@amir83nasr/10router 0.1.14 → 0.1.16

package/open-sse/providers/registry/index.js

@@ -4,56 +4,17 @@ import p1 from "./claude.js";
 import p2 from "./codex.js";
 import p3 from "./cursor.js";
 import p4 from "./deepseek.js";
-import p5 from "./fish-audio.js";
-import p6 from "./gemini.js";
-import p7 from "./glm-cn.js";
-import p8 from "./glm.js";
-import p9 from "./groq.js";
-import p10 from "./huggingface.js";
-import p11 from "./iflow.js";
-import p12 from "./minimax-cn.js";
-import p13 from "./minimax.js";
-import p14 from "./openai.js";
-import p15 from "./opencode-go.js";
-import p16 from "./opencode.js";
-import p17 from "./openrouter.js";
-import p18 from "./siliconflow.js";
-import p19 from "./vercel-ai-gateway.js";
-import p20 from "./xai.js";
-import p21 from "./xquik.js";
-import p22 from "./xiaomi-mimo.js";
-import p23 from "./xiaomi-tokenplan.js";
-import p24 from "./zed.js";
-import p25 from "./mimo-free.js";
-import p26 from "./opencode-zen.js";
-import p27 from "./qoder.js";
-export default [
-  p0,
-  p1,
-  p2,
-  p3,
-  p4,
-  p5,
-  p6,
-  p7,
-  p8,
-  p9,
-  p10,
-  p11,
-  p12,
-  p13,
-  p14,
-  p15,
-  p16,
-  p17,
-  p18,
-  p19,
-  p20,
-  p21,
-  p22,
-  p23,
-  p24,
-  p25,
-  p26,
-  p27,
-];
+import p5 from "./gemini.js";
+import p6 from "./glm.js";
+import p7 from "./iflow.js";
+import p8 from "./openai.js";
+import p9 from "./opencode-go.js";
+import p10 from "./opencode.js";
+import p11 from "./openrouter.js";
+import p12 from "./xai.js";
+import p13 from "./xiaomi-mimo.js";
+import p14 from "./zed.js";
+import p15 from "./mimo-free.js";
+import p16 from "./opencode-zen.js";
+import p17 from "./qoder.js";
+export default [p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17];

package/open-sse/providers/registry/opencode-go.js

@@ -17,7 +17,7 @@ export default {
   },
   category: "apikey",
   transport: {
-    baseUrl: "https://opencode.ai/inference/go/v1/chat/completions",
+    baseUrl: "https://opencode.ai/inference/go/openai/v1/chat/completions",
     headers: {},
     usage: {
       url: "https://opencode.ai/inference/go/v1/usage",
@@ -29,7 +29,7 @@ export default {
   transports: [
     {
       format: "openai",
-      baseUrl: "https://opencode.ai/inference/go/v1/chat/completions",
+      baseUrl: "https://opencode.ai/inference/go/openai/v1/chat/completions",
       auth: { combined: true, header: "Authorization", scheme: "bearer" },
       // OpenAI dialect only: the gateway rejects `thinking`/`enable_thinking`
       // (400 unknown field "thinking"), so normalize reasoning to reasoning_effort.
@@ -37,12 +37,12 @@ export default {
     },
     {
       format: "claude",
-      baseUrl: "https://opencode.ai/inference/go/v1/messages",
+      baseUrl: "https://opencode.ai/inference/go/anthropic/v1/messages",
       auth: { combined: true, header: "x-api-key", scheme: "raw", anthropicVersion: true },
     },
     {
       format: "openai-responses",
-      baseUrl: "https://opencode.ai/inference/go/v1/responses",
+      baseUrl: "https://opencode.ai/inference/go/openai/v1/responses",
       auth: { combined: true, header: "Authorization", scheme: "bearer" },
       thinkingFormat: "openai",
     },
@@ -57,6 +57,7 @@ export default {
     { id: "kimi-k2.7-code", name: "Kimi K2.7 Code", supportedFormats: ["openai"] },
     { id: "kimi-k2.6", name: "Kimi K2.6", supportedFormats: ["openai"] },
     { id: "kimi-k3", name: "Kimi K3", supportedFormats: ["openai"] },
+    { id: "kimi-k3.1-pro", name: "Kimi K3.1 Pro", supportedFormats: ["openai"] },
     {
       id: "deepseek-v4-pro",
       name: "DeepSeek V4 Pro",
@@ -78,19 +79,27 @@ export default {
     { id: "mimo-v2.5", name: "MiMo V2.5", supportedFormats: ["openai"] },
     { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro", supportedFormats: ["openai"] },
     { id: "minimax-m3", name: "MiniMax M3", supportedFormats: ["openai", "claude"] },
+    { id: "minimax-m3.1-flash", name: "MiniMax M3.1 Flash", supportedFormats: ["openai"] },
     { id: "minimax-m2.7", name: "MiniMax M2.7", supportedFormats: ["openai", "claude"] },
+    { id: "claude-haiku-5-5", name: "Claude Haiku 5.5", supportedFormats: ["claude"] },
     { id: "qwen3.8-max", name: "Qwen 3.8 Max", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.8-flash", name: "Qwen 3.8 Flash", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.7-plus", name: "Qwen 3.7 Plus", supportedFormats: ["openai", "claude"] },
     { id: "hy4-preview", name: "Hy4 Preview", supportedFormats: ["openai"] },
     { id: "hy3", name: "Hy3", supportedFormats: ["openai"] },
-    { id: "space-bunny-free", name: "Space Bunny Free", supportedFormats: ["openai"] },
+    {
+      id: "space-bunny-free",
+      name: "Space Bunny Free",
+      supportedFormats: ["openai"],
+      tier: "free",
+    },
     {
       id: "longcat-2.5-preview-free",
       name: "LongCat 2.5 Preview Free",
       supportedFormats: ["openai"],
+      tier: "free",
     },
-    // Served by /inference/go/v1/responses only — the responses-only entry forces chatCore
+    // Served by /inference/go/openai/v1/responses only — the responses-only entry forces chatCore
     // past the sourceFormat-matched transports into translation (see chatCore guard).
     {
       id: "grok-4.7",

package/open-sse/providers/registry/opencode-zen.js

@@ -19,9 +19,6 @@ export default {
   transport: {
     baseUrl: "https://opencode.ai/inference/openai/v1/chat/completions",
     headers: {},
-    usage: {
-      url: "https://opencode.ai/zen/v1/usage",
-    },
   },
   // Multi-endpoint: pick the transport matching the client sourceFormat to skip
   // translation. Mirrors opencode-go, pointed at /inference (endpoint families per
@@ -254,19 +251,36 @@ export default {
     { id: "big-pickle", name: "Big Pickle", supportedFormats: ["openai"] },
     { id: "union-alpha", name: "Union Alpha", supportedFormats: ["claude"] },
     // Free tier on the keyed lane (chat completions)
-    { id: "deepseek-v4-flash-free", name: "DeepSeek V4 Flash Free", supportedFormats: ["openai"] },
-    { id: "mimo-v2.6-flash-free", name: "MiMo V2.6 Flash Free", supportedFormats: ["openai"] },
-    { id: "mimo-v2.5-free", name: "MiMo V2.5 Free", supportedFormats: ["openai"] },
+    {
+      id: "deepseek-v4-flash-free",
+      name: "DeepSeek V4 Flash Free",
+      supportedFormats: ["openai"],
+      tier: "free",
+    },
+    {
+      id: "mimo-v2.6-flash-free",
+      name: "MiMo V2.6 Flash Free",
+      supportedFormats: ["openai"],
+      tier: "free",
+    },
+    { id: "mimo-v2.5-free", name: "MiMo V2.5 Free", supportedFormats: ["openai"], tier: "free" },
     {
       id: "ling-3.0-flash-fin-free",
       name: "Ling 3.0 Flash Fin Free",
       supportedFormats: ["openai"],
+      tier: "free",
+    },
+    {
+      id: "nemotron-3-ultra-free",
+      name: "Nemotron 3 Ultra Free",
+      supportedFormats: ["openai"],
+      tier: "free",
     },
-    { id: "nemotron-3-ultra-free", name: "Nemotron 3 Ultra Free", supportedFormats: ["openai"] },
     {
       id: "nemotron-3.5-lightning-free",
       name: "Nemotron 3.5 Lightning Free",
       supportedFormats: ["openai"],
+      tier: "free",
     },
     // Free tier on the keyed lane (responses)
     {
@@ -274,20 +288,22 @@ export default {
       name: "Muse Spark 1.3 Contributor Free",
       targetFormat: "openai-responses",
       supportedFormats: ["openai-responses"],
+      tier: "free",
     },
     {
       id: "muse-spark-1.2-contributor-free",
       name: "Muse Spark 1.2 Contributor Free",
       targetFormat: "openai-responses",
       supportedFormats: ["openai-responses"],
+      tier: "free",
     },
     // System One (Jev) decision models on the native /systemone endpoint
     { id: "jev-1.13", name: "Jev 1.13", kind: "systemone" },
-    { id: "jev-1.13-free", name: "Jev 1.13 Free", kind: "systemone" },
+    { id: "jev-1.13-free", name: "Jev 1.13 Free", kind: "systemone", tier: "free" },
   ],
   serviceKinds: ["llm", "systemone"],
   systemoneConfig: {
-    baseUrl: "https://opencode.ai/inference/v1/systemone",
+    baseUrl: "https://opencode.ai/inference/systemone/v1/systemone",
     headers: {
       "x-opencode-client": "cli",
       "User-Agent": "opencode/latest/2.0.26/cli",

package/open-sse/providers/registry/opencode.js

@@ -35,18 +35,20 @@ export default {
       id: "muse-spark-1.2-contributor-free",
       name: "Muse Spark 1.2 Contributor Free",
       targetFormat: "openai-responses",
+      tier: "free",
     },
     {
       id: "muse-spark-1.3-contributor-free",
       name: "Muse Spark 1.3 Contributor Free",
       targetFormat: "openai-responses",
+      tier: "free",
     },
-    { id: "union-alpha", name: "Union Alpha Free", targetFormat: "claude" },
-    { id: "jev-1.13-free", name: "Jev 1.13 Free", kind: "systemone" },
+    { id: "union-alpha", name: "Union Alpha Free", targetFormat: "claude", tier: "free" },
+    { id: "jev-1.13-free", name: "Jev 1.13 Free", kind: "systemone", tier: "free" },
   ],
   serviceKinds: ["llm", "systemone"],
   systemoneConfig: {
-    baseUrl: "https://opencode.ai/inference/v1/systemone",
+    baseUrl: "https://opencode.ai/inference/systemone/v1/systemone",
     headers: {
       "x-opencode-client": "cli",
       "User-Agent": "opencode/latest/2.0.26/cli",

package/open-sse/services/accountFallback.js

@@ -28,6 +28,7 @@ export function checkFallbackError(status, errorText, backoffLevel = 0) {
   for (const rule of ERROR_RULES) {
     // Text-based rule: match substring in error message
     if (rule.text && lowerError && lowerError.includes(rule.text)) {
+      if (rule.noFallback) return { shouldFallback: false, cooldownMs: 0 };
       if (rule.backoff) {
         const newLevel = Math.min(backoffLevel + 1, BACKOFF_CONFIG.maxLevel);
         return {

package/open-sse/services/opencodeGoProbe.js

@@ -0,0 +1,39 @@
+import { getProviderModels } from "../config/providerModels.js";
+import { OPENCODE_GO_ENDPOINTS } from "../config/opencodeGo.js";
+import { OpenCodeGoExecutor } from "../executors/opencode-go.js";
+import { resolveTransport } from "./provider.js";
+
+const OPENCODE_GO = "opencode-go";
+
+// Connection probe for one OpenCode Go endpoint. The transport, URL, auth and the
+// x-opencode-session header come from the executor, so the probe sends what real
+// traffic sends. A hand-built request to /inference/go/* is rejected with MissingSessionID.
+export function buildOpenCodeGoProbe(apiKey, format = "openai") {
+  const transport = resolveTransport(OPENCODE_GO, format);
+  const model = getProviderModels(OPENCODE_GO).find((m) =>
+    m.supportedFormats?.includes(format),
+  )?.id;
+  if (!transport || !model) throw new Error(`OpenCode Go has no ${format} endpoint`);
+
+  const executor = new OpenCodeGoExecutor();
+  const credentials = executor.prepareRequestCredentials({
+    credentials: { apiKey, runtimeTransport: transport },
+  });
+  const url = executor.buildUrl(model, false, 0, credentials);
+
+  return {
+    url,
+    headers: executor.buildHeaders(credentials, false, url, model),
+    body: JSON.stringify({
+      model,
+      messages: [{ role: "user", content: "ping" }],
+      max_tokens: 1,
+      stream: false,
+    }),
+  };
+}
+
+// One probe per endpoint in OPENCODE_GO_ENDPOINTS.
+export function buildOpenCodeGoProbes(apiKey) {
+  return OPENCODE_GO_ENDPOINTS.map(({ format }) => buildOpenCodeGoProbe(apiKey, format));
+}

package/open-sse/services/usage/glm.js

@@ -1,31 +1,25 @@
 /**
- * GLM Coding Plan usage (international + China regions)
+ * GLM Coding Plan usage
  */
 
 import { proxyAwareFetch } from "../../utils/proxyFetch.js";
 import { U } from "./shared.js";
 
-// GLM quota endpoints (region-aware) — url from registry transport.usage
-const GLM_QUOTA_URLS = {
-  international: U("glm").url,
-  china: U("glm-cn").url,
-};
+// GLM quota endpoint — url from registry transport.usage
+const GLM_QUOTA_URL = U("glm").url;
 
 /**
- * GLM Coding Plan usage (international + China regions)
+ * GLM Coding Plan usage
  * Supports both TOKENS_LIMIT and CREDIT_LIMIT and dynamic intervals (e.g. session 5h, weekly 7d).
  */
-export async function getGlmUsage(apiKey, provider, proxyOptions = null) {
+export async function getGlmUsage(apiKey, proxyOptions = null) {
   if (!apiKey) {
     return { message: "GLM API key not available." };
   }
 
-  const region = provider === "glm-cn" ? "china" : "international";
-  const quotaUrl = GLM_QUOTA_URLS[region];
-
   try {
     const response = await proxyAwareFetch(
-      quotaUrl,
+      GLM_QUOTA_URL,
       {
         headers: {
           Authorization: `Bearer ${apiKey}`,

package/open-sse/services/usage/misc.js

@@ -1,5 +1,5 @@
 /**
- * Misc usage handlers (iFlow, Ollama, GLM, Vercel AI Gateway, Qoder)
+ * Misc usage handlers (iFlow, Ollama, GLM, Qoder)
  */
 
 import { proxyAwareFetch } from "../../utils/proxyFetch.js";
@@ -7,10 +7,6 @@ import { U } from "./shared.js";
 
 export { getGlmUsage } from "./glm.js";
 
-// Vercel AI Gateway credits endpoint
-// Returns { balance: "95.50", total_used: "4.50" } (USD as decimal strings).
-const VERCEL_AI_GATEWAY_CREDITS_URL = U("vercel-ai-gateway").url;
-
 /**
  * iFlow Usage
  */
@@ -124,93 +120,6 @@ export async function getOllamaUsage(apiKey, providerSpecificData, proxyOptions
   }
 }
 
-/**
- * Vercel AI Gateway usage — credit balance for the API key
- *
- * Calls GET /v1/credits which returns:
- *   { "balance": "95.50", "total_used": "4.50" }   (USD as decimal strings)
- *
- * We surface this as a single "Balance ($)" quota row so the existing
- * QuotaTable / progress-bar UI can render it. used = total_used,
- * total = balance + total_used (the original credit allotment), so the
- * remaining percentage equals balance / total.
- *
- * Docs: https://vercel.com/docs/ai-gateway/usage
- */
-export async function getVercelAiGatewayUsage(apiKey, proxyOptions = null) {
-  if (!apiKey) {
-    return { message: "Vercel AI Gateway API key not available." };
-  }
-
-  try {
-    const response = await proxyAwareFetch(
-      VERCEL_AI_GATEWAY_CREDITS_URL,
-      {
-        method: "GET",
-        headers: {
-          Authorization: `Bearer ${apiKey}`,
-          Accept: "application/json",
-        },
-      },
-      proxyOptions,
-    );
-
-    if (response.status === 401 || response.status === 403) {
-      return { message: "Vercel AI Gateway API key invalid or expired." };
-    }
-
-    if (!response.ok) {
-      const errorText = await response.text().catch(() => "");
-      const trimmed = errorText ? `: ${errorText.slice(0, 200)}` : "";
-      return { message: `Vercel AI Gateway credits API error (${response.status})${trimmed}` };
-    }
-
-    const data = await response.json();
-
-    // Vercel returns numeric strings; coerce safely.
-    const balance = Number(data?.balance) || 0;
-    const totalUsed = Number(data?.total_used) || 0;
-
-    // Vercel gives $5/month free credit. The API doesn't return the
-    // monthly allocation so we use the known constant as the denominator.
-    const MONTHLY_CREDIT = 5;
-    const remainingPercentage = (balance / MONTHLY_CREDIT) * 100;
-
-    if (balance <= 0 && totalUsed <= 0) {
-      return {
-        plan: "Pay-as-you-go",
-        message:
-          "Vercel AI Gateway connected. No credit allocation found (BYOK or unfunded account).",
-        quotas: {},
-      };
-    }
-
-    // "Used (USD)": how much has been spent this month (no fixed cap → unlimited).
-    // "Remaining (USD)": balance remaining out of the $5 monthly allocation.
-    return {
-      plan: "Pay-as-you-go",
-      quotas: {
-        "Used (USD)": {
-          used: totalUsed,
-          total: 0,
-          remaining: 0,
-          remainingPercentage: 100,
-          unlimited: true,
-        },
-        "Remaining (USD)": {
-          used: balance,
-          total: MONTHLY_CREDIT,
-          remaining: balance,
-          remainingPercentage,
-          unlimited: false,
-        },
-      },
-    };
-  } catch (error) {
-    return { message: `Vercel AI Gateway error: ${error.message}` };
-  }
-}
-
 export async function getQoderUsage(accessToken, proxyOptions = null) {
   if (!accessToken) {
     return { message: "Qoder usage unavailable: no access token" };

package/open-sse/services/usage/zed.js

@@ -125,11 +125,24 @@ export function parseZedAuthenticatedUserUsage(userInfo) {
   const usage = plan.usage || {};
 
   const editPredictions = usage.edit_predictions || usage.editPredictions;
-  if (editPredictions) {
+  // `{ limited: 0 }` is the token-billed shape, not a zero-cap quota. Emitting a
+  // 0%-remaining row here made the auth pre-filter skip the connection for the
+  // whole billing cycle even though nothing was exhausted.
+  const editPredictionsLimitInfo = editPredictions
+    ? parseZedUsageLimit(
+        editPredictions.limit != null
+          ? editPredictions.limit
+          : usageBucketLimit(editPredictions)?.limit,
+      )
+    : null;
+  const hasEditPredictionsQuota =
+    !!editPredictionsLimitInfo &&
+    (editPredictionsLimitInfo.unlimited || editPredictionsLimitInfo.total > 0);
+  if (hasEditPredictionsQuota) {
     quotas["Edit Predictions"] = makeZedQuotaRow(
       "Edit Predictions",
       editPredictions.used,
-      editPredictions.limit,
+      editPredictions.limit ?? usageBucketLimit(editPredictions)?.limit,
       resetAt,
     );
   }
@@ -155,7 +168,9 @@ export function parseZedAuthenticatedUserUsage(userInfo) {
     isZedTokenBillingModelRequestsLimit(
       modelRequests.limit ?? usageBucketLimit(modelRequests)?.limit,
     )
-      ? "Hosted AI models are billed per token (not request count). Edit Predictions are tracked below. Token spend is on dashboard.zed.dev."
+      ? hasEditPredictionsQuota
+        ? "Hosted AI models are billed per token (not request count). Edit Predictions are tracked below. Token spend is on dashboard.zed.dev."
+        : "Hosted AI models are billed per token (not request count). Token spend is on dashboard.zed.dev."
       : null;
 
   let planLabel = formatZedPlanLabel(planId);

package/open-sse/services/usage.js

@@ -12,23 +12,16 @@ import {
 } from "./usage/codex.js";
 
 export { consumeCodexRateLimitResetCredit, getCodexRateLimitResetCredits };
-import { getMiniMaxUsage } from "./usage/minimax.js";
 import { getCodeBuddyCnUsage, getCodeBuddyIntlUsage } from "./usage/codebuddy-cn.js";
 import { getGrokCliUsage } from "./usage/grok-cli.js";
 import { getDeepseekUsage } from "./usage/deepseek.js";
 import { getOpenCodeGoUsage } from "./usage/opencode-go.js";
-import { getGroqUsage } from "./usage/groq.js";
 import { getZedUsage } from "./usage/zed.js";
 import { getXiaomiMimoUsage } from "./usage/xiaomi-mimo.js";
 import { resolveQoderCredentials } from "./qoderModels.js";
 import { getGlmUsage } from "./usage/glm.js";
 import { getCommandCodeUsage } from "./usage/commandcode.js";
-import {
-  getIflowUsage,
-  getOllamaUsage,
-  getVercelAiGatewayUsage,
-  getQoderUsage,
-} from "./usage/misc.js";
+import { getIflowUsage, getOllamaUsage, getQoderUsage } from "./usage/misc.js";
 
 /**
  * Get usage data for a provider connection
@@ -49,11 +42,7 @@ const USAGE_HANDLERS = {
   },
   iflow: (c) => getIflowUsage(c.accessToken),
   ollama: (c) => getOllamaUsage(c.apiKey, c.providerSpecificData, c.proxyOptions),
-  glm: (c) => getGlmUsage(c.apiKey, c.provider, c.proxyOptions),
-  "glm-cn": (c) => getGlmUsage(c.apiKey, c.provider, c.proxyOptions),
-  minimax: (c) => getMiniMaxUsage(c.apiKey, c.provider, c.proxyOptions),
-  "minimax-cn": (c) => getMiniMaxUsage(c.apiKey, c.provider, c.proxyOptions),
-  "vercel-ai-gateway": (c) => getVercelAiGatewayUsage(c.apiKey, c.proxyOptions),
+  glm: (c) => getGlmUsage(c.apiKey, c.proxyOptions),
   "codebuddy-cn": (c) =>
     getCodeBuddyCnUsage(c.accessToken, c.apiKey, c.providerSpecificData, c.proxyOptions),
   "codebuddy-intl": (c) =>
@@ -61,7 +50,6 @@ const USAGE_HANDLERS = {
   "grok-cli": (c) => getGrokCliUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
   "opencode-go": (c) => getOpenCodeGoUsage(c.apiKey, c.proxyOptions),
   deepseek: (c) => getDeepseekUsage(c.apiKey, c.proxyOptions),
-  groq: (c) => getGroqUsage(c.apiKey, c.proxyOptions),
   zed: (c) => getZedUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
   "xiaomi-mimo": (c) => getXiaomiMimoUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
   commandcode: (c) => getCommandCodeUsage(c.apiKey, c.proxyOptions),

package/open-sse/shared/zedAuth.js

@@ -11,7 +11,11 @@
 
 import crypto from "node:crypto";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
-import { DEFAULT_RETRY_CONFIG, resolveRetryEntry } from "../config/runtimeConfig.js";
+import {
+  DEFAULT_RETRY_CONFIG,
+  FETCH_CONNECT_TIMEOUT_MS,
+  resolveRetryEntry,
+} from "../config/runtimeConfig.js";
 
 export const ZED_WEB_BASE_URL = "https://zed.dev";
 export const ZED_CLOUD_BASE_URL = "https://cloud.zed.dev";
@@ -208,20 +212,61 @@ function isTransientNetError(error) {
   return /socket hang up|ECONNRESET|other side closed/i.test(String(error.message || ""));
 }
 
-async function zedFetchWithRetry(url, options, proxyOptions = null, statusKey = 502) {
+// Abort an upstream call whose response headers never arrive. A hung socket
+// otherwise blocks the request until the client gives up (Zed rides the same
+// BaseExecutor pipeline, whose executors all bound the connect phase).
+function withConnectTimeout(options, connectTimeoutMs) {
+  const { signal, ...rest } = options || {};
+  const timeoutCtrl = new AbortController();
+  const timer = setTimeout(
+    () => timeoutCtrl.abort(new Error("fetch connect timeout")),
+    connectTimeoutMs,
+  );
+  const merged = signal ? AbortSignal.any([signal, timeoutCtrl.signal]) : timeoutCtrl.signal;
+  return {
+    options: { ...rest, signal: merged },
+    clear: () => clearTimeout(timer),
+    timedOut: () => timeoutCtrl.signal.aborted,
+  };
+}
+
+async function zedFetchWithRetry(
+  url,
+  options,
+  proxyOptions = null,
+  { statusKey = 502, connectTimeoutMs = FETCH_CONNECT_TIMEOUT_MS } = {},
+) {
   const { attempts, delayMs } = resolveRetryEntry(DEFAULT_RETRY_CONFIG[statusKey]);
   for (let attempt = 0; ; attempt++) {
+    const attemptOptions = withConnectTimeout(options, connectTimeoutMs);
     try {
-      return await proxyAwareFetch(url, options, proxyOptions);
+      return await proxyAwareFetch(url, attemptOptions.options, proxyOptions);
     } catch (error) {
+      // A connect timeout is an internal deadline, not a caller abort: treat it
+      // like a transient socket failure so the retry/fallback path can act.
+      if (attemptOptions.timedOut()) {
+        if (attempt >= attempts) {
+          throw new Error(`Zed request timed out after ${connectTimeoutMs}ms (no response)`);
+        }
+        await sleep(delayMs);
+        continue;
+      }
       if (!isTransientNetError(error) || attempt >= attempts) throw error;
       await sleep(delayMs);
+    } finally {
+      attemptOptions.clear();
     }
   }
 }
 
-async function fetchJson(url, options, proxyOptions = null) {
-  const res = await zedFetchWithRetry(url, options, proxyOptions);
+/** Only override the connect deadline when the caller set one (a bare
+ * `undefined` would reach setTimeout and fire immediately). */
+function timeoutOption(connectTimeoutMs) {
+  return connectTimeoutMs == null ? {} : { connectTimeoutMs };
+}
+
+async function fetchJson(url, options, proxyOptions = null, timeouts = {}) {
+  const res = await zedFetchWithRetry(url, options, proxyOptions, timeouts);
   const text = await res.text();
   let data = null;
   if (text) {
@@ -259,6 +304,7 @@ export async function fetchZedAuthenticatedUser(credentials, options = {}) {
       signal: options.signal ?? undefined,
     },
     options.proxyOptions ?? null,
+    timeoutOption(options.connectTimeoutMs),
   );
 }
 
@@ -329,6 +375,7 @@ export async function fetchZedLlmToken(credentials, options = {}) {
       signal: options.signal ?? undefined,
     },
     options.proxyOptions ?? null,
+    timeoutOption(options.connectTimeoutMs),
   );
   const token =
     typeof data?.token === "string" ? data.token : data?.token?.[0] || data?.token?.value;
@@ -361,6 +408,7 @@ export async function zedLlmFetch(credentials, path, options = {}) {
         signal: options.signal ?? undefined,
       },
       options.proxyOptions ?? null,
+      { connectTimeoutMs: options.connectTimeoutMs },
     );
   };