From 9aa0c0e6c0a56b0ab649dc563f7a31a62dc88de8 Mon Sep 17 00:00:00 2001
From: Raden Salman Alfaridzi <radensalman@mhs.pelitabangsa.ac.id>
Date: Wed, 22 Jul 2026 01:00:39 +0700
Subject: [PATCH] feat(v0.7.4): C1 auth bypass fix,
 Forge/TokenRouter/Qwen/Alibaba/hcnsec providers, GitHub Copilot /v1/messages,
 opencode effort tiers, xAI quota handler, ModelAccessModal crash fix

Features:
- New providers: Forge Workspace (33 models, 3 tiers), TokenRouter (with
  management key + Quota Tracker), Huancheng (hcnsec), Qwen Cloud + Qwen Cloud
  Token Plan, Alibaba + Alibaba CN
- GitHub Copilot native /v1/messages routing for Claude models
  (targetFormat:claude, anthropic-version header)
- GitHub Copilot model catalog synced to OmniRoute fea1d54 (20 models)
- opencode-go effort-tier aliases (9 models) + EFFORT_TIERS table
  (port of OmniRoute 1843b34)
- Qwen3.8 Max Preview on Qwen Web (Subscription) with REQUIRED_THINKING_MODELS

Fixes:
- C1 Critical: dashboardGuard method-based auth routing (mutations always
  require JWT/CLI token even when requireLogin=false)
- Cline/GLM 500 stream_options (inject stream_options on streaming requests)
- HuggingChat HTML 200 (DEFAULT_MODEL omni + preprompt + native fetch)
- xAI Quota Tracker empty card (billing API removed; show informative message)
- ModelAccessModal crash (Cannot read properties of null (reading 'length'))
- TokenRouter quota null plan crash
- Playground value/key/object Object crashes (3 separate fixes)

Improvements:
- SanitizeHtml utility + unit tests
- Playground MessageContent markdown renderer
- AddApiKeyModal TokenRouter management key field
- 8 new provider SVG icons

Bumps package.json + cli/package.json to 0.7.4
---
 CHANGELOG.md                                  |  30 ++++
 cli/package.json                              |   2 +-
 open-sse/executors/base.js                    |   2 +-
 open-sse/executors/github.js                  |  73 +++++++--
 open-sse/executors/huggingchat.js             |  35 ++++-
 open-sse/executors/opencode-go.js             |  71 ++++++++-
 open-sse/executors/qwen-web.js                |  73 ++++++++-
 open-sse/providers/capabilities.js            |   2 +
 open-sse/providers/pricing.js                 |  36 +++++
 open-sse/providers/registry/alibaba-cn.js     |  50 +++++++
 open-sse/providers/registry/alibaba.js        |  52 +++++++
 open-sse/providers/registry/forge.js          |  97 ++++++++++++
 open-sse/providers/registry/github.js         |  55 ++++---
 open-sse/providers/registry/hcnsec.js         |  47 ++++++
 open-sse/providers/registry/huggingchat.js    |   5 +
 open-sse/providers/registry/index.js          |  16 +-
 open-sse/providers/registry/opencode-go.js    |  15 ++
 .../registry/qwen-cloud-token-plan.js         |  76 ++++++++++
 open-sse/providers/registry/qwen-cloud.js     |  55 +++++++
 open-sse/providers/registry/qwen-web.js       |  30 +++-
 open-sse/providers/registry/tokenrouter.js    |  55 +++++++
 open-sse/providers/registry/xai.js            |   6 +-
 open-sse/services/usage.js                    |   2 +
 open-sse/services/usage/tokenrouter.js        | 116 +++++++++++++++
 open-sse/services/usage/xai.js                | 140 +++++++-----------
 package.json                                  |   4 +-
 public/providers/alibaba-cn.svg               |  11 ++
 public/providers/alibaba.svg                  |  11 ++
 public/providers/forge.svg                    |   9 ++
 public/providers/hcnsec.svg                   |  10 ++
 public/providers/qwen-cloud-token-plan.svg    |   8 +
 public/providers/qwen-cloud.svg               |   8 +
 public/providers/tokenrouter.svg              |  15 ++
 .../dashboard/endpoint/EndpointPageClient.js  |  12 +-
 .../providers/[id]/AddApiKeyModal.js          |  23 +++
 .../quota/components/ProviderLimits/index.js  |  11 +-
 .../quota/components/ProviderLimits/utils.js  |  18 +++
 src/dashboardGuard.js                         |  21 +++
 src/shared/utils/providerIcon.js              |   3 +-
 src/shared/utils/sanitizeHtml.js              |  77 ++++++----
 tests/unit/sanitize-html.test.js              |  90 +++++++++++
 41 files changed, 1289 insertions(+), 183 deletions(-)
 create mode 100644 open-sse/providers/registry/alibaba-cn.js
 create mode 100644 open-sse/providers/registry/alibaba.js
 create mode 100644 open-sse/providers/registry/forge.js
 create mode 100644 open-sse/providers/registry/hcnsec.js
 create mode 100644 open-sse/providers/registry/qwen-cloud-token-plan.js
 create mode 100644 open-sse/providers/registry/qwen-cloud.js
 create mode 100644 open-sse/providers/registry/tokenrouter.js
 create mode 100644 open-sse/services/usage/tokenrouter.js
 create mode 100644 public/providers/alibaba-cn.svg
 create mode 100644 public/providers/alibaba.svg
 create mode 100644 public/providers/forge.svg
 create mode 100644 public/providers/hcnsec.svg
 create mode 100644 public/providers/qwen-cloud-token-plan.svg
 create mode 100644 public/providers/qwen-cloud.svg
 create mode 100644 public/providers/tokenrouter.svg
 create mode 100644 tests/unit/sanitize-html.test.js

