feat: import upstream v3.8.51 retained-core routing, translation, and security updates

.env.example

@@ -1302,6 +1302,11 @@ CURSOR_USER_AGENT="Cursor/3.4"
 # Override the advertised GitHub Copilot CLI version independently of
 # GITHUB_USER_AGENT. Used by: open-sse/config/providerHeaderProfiles.ts.
 # GITHUB_COPILOT_CLI_VERSION=1.0.82
+#
+# Override the advertised Grok Build (grok-cli) client version independently
+# of the full User-Agent string. xAI enforces minimum client version gates
+# (HTTP 426). Used by: open-sse/config/grokBuild.ts.
+# GROK_CLI_CLIENT_VERSION=1.0.44
 
 # Kill-switch to strip non-standard `codex.*` SSE events (e.g. codex.rate_limits)
 # from the Codex Responses stream. These frames break the OpenAI SDK's
@@ -1601,6 +1606,7 @@ CURSOR_USER_AGENT="Cursor/3.4"
 
 # ── TLS client (wreq-js fingerprint proxy) ──
 # TLS_CLIENT_TIMEOUT_MS=600000       # Inherits from FETCH_TIMEOUT_MS by default
+# TLS_FIRST_BYTE_WATCHDOG_MS=10000   # #12656: bounds time-to-first-byte on the wreq body (0 disables)
 
 # ── API Bridge (/v1 proxy server) ──
 # API_BRIDGE_PROXY_TIMEOUT_MS=600000             # Proxy hop timeout (default: 10min)

changelog.d/fixes/v1beta-gemini-format-detection.md

@@ -0,0 +1 @@
+- **fix(api):** `/v1beta` Gemini requests with `generationConfig.maxOutputTokens` now return properly shaped Gemini replies ([#14165](https://github.com/diegosouzapw/OmniRoute/issues/14165)). The ingress converts the body to OpenAI chat format before re-entering `handleChat` while the URL keeps its `/v1beta` path, and with no path branch `detectFormat`'s `max_tokens` heuristic misread the converted body as `claude` — non-streaming callers got an anthropic-shaped body the route's OpenAI→Gemini converter silently dropped, and streaming callers got a 200 SSE response with zero bytes, looping the antigravity CLI and the Google GenAI SDK. `detectFormatFromEndpoint` now treats a `/v1beta` path as OpenAI chat unless the body still carries the raw Gemini `contents` envelope. Raw Gemini bodies (client-raw-request contexts) are unaffected.

docs/reference/ENVIRONMENT.md

@@ -621,6 +621,7 @@ process.env[`${PROVIDER_ID}_USER_AGENT`]
 | `CODEX_CLIENT_VERSION`           | `0.131.0`                                     | Override Codex client version independently of full UA string                                    |
 | `CLAUDE_CODE_CLIENT_VERSION`     | `2.1.258`                                     | Override advertised Claude Code version independently of `CLAUDE_USER_AGENT`. Anthropic gates some models on this value (#12417). |
 | `GITHUB_COPILOT_CLI_VERSION`     | `1.0.81-6`                                    | Override advertised Copilot CLI version independently of `GITHUB_USER_AGENT`                     |
+| `GROK_CLI_CLIENT_VERSION`        | `1.0.44`                                      | Override advertised Grok Build (grok-cli) client version independently of the full User-Agent string. xAI enforces minimum client version gates (HTTP 426). Values not matching `^[A-Za-z0-9][A-Za-z0-9._-]{0,31}$` are ignored (`open-sse/config/grokBuild.ts`). |
 | `GITHUB_USER_AGENT`              | `GitHubCopilotChat/0.54.0`                    | When GitHub Copilot Chat updates                                                                 |
 | `ANTIGRAVITY_USER_AGENT`         | `antigravity/2.0.1 darwin/arm64`              | When Antigravity IDE updates                                                                     |
 | `KIRO_USER_AGENT`                | `AWS-SDK-JS/3.0.0 kiro-ide/1.0.0`             | When Kiro IDE updates                                                                            |
@@ -749,6 +750,7 @@ REQUEST_TIMEOUT_MS (global override)
 | `FETCH_CONNECT_TIMEOUT_MS`                | `30000`              | TCP connection establishment timeout.                                                                                                                           |
 | `FETCH_KEEPALIVE_TIMEOUT_MS`              | `4000`               | Keep-alive socket idle timeout.                                                                                                                                 |
 | `TLS_CLIENT_TIMEOUT_MS`                   | = `FETCH_TIMEOUT_MS` | TLS fingerprint proxy (wreq-js) timeout.                                                                                                                        |
+| `TLS_FIRST_BYTE_WATCHDOG_MS`              | `10000`              | Time-to-first-byte bound (ms) on the wreq-js TLS fingerprint transport body, so a wedged body fails fast instead of riding `TLS_CLIENT_TIMEOUT_MS` (#12656). Set `0` to disable; invalid values fall back to the default. |
 | `API_BRIDGE_PROXY_TIMEOUT_MS`             | `30000`              | Proxy hop timeout for `/v1` bridge requests.                                                                                                                    |
 | `FIRECRAWL_BASE_URL`                      | `https://api.firecrawl.dev` | Point the Firecrawl web-fetch executor at a self-hosted instance (API key optional off-cloud).                                                          |
 | `FIRECRAWL_TIMEOUT_MS`                    | `30000`              | Per-request timeout for the Firecrawl web-fetch executor.                                                                                                       |

open-sse/config/codexClient.ts

@@ -3,6 +3,7 @@ import {
   DEFAULT_CODEX_CLIENT_VERSION,
   getCodexCliRsHeaders as buildCodexCliRsHeaders,
 } from "@/shared/constants/codexClient";
+import { getActiveClientVersion } from "@/lib/client-versions/registry";
 
 export {
   DEFAULT_CODEX_CLIENT_VERSION,
@@ -27,6 +28,8 @@ function getSafeEnvValue(name: string, pattern: RegExp): string | null {
 }
 
 export function getCodexClientVersion(): string {
+  const dynamic = getActiveClientVersion("codex");
+  if (dynamic) return dynamic;
   return (
     getSafeEnvValue(CODEX_VERSION_OVERRIDE_ENV, SAFE_HEADER_TOKEN_PATTERN) ||
     DEFAULT_CODEX_CLIENT_VERSION

open-sse/config/codexIdentity.ts

@@ -570,3 +570,37 @@ export function isVerifiedNativeCodexRequest(
 ): boolean {
   return isCodexOriginatedHeaders(headers) && hasNativeCodexTurnBinding(body);
 }
+
+/**
+ * Detect the Claude Code CLI as the request *client* from request headers.
+ * Used to auto-enable model echo so session restores work when the resolved
+ * upstream model (e.g. `oc/nemotron-3-ultra-free`) is not recognized by the
+ * Claude Code client on `--resume`.
+ */
+export function isClaudeCodeOriginatedHeaders(
+  headers: Headers | Record<string, unknown> | null | undefined
+): boolean {
+  const getHeader = (name: string): string => {
+    if (headers instanceof Headers) {
+      return headers.get(name)?.toLowerCase() ?? "";
+    }
+    if (headers && typeof headers === "object") {
+      for (const [key, value] of Object.entries(headers as Record<string, unknown>)) {
+        if (key.toLowerCase() === name && typeof value === "string") {
+          return value.toLowerCase();
+        }
+      }
+    }
+    return "";
+  };
+
+  // Claude Code identifies itself via the user-agent header
+  const userAgent = getHeader("user-agent");
+  if (userAgent.includes("claude-code") || userAgent.includes("anthropic-ai/claude-code")) {
+    return true;
+  }
+  // Also check originator if present
+  const originator = getHeader("originator");
+  if (originator.startsWith("claude-code")) return true;
+  return false;
+}

open-sse/config/grokBuild.ts

@@ -8,7 +8,7 @@ export const GROK_BUILD_OAUTH_ISSUER = "https://auth.x.ai";
 export const GROK_BUILD_DEVICE_CODE_URL = `${GROK_BUILD_OAUTH_ISSUER}/oauth2/device/code`;
 export const GROK_BUILD_TOKEN_URL = `${GROK_BUILD_OAUTH_ISSUER}/oauth2/token`;
 
-export const GROK_BUILD_DEFAULT_CLIENT_VERSION = "0.2.106";
+export const GROK_BUILD_DEFAULT_CLIENT_VERSION = "1.0.44";
 export const GROK_BUILD_DEFAULT_CONTEXT_WINDOW = 256_000;
 export const GROK_BUILD_DEFAULT_REASONING_EFFORT = "high";
 export const GROK_BUILD_SUPPORTED_REASONING_EFFORTS = Object.freeze(["low", "medium", "high"]);
@@ -17,6 +17,9 @@ export const GROK_BUILD_TOKEN_AUTH = "xai-grok-cli";
 export const GROK_BUILD_REASONING_INCLUDE = "reasoning.encrypted_content";
 export const GROK_BUILD_OAUTH_REFERRER = "grok-build";
 
+const GROK_CLI_VERSION_OVERRIDE_ENV = "GROK_CLI_CLIENT_VERSION";
+const SAFE_HEADER_TOKEN_PATTERN = /^[A-Za-z0-9][A-Za-z0-9._-]{0,31}$/;
+
 export const GROK_BUILD_OAUTH_SCOPES = Object.freeze([
   "openid",
   "profile",
@@ -62,7 +65,21 @@ function mapArch(arch: string): string {
   return arch;
 }
 
+function getSafeEnvValue(name: string, pattern: RegExp): string | null {
+  const raw = process.env[name];
+  if (typeof raw !== "string") return null;
+  const normalized = raw.trim();
+  if (!normalized || !pattern.test(normalized)) {
+    return null;
+  }
+  return normalized;
+}
+
 export function getGrokBuildClientVersion(): string {
+  const envOverride = getSafeEnvValue(GROK_CLI_VERSION_OVERRIDE_ENV, SAFE_HEADER_TOKEN_PATTERN);
+  if (envOverride) {
+    return envOverride;
+  }
   return GROK_BUILD_DEFAULT_CLIENT_VERSION;
 }
 

open-sse/executors/codex.ts

@@ -59,6 +59,7 @@ import {
 } from "./codex/reasoningSuffix.ts";
 import { repairMissingCodexToolCallOutputs } from "./codex/toolCallRepair.ts";
 import { resolveAppServerConfig } from "./codex/appServerConfig.ts";
+import { getResponsesSubpath } from "./codex/responsesSubpath.ts";
 import { CodexAppServerExecutor } from "./codex-app-server.ts";
 // Re-exported for external importers (tests + provider services).
 export { isCodexFreePlan, normalizeCodexTools } from "./codex/tools.ts";
@@ -287,30 +288,6 @@ function stripOrphanedCodexFunctionCallOutputs(body: Record<string, unknown>): v
   }
 }
 
-function getResponsesSubpath(endpointPath: unknown): string | null {
-  let normalizedEndpoint = String(endpointPath || "");
-  while (normalizedEndpoint.endsWith("/") && normalizedEndpoint.length > 0) {
-    normalizedEndpoint = normalizedEndpoint.slice(0, -1);
-  }
-
-  const lower = normalizedEndpoint.toLowerCase();
-  if (lower === "responses" || lower.endsWith("/responses")) {
-    return "";
-  }
-
-  const responsesSlash = "/responses/";
-  const idx = lower.lastIndexOf(responsesSlash);
-  if (idx !== -1) {
-    return normalizedEndpoint.slice(idx + "/responses".length);
-  }
-
-  if (lower.startsWith("responses/")) {
-    return normalizedEndpoint.slice("responses".length);
-  }
-
-  return null;
-}
-
 export function isCompactResponsesEndpoint(endpointPath: unknown): boolean {
   return getResponsesSubpath(endpointPath)?.toLowerCase() === "/compact";
 }

open-sse/executors/codex/responsesSubpath.ts

@@ -0,0 +1,45 @@
+/**
+ * Extracts the `/v1/responses/<subpath>` suffix the Codex executor forwards upstream
+ * (e.g. `/compact`, `/<id>/cancel`). Returns "" for the plain endpoint and null when the
+ * path is not a Responses path or the subpath is unsafe to append.
+ */
+
+// The subpath comes from the request URL and is appended to the upstream URL verbatim, so
+// anything the upstream (or fetch's URL parser) would read as path traversal or as the end of
+// the path is refused. The caller then falls back to the plain /responses endpoint.
+const UNSAFE_SUBPATH_ESCAPE = /%(?:2e|2f|5c|23|3f|00)/i;
+
+function isSafeResponsesSubpath(subpath: string): boolean {
+  if (subpath === "") return true;
+  if (/[\\?#\u0000]/.test(subpath) || UNSAFE_SUBPATH_ESCAPE.test(subpath)) return false;
+  return !subpath.split("/").some((segment) => segment === "." || segment === "..");
+}
+
+export function getResponsesSubpath(endpointPath: unknown): string | null {
+  const subpath = findResponsesSubpath(endpointPath);
+  return subpath !== null && isSafeResponsesSubpath(subpath) ? subpath : null;
+}
+
+function findResponsesSubpath(endpointPath: unknown): string | null {
+  let normalizedEndpoint = String(endpointPath || "");
+  while (normalizedEndpoint.endsWith("/") && normalizedEndpoint.length > 0) {
+    normalizedEndpoint = normalizedEndpoint.slice(0, -1);
+  }
+
+  const lower = normalizedEndpoint.toLowerCase();
+  if (lower === "responses" || lower.endsWith("/responses")) {
+    return "";
+  }
+
+  const responsesSlash = "/responses/";
+  const idx = lower.lastIndexOf(responsesSlash);
+  if (idx !== -1) {
+    return normalizedEndpoint.slice(idx + "/responses".length);
+  }
+
+  if (lower.startsWith("responses/")) {
+    return normalizedEndpoint.slice("responses".length);
+  }
+
+  return null;
+}

open-sse/executors/geminiCli.ts

@@ -19,6 +19,7 @@ import {
   reassembleGeminiCliChunks,
   type GeminiCliResponseAccumulator,
 } from "../translator/response/geminiCli.ts";
+import { getActiveClientVersion } from "@/lib/client-versions/registry";
 
 export const GEMINI_CLI_ENDPOINT_FALLBACKS = [
   "https://cloudcode-pa.googleapis.com/v1internal",
@@ -75,6 +76,7 @@ export function buildGeminiCliHeaders(
 ): Record<string, string> {
   const uaVer =
     env.uaVersion ||
+    getActiveClientVersion("gemini-cli") ||
     process.env.GEMINI_CLI_UA_VERSION ||
     process.env.GEMINI_CLI_CLIENT_VERSION ||
     GEMINI_CLI_UA_VERSION;

open-sse/executors/grok-web/tool-bridge.ts

@@ -1,5 +1,6 @@
 // OpenAI <-> Grok tool-call translation (pure). Extracted verbatim from grok-web.ts.
 import type { GrokStreamResponse } from "./types.ts";
+import { findTagBlocks } from "../../utils/tagBlocks.ts";
 
 // ─── OpenAI message → Grok query translation ───────────────────────────────
 
@@ -32,12 +33,29 @@ export interface ToolBridgeContext {
   lastUserText: string;
 }
 
+/**
+ * Where a reminder block starts once the `---` separator line the client put in front of it is
+ * counted: `\n?---`, blanks, a newline and any further whitespace, right before the tag.
+ */
+function reminderBlockStart(text: string, floor: number, tagStart: number): number {
+  let runStart = tagStart;
+  while (runStart > floor && /\s/.test(text[runStart - 1])) runStart -= 1;
+  if (!text.slice(runStart, tagStart).includes("\n")) return tagStart;
+  if (runStart - 3 < floor || text.slice(runStart - 3, runStart) !== "---") return tagStart;
+  const separatorStart = runStart - 3;
+  return separatorStart > floor && text[separatorStart - 1] === "\n"
+    ? separatorStart - 1
+    : separatorStart;
+}
+
 export function stripInjectedRuntimeReminders(text: string): string {
-  return text
-    .replace(/\n?---\s*\n\s*<internal_reminder>[\s\S]*?<\/internal_reminder>/gi, "")
-    .replace(/<internal_reminder>[\s\S]*?<\/internal_reminder>/gi, "")
-    .replace(/\n{3,}/g, "\n\n")
-    .trim();
+  let kept = "";
+  let position = 0;
+  for (const block of findTagBlocks(text, /<internal_reminder>/gi, /<\/internal_reminder>/gi)) {
+    kept += text.slice(position, reminderBlockStart(text, position, block.start));
+    position = block.end;
+  }
+  return (kept + text.slice(position)).replace(/\n{3,}/g, "\n\n").trim();
 }
 
 export function extractTextContent(msg: Record<string, unknown>): string {
@@ -629,11 +647,10 @@ export function parseClientToolCallMarkup(
 ): OpenAIToolCall[] | null {
   if (!toolRegistry.enabled || !text.includes("<tool_call>")) return null;
   const calls: OpenAIToolCall[] = [];
-  const re = /<tool_call>\s*([\s\S]*?)\s*<\/tool_call>/g;
-  for (const match of text.matchAll(re)) {
+  for (const block of findTagBlocks(text, /<tool_call>/g, /<\/tool_call>/g)) {
     let parsed: unknown;
     try {
-      parsed = JSON.parse(match[1]);
+      parsed = JSON.parse(block.inner.trim());
     } catch {
       continue;
     }

open-sse/executors/trae.ts

@@ -20,6 +20,7 @@ import { BaseExecutor, mergeUpstreamExtraHeaders } from "./base.ts";
 import { PROVIDERS } from "../config/constants.ts";
 import { sanitizeErrorMessage } from "../utils/error.ts";
 import { resolvePublicCred } from "../utils/publicCreds.ts";
+import { resolveTraeApiHost } from "../utils/traeHost.ts";
 
 type JsonRecord = Record<string, unknown>;
 type ChatMessage = { role?: string; content?: unknown };
@@ -438,7 +439,7 @@ export class TraeExecutor extends BaseExecutor {
     const psd = (credentials?.providerSpecificData as JsonRecord) || {};
     const refreshToken = credentials?.refreshToken as string | undefined;
     if (!refreshToken) return null;
-    const host = ((psd.host as string) || "https://api-us-east.trae.ai").replace(/\/$/, "");
+    const host = resolveTraeApiHost(psd.host);
     const clientId =
       (psd.clientId as string) || resolvePublicCred("trae_id", "TRAE_OAUTH_CLIENT_ID");
     const url = `${host}/cloudide/api/v3/trae/oauth/ExchangeToken`;