, false=convert).
+- **DB**: New `getModelPreserveOpenAIDeveloperRole()` and `mergeModelCompatOverride()` in `models.ts`.
+
+### Bug Fixes
+
+- **KIRO MITM dashboard** (#481/#487): `CLIToolsPageClient` now routes any `configType: "mitm"` tool to `AntigravityToolCard` (MITM Start/Stop controls). Previously only Antigravity was hardcoded.
+- **AntigravityToolCard generic**: Uses `tool.image`, `tool.description`, `tool.id` instead of hardcoded Antigravity values. Guards against missing `defaultModels`.
+
+### Cleanup
+
+- Removed `ZWS_README_V2.md` (development-only docs from PR #494).
+
+### Issues Triaged (8)
+
+- **#487** — Closed (KIRO MITM fixed in this release)
+- **#486** — needs-info (Windows REG.exe PATH issue)
+- **#489** — needs-info (Antigravity projectId missing, OAuth reconnect needed)
+- **#492** — needs-info (missing app/server.js on mise-managed Node)
+- **#490** — Acknowledged (streaming + context cache blocking, fix planned)
+- **#491** — Acknowledged (Codex auth state inconsistency)
+- **#493** — Acknowledged (Modal provider model name prefix, workaround provided)
+- **#488** — Feature request backlog (auto-update model lists)
+
+---
+
 ## [2.8.5] — 2026-03-19
 
 > Sprint: Fix zombie SSE streams, context cache first-turn, KIRO MITM, and triage 5 external issues.

ZWS_README_V2.md

@@ -1,97 +0,0 @@
-# ZWS_README_V2 — developer 角色与「role param error」修复说明
-
-## 一、为什么要修
-
-### 现象
-
-- 使用 **OpenAI Responses API**（`/v1/responses`，body 带 `input`）经 OmniRoute 转发到 **MiniMax** 等 OpenAI 兼容网关时，上游返回 **422**，报错文案为 **`role param error`**。
-
-### 原因
-
-- **Responses API** 允许并会下发消息角色 **`developer`**（与 `system` 语义接近，用于模型指令）。
-- 多数 **OpenAI 兼容网关**（如 MiniMax）只接受 **`system` / `user` / `assistant` / `tool`**，不识别 `developer`，收到即报 422。
-- 修复前，OmniRoute 对这类请求**未做角色转换**，直接把带 `developer` 的 body 转给上游，因此触发 `role param error`。
-
----
-
-## 二、怎么修的
-
-### 1. 后端：统一用「是否保留 developer」控制是否转换
-
-- **`open-sse/services/roleNormalizer.ts`**
-  - `normalizeDeveloperRole(messages, targetFormat, preserveDeveloperRole)`：
-    - 当 **`targetFormat === "openai"` 且 `preserveDeveloperRole !== false`** 时：**保留** `developer`（不转换），兼容官方 OpenAI 等支持 developer 的后端。
-    - 否则：将消息中的 **`developer` 改为 `system`**，避免 MiniMax 等报 422。
-  - 即：**默认保留**（与「以前没有此功能」时的行为一致）；只有**显式关闭保留**（见下文「开关」）时才做 developer→system。
-
-- **不在 translator 里硬编码**
-  - **`open-sse/translator/request/openai-responses.ts`** 中 **不再** 在从 `input` 构建 `messages` 时写死 developer→system，保持 `messages.push({ role: toString(item.role), content })`。
-  - 所有「是否转换 developer」由 **`normalizeRoles`** 根据 `preserveDeveloperRole` 统一处理。
-
-- **Responses 路径补跑一遍 role 管道**
-  - 在 **`open-sse/translator/index.ts`** 的 `translateRequest` 中：当 **`sourceFormat === OPENAI_RESPONSES`** 且已有 `result.messages` 时，在翻译完成后**再执行一次 `normalizeRoles`**，这样从 `input` 刚转出来的 `messages` 也会按开关做 developer→system，与 flag 一致。
-
-- **三态与存储**
-  - **`src/lib/db/models.ts`** 中 **`getModelPreserveOpenAIDeveloperRole(providerId, modelId)`** 返回 **`boolean | undefined`**：
-    - **`undefined`**：未配置 → 路由侧视为「保留 developer」。
-    - **`true`**：显式保留。
-    - **`false`**：显式不保留（developer→system，修 MiniMax 422）。
-  - 配置来源：**custom model 行** 或 **modelCompatOverrides**（无完整 custom 行时用 compat 存该模型的两项兼容选项）；**`mergeModelCompatOverride`** 可写入 **`preserveOpenAIDeveloperRole: false`**，便于「不保留」持久化。
-
-- **chatCore**
-  - 从 **`getModelPreserveOpenAIDeveloperRole`** 取值传给 **`translateRequest`** 的 `options.preserveDeveloperRole`，不再强制 `=== true`，从而支持 `undefined` 的默认保留语义。
-
-### 2. 前端：兼容性入口与「不保留」开关
-
-- **一个「兼容性」按钮 + 弹层**
-  - 每个模型（内置、OpenRouter/兼容、自定义）行上有一个 **「兼容性」** 按钮；点击后弹出**不透明**下拉面板（白/深色背景 + 阴影），内含：
-    - **工具 ID 9 位**：原有「将 tool call id 规范为 9 位」选项。
-    - **不保留 developer 角色**：勾选 = 不保留 = 写入 **`preserveOpenAIDeveloperRole: false`**，路由时 developer→system；**默认不勾选** = 保留 = 与历史行为一致。
-  - 弹层点击外部关闭；结构便于后续增加更多兼容项。
-
-- **绑定关系**
-  - 后端仍只存「是否保留」：`preserveOpenAIDeveloperRole`（true/false/未设置）。
-  - 弹层内「不保留 developer 角色」开关：**勾选 ⟺ `preserveDeveloperRole === false`**，`onChange(checked) => onPreserveChange(!checked)`，不改变后端字段含义。
-
-- **角标**
-  - 当某模型为「不保留」时，在列表上显示短角标（如「不保留」），便于一眼看出该模型已开启 developer→system。
-
-### 3. 文档与 i18n
-
-- 新增/沿用 i18n：**兼容性** 按钮、**不保留 developer 角色** 选项、角标「不保留」等（中/英），见 `src/i18n/messages/`。
-
----
-
-## 三、使用方式（如何避免 422）
-
-- **默认**：不勾选「不保留 developer 角色」→ 保留 developer，行为与修复前一致。
-- **遇到 MiniMax 等 422**：在该模型（或对应兼容节点下的模型）上点击 **「兼容性」**，勾选 **「不保留 developer 角色」** 并保存；之后该模型请求会做 developer→system，422 消失。
-- 仅在使用**官方 OpenAI 且确实需要 developer 角色**时，再保持不勾选或显式保留。
-
----
-
-## 四、当前范围与后续扩展（按协议区分兼容性）
-
-### 当前实现仅针对 OpenAI 协议
-
-- 本轮的「兼容性」配置（**工具 ID 9 位**、**不保留 developer 角色**）在实现上**只对 OpenAI 系协议生效**：请求经 OpenAI Chat Completions / Responses 格式转发时，才会应用这些选项。
-- **同一个模型**可能被多种协议调用（例如同一物理模型既走 OpenAI 兼容端点，也走 Anthropic Messages、或其它网关），理想情况下**兼容性应按「协议」维度**配置，由用户为每个协议单独选择是否做 9 位 tool id、是否不保留 developer 等。
-- **当前页面的「兼容性」入口没有标明协议**，容易让用户误以为这些选项对该模型的所有调用方式都生效；实际上只影响 **OpenAI 协议** 下的行为。
-
-### 后续计划
-
-- **按协议维护兼容性**：将兼容项（如 `normalizeToolCallId`、`preserveOpenAIDeveloperRole`）归属到「协议」维度（如 `openai`、`anthropic` 等），存储与 UI 均按协议展示与编辑，避免误导并支持多协议并存时的差异化配置。
-
----
-
-## 五、涉及文件摘要
-
-| 区域       | 文件                                                                                                                     |
-| ---------- | ------------------------------------------------------------------------------------------------------------------------ |
-| 角色转换   | `open-sse/services/roleNormalizer.ts`，`open-sse/translator/index.ts`，`open-sse/translator/request/openai-responses.ts` |
-| 配置与读写 | `src/lib/db/models.ts`，`src/lib/localDb.ts`，`src/app/api/provider-models/route.ts`                                     |
-| 请求管线   | `open-sse/handlers/chatCore.ts`                                                                                          |
-| 前端 UI    | `src/app/(dashboard)/dashboard/providers/[id]/page.tsx`（兼容性按钮、弹层、不保留开关与角标）                            |
-| 文案       | `src/i18n/messages/zh-CN.json`，`src/i18n/messages/en.json`                                                              |
-
-以上即为「为什么修」与「怎么修」的说明；按 CONTRIBUTING 流程在分支上提交即可。

docs/openapi.yaml

@@ -1,7 +1,7 @@
 openapi: 3.1.0
 info:
   title: OmniRoute API
-  version: 2.8.5
+  version: 2.8.7
   description: |
     OmniRoute is a local-first AI API proxy router. It provides an OpenAI-compatible
     endpoint that routes requests to multiple AI providers with load balancing,

open-sse/services/rateLimitManager.ts

@@ -339,14 +339,19 @@ export function updateFromHeaders(provider, connectionId, headers, status, model
   // Handle 429 — rate limited
   if (status === 429) {
     const retryAfterMs = parseResetTime(retryAfterStr) || 60000; // Default 60s
+    const counts = limiter.counts();
+    const limiterKey = `${provider}:${connectionId}`;
     console.log(
-      `🚫 [RATE-LIMIT] ${provider}:${connectionId.slice(0, 8)} — 429 received, pausing for ${Math.ceil(retryAfterMs / 1000)}s`
+      `🚫 [RATE-LIMIT] ${provider}:${connectionId.slice(0, 8)} — 429 received, pausing for ${Math.ceil(retryAfterMs / 1000)}s, dropping ${counts.QUEUED} queued request(s)`
     );
 
-    limiter.updateSettings({
-      reservoir: 0,
-      reservoirRefreshAmount: limit || 60,
-      reservoirRefreshInterval: retryAfterMs,
+    // Stop the limiter and drop all waiting jobs so they fail immediately
+    // instead of hanging in the queue until reservoir refreshes (which can
+    // be hours for providers like Codex with long rate limit windows).
+    // This lets upstream callers (e.g. LiteLLM) trigger fallback to other providers.
+    // After stop, delete from Map so getLimiter() creates a fresh instance.
+    limiter.stop({ dropWaitingJobs: true }).finally(() => {
+      limiters.delete(limiterKey);
     });
     return;
   }

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "omniroute",
-  "version": "2.8.5",
+  "version": "2.8.7",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omniroute",
-      "version": "2.8.5",
+      "version": "2.8.7",
       "hasInstallScript": true,
       "license": "MIT",
       "workspaces": [

package.json

@@ -1,6 +1,6 @@
 {
   "name": "omniroute",
-  "version": "2.8.5",
+  "version": "2.8.7",
   "description": "Smart AI Router with auto fallback — route to FREE & cheap models, zero downtime. Works with Cursor, Cline, Claude Desktop, Codex, and any OpenAI-compatible tool.",
   "type": "module",
   "bin": {

src/app/(dashboard)/dashboard/cli-tools/CLIToolsPageClient.tsx

@@ -267,6 +267,18 @@ export default function CLIToolsPageClient({ machineId }) {
           />
         );
       default:
+        // #487: Any tool with configType "mitm" should use the MITM card (Start/Stop controls)
+        if (tool.configType === "mitm") {
+          return (
+            <AntigravityToolCard
+              key={toolId}
+              {...commonProps}
+              activeProviders={getActiveProviders()}
+              hasActiveProviders={hasActiveProviders}
+              cloudEnabled={cloudEnabled}
+            />
+          );
+        }
         return (
           <DefaultToolCard
             key={toolId}

src/app/(dashboard)/dashboard/cli-tools/components/AntigravityToolCard.tsx

@@ -41,7 +41,7 @@ export default function AntigravityToolCard({
 
   const loadSavedMappings = async () => {
     try {
-      const res = await fetch("/api/cli-tools/antigravity-mitm/alias?tool=antigravity");
+      const res = await fetch(`/api/cli-tools/antigravity-mitm/alias?tool=${tool.id}`);
       if (res.ok) {
         const data = await res.json();
         const aliases = data.aliases || {};
@@ -187,7 +187,7 @@ export default function AntigravityToolCard({
       const res = await fetch("/api/cli-tools/antigravity-mitm/alias", {
         method: "PUT",
         headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({ tool: "antigravity", mappings: modelMappings }),
+        body: JSON.stringify({ tool: tool.id, mappings: modelMappings }),
       });
 
       if (!res.ok) {
@@ -211,7 +211,7 @@ export default function AntigravityToolCard({
         <div className="flex items-center gap-3">
           <div className="size-8 flex items-center justify-center shrink-0">
             <Image
-              src="/providers/antigravity.png"
+              src={tool.image || "/providers/antigravity.png"}
               alt={tool.name}
               width={32}
               height={32}
@@ -235,7 +235,7 @@ export default function AntigravityToolCard({
                 </Badge>
               )}
             </div>
-            <p className="text-xs text-text-muted truncate">{t("toolDescriptions.antigravity")}</p>
+            <p className="text-xs text-text-muted truncate">{tool.description}</p>
           </div>
         </div>
         <span
@@ -306,7 +306,7 @@ export default function AntigravityToolCard({
                 )}
               </div>
 
-              {tool.defaultModels.map((model) => (
+              {(tool.defaultModels || []).map((model) => (
                 <div key={model.alias} className="flex items-center gap-2">
                   <span className="w-32 shrink-0 text-sm font-semibold text-text-main text-right">
                     {model.name}

src/app/api/v1/embeddings/route.ts

@@ -12,6 +12,7 @@ import {
   getEmbeddingProvider,
   buildDynamicEmbeddingProvider,
   type EmbeddingProviderNodeRow,
+  type EmbeddingProvider,
 } from "@omniroute/open-sse/config/embeddingRegistry.ts";
 import { errorResponse } from "@omniroute/open-sse/utils/error.ts";
 import { HTTP_STATUS } from "@omniroute/open-sse/config/constants.ts";
@@ -116,9 +117,9 @@ export async function POST(request) {
   // Load local provider_nodes for embedding routing (only localhost — prevents auth bypass/SSRF)
   let dynamicProviders: ReturnType<typeof buildDynamicEmbeddingProvider>[] = [];
   try {
-    const nodes = await getProviderNodes();
+    const nodes = (await getProviderNodes()) as unknown as EmbeddingProviderNodeRow[];
     dynamicProviders = (Array.isArray(nodes) ? nodes : [])
-      .filter((n: EmbeddingProviderNodeRow) => {
+      .filter((n) => {
         // provider_nodes apiType is "chat" or "responses" (not "embeddings") — local OpenAI-compatible
         // backends expose /embeddings under the same base URL as chat, so we build the URL as baseUrl + /embeddings.
         if (n.apiType !== "chat" && n.apiType !== "responses") return false;
@@ -157,9 +158,38 @@ export async function POST(request) {
   }
 
   // Resolve provider config — dynamic first (local override), then hardcoded
-  const providerConfig =
+  let providerConfig: EmbeddingProvider | null =
     dynamicProviders.find((dp) => dp.id === provider) || getEmbeddingProvider(provider) || null;
 
+  // #496: Fallback — resolve from ALL provider_nodes (not just localhost)
+  // This enables custom embedding models (e.g. google/gemini-embedding-001) whose
+  // providers have remote baseUrls. Safe because getProviderCredentials() authenticates.
+  if (!providerConfig) {
+    try {
+      const allNodes = (await getProviderNodes()) as unknown as EmbeddingProviderNodeRow[];
+      const matchingNode = (Array.isArray(allNodes) ? allNodes : []).find(
+        (n) =>
+          n.prefix === provider && (n.apiType === "chat" || n.apiType === "responses") && n.baseUrl
+      );
+      if (matchingNode) {
+        const baseUrl = String(matchingNode.baseUrl).replace(/\/+$/, "");
+        providerConfig = {
+          id: matchingNode.prefix,
+          baseUrl: `${baseUrl}/embeddings`,
+          authType: "apikey",
+          authHeader: "bearer",
+          models: [],
+        };
+        log.info(
+          "EMBED",
+          `Resolved custom embedding provider: ${provider} → ${providerConfig.baseUrl}`
+        );
+      }
+    } catch (err) {
+      log.error("EMBED", `Failed to resolve custom embedding provider ${provider}: ${err}`);
+    }
+  }
+
   if (!providerConfig) {
     return errorResponse(
       HTTP_STATUS.BAD_REQUEST,