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 69124a9..2853a66 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,33 @@
+# v0.7.4 (2026-07-19)
+
+## Features
+- **Forge Workspace provider**: new API-key provider (forge) with 33 models across 3 pricing tiers (free/pro/enterprise), live model discovery via modelsFetcher, dedicated pricing block, and SVG brand icon.
+- **TokenRouter provider + Quota Tracker**: new API-key provider (tokenrouter) with a separate Management API key (mirrors TokenRouter's two-credential design). Quota card surfaces Wallet / Top-up / Voucher balances from the management endpoint. Handler returns a graceful `message` when no management key is configured (no more UI crash).
+- **Huancheng Public API (hcnsec) provider**: new OpenAI-compatible regional API-key provider.
+- **Qwen Cloud + Qwen Cloud Token Plan providers**: two new dedicated API-key providers for Alibaba Cloud's Bailian Qwen endpoints (pay-as-you-go + token-plan variants).
+- **Alibaba + Alibaba CN providers**: regional Alibaba DashScope API-key providers (international + China mainland base URLs).
+- **GitHub Copilot native /v1/messages routing**: Claude models routed directly to GitHub's `/messages` endpoint (targetFormat:"claude") with `anthropic-version:2023-06-01` header, bypassing Chat Completions quirks and `sanitizeMessagesForChatCompletions`. Result: native Anthropic-format responses for Claude Code, Cline, etc.
+- **GitHub Copilot model catalog sync**: synced to OmniRoute fea1d54 (20 models, including the latest Claude 4.5/4.10 family).
+- **opencode-go effort-tier aliases**: 9 new alias models (`glm-5.2-high/max`, `mimo-v2.5-high/max`, `deepseek-v4-pro-low/medium/high/max`) auto-rewriting model id + injecting `reasoning_effort` via a new EFFORT_TIERS table + `parseEffortLevel()` suffix parser (port of OmniRoute commit 1843b34).
+- **Qwen3.8 Max Preview model**: added to Qwen Web (Subscription) catalog. Auto-enables thinking mode via `REQUIRED_THINKING_MODELS` set; SSE parser handles `thinking_summary` via `delta.extra.summary_thought.content[]`.
+
+## Fixes
+- **C1 Critical — Dashboard auth bypass**: `dashboardGuard.js` now uses method-based routing. Mutation routes (POST/PUT/PATCH/DELETE) **always** require a real JWT/CLI token even when `requireLogin=false`. Previously anyone with network access to the dashboard port could mutate state (create/delete providers, keys, settings) without authentication.
+- **Cline/GLM 500 `stream_options` error**: `DefaultExecutor.transformRequest` now injects `stream_options: { include_usage: true }` on streaming requests so providers that require usage-on-stream don't 500. Signature expanded to `(model, body, stream, credentials)`.
+- **HuggingChat provider dead (HTML 200 response)**: `zai-org/GLM-5.2` was retired by HF. Switched `DEFAULT_MODEL` to `omni` (HF's auto-router), always send `preprompt: ""`, switched `tlsFetch` → native `proxyAwareFetch` (no longer needs TLS impersonation), and added `isEncryptedCredentialBlob` guard.
+- **xAI Quota Tracker empty card**: xAI removed both `/v1/billing?format=credits` and `/v1/user?include=subscription` (now 404) and migrated billing to a separate Management API on `management-api.x.ai` (requires a different key). Handler rewritten to return a clear `message` explaining where to find usage (`console.x.ai → Billing`) instead of leaving the card blank. No network calls — no more 404 spam.
+- **ModelAccessModal crash (`Cannot read properties of null (reading 'length')`)**: `allowedModels` was `useState(null)` and only synced to an array inside `useEffect` (post-render), so the first render hit `.length` on null. Now uses lazy initializer `useState(() => [...])` so the first render already has a stable array.
+- **TokenRouter quota `Cannot read properties of null (reading 'plan')`**: handler previously returned `null` when no management key was present; UI didn't guard. Fixed on both sides — handler returns an object with a `message` field (never null), and `ProviderLimits/index.js` uses `data ?? {}` for null-safe access.
+- **Playground value/key TypeErrors (3 separate crashes)**: added `valueStr` coercion in ModelPicker, defensive coercion in `useModelCaps.getCaps`, and `model.value || model.name || model.id` extraction in the modal `onSelect` handler (modal passes the whole object, not a string).
+- **Playground `[object Object]` model name after model swap**: extraction fix above also resolved the rendering bug.
+
+## Improvements
+- **SanitizeHtml utility** (`src/shared/utils/sanitizeHtml.js`): regex-based HTML sanitizer for markdown model output. Strips `<script>`/`<iframe>`/`<object>`/`<embed>`/`<form>`/`<style>`, neutralizes `on*` event handlers and `javascript:`/`data:` URLs. Defense-in-depth on top of `marked`'s AST (already constrained).
+- **Playground MessageContent component**: renders assistant content as sanitized markdown, user/error as plain text. Code blocks get monospace styling + copy button.
+- **AddApiKeyModal**: added TokenRouter management-key field UI (paired with chat API key).
+- **Provider icon assets**: added forge, tokenrouter, qwen-cloud, qwen-cloud-token-plan, alibaba, alibaba-cn, hcnsec SVG icons + registered them in `providerIcon.js`'s `SVG_ICON_IDS`.
+- **SanitizeHtml unit tests** (`tests/unit/sanitize-html.test.js`): regression coverage for the sanitizer.
+
 # v0.7.2 (2026-07-18)
 
 ## Features
diff --git a/cli/package.json b/cli/package.json
index 26d2963..a4b40e0 100644
--- a/cli/package.json
+++ b/cli/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@rsalmn/extremerouter",
-  "version": "0.7.3",
+  "version": "0.7.4",
   "description": "ExtremeRouter CLI - AI Gateway control plane",
   "bin": {
     "extremerouter": "./cli.js"
diff --git a/open-sse/executors/base.js b/open-sse/executors/base.js
index 564a116..bff78e3 100644
--- a/open-sse/executors/base.js
+++ b/open-sse/executors/base.js
@@ -136,7 +136,7 @@ export class BaseExecutor {
     for (let urlIndex = 0; urlIndex < fallbackCount; urlIndex++) {
       const url = this.buildUrl(model, stream, urlIndex, credentials);
       const transformedBody = this.transformRequest(model, body, stream, credentials);
-      const headers = this.buildHeaders(credentials, stream);
+      const headers = this.buildHeaders(credentials, stream, model);
 
       if (!retryAttemptsByUrl[urlIndex]) retryAttemptsByUrl[urlIndex] = 0;
 
diff --git a/open-sse/executors/github.js b/open-sse/executors/github.js
index e03b9ef..e998250 100644
--- a/open-sse/executors/github.js
+++ b/open-sse/executors/github.js
@@ -9,6 +9,7 @@ import { parseSSELine, formatSSE } from "../utils/streamHelpers.js";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
 import { stripUnsupportedParams } from "../translator/concerns/paramSupport.js";
 import { SSE_DONE } from "../utils/sseConstants.js";
+import { getModelTargetFormat } from "../config/providerModels.js";
 import crypto from "crypto";
 
 export class GithubExecutor extends BaseExecutor {
@@ -38,12 +39,22 @@ export class GithubExecutor extends BaseExecutor {
   }
 
   buildUrl(model, stream, urlIndex = 0) {
+    // Claude models: route to Copilot's Anthropic-native /v1/messages shim — the
+    // only Copilot endpoint that surfaces prompt-cache token counts for Claude
+    // and avoids a lossy round-trip of tool_use/tool_result/thinking content
+    // blocks through the OpenAI shape. Driven by the registry's per-model
+    // targetFormat (see registry/github.js), which chatCore also uses to
+    // translate the request to Claude shape before the executor ever sees it.
+    // Port of decolua/9router#2608.
+    if (getModelTargetFormat("gh", model) === "claude" && this.config.messagesUrl) {
+      return this.config.messagesUrl;
+    }
     return this.config.baseUrl;
   }
 
-  buildHeaders(credentials, stream = true) {
+  buildHeaders(credentials, stream = true, model = null) {
     const token = credentials.copilotToken || credentials.accessToken;
-    return {
+    const headers = {
       "Authorization": `Bearer ${token}`,
       "Content-Type": "application/json",
       "copilot-integration-id": "vscode-chat",
@@ -57,6 +68,14 @@ export class GithubExecutor extends BaseExecutor {
       "X-Initiator": "user",
       "Accept": stream ? "text/event-stream" : "application/json"
     };
+    // Claude models routed to the Anthropic-native /v1/messages shim require
+    // the anthropic-version header (harmless no-op on /chat/completions and
+    // /responses, but /v1/messages rejects the request without it).
+    // Port of decolua/9router#2608.
+    if (model && getModelTargetFormat("gh", model) === "claude") {
+      headers["anthropic-version"] = "2023-06-01";
+    }
+    return headers;
   }
 
   // Sanitize messages for GitHub Copilot /chat/completions endpoint.
@@ -131,14 +150,26 @@ export class GithubExecutor extends BaseExecutor {
   }
 
   transformRequest(model, body, stream, credentials) {
+    // Claude models arrive here already translated to Anthropic-native shape by
+    // chatCore (registry targetFormat: "claude") and are dispatched at /v1/messages
+    // (buildUrl above), which behaves like the real Anthropic API. The quirks
+    // below (max_tokens→max_completion_tokens rename, reasoning_effort="none"
+    // strip) are /chat/completions-only OpenAI-shape concerns — applying them
+    // to a Claude-shape body would either rename away a valid Anthropic field
+    // or strip thinking config the native endpoint honors. Skip them entirely
+    // for the native path. Port of decolua/9router#2608.
+    const isClaudeNative = getModelTargetFormat("gh", model) === "claude";
+
     const transformed = { ...body };
-    if (this.requiresMaxCompletionTokens(model) && transformed.max_tokens !== undefined) {
-      transformed.max_completion_tokens = transformed.max_tokens;
-      delete transformed.max_tokens;
-    }
-    // "none" means no thinking — strip it so models that don't support "none" don't 400
-    if (transformed.reasoning_effort === "none") {
-      delete transformed.reasoning_effort;
+    if (!isClaudeNative) {
+      if (this.requiresMaxCompletionTokens(model) && transformed.max_tokens !== undefined) {
+        transformed.max_completion_tokens = transformed.max_tokens;
+        delete transformed.max_tokens;
+      }
+      // "none" means no thinking — strip it so models that don't support "none" don't 400
+      if (transformed.reasoning_effort === "none") {
+        delete transformed.reasoning_effort;
+      }
     }
     // Config-driven strip of params unsupported by this provider/model
     stripUnsupportedParams("github", model, transformed);
@@ -165,19 +196,29 @@ export class GithubExecutor extends BaseExecutor {
       return this.executeWithResponsesEndpoint(options);
     }
 
-    // Sanitize messages before sending to /chat/completions
-    // This handles Claude models on GitHub Copilot which reject non-text/image_url content types
-    const sanitizedOptions = {
-      ...options,
-      body: this.sanitizeMessagesForChatCompletions(options.body)
-    };
+    // Claude models with targetFormat: "claude" are routed to the Anthropic-native
+    // /v1/messages shim (buildUrl) and have already been translated to Claude shape
+    // by chatCore. The /chat/completions sanitization below would corrupt native
+    // tool_use/tool_result/thinking content blocks (it serializes them as text),
+    // and the response_format-as-system-prompt workaround is unnecessary because
+    // /v1/messages honors JSON-mode natively. Skip sanitization for the native
+    // path. Port of decolua/9router#2608.
+    const isClaudeNative = getModelTargetFormat("gh", model) === "claude";
+
+    // Sanitize messages before sending to /chat/completions.
+    // This handles Claude models on GitHub Copilot which reject non-text/image_url
+    // content types — only applies to the legacy /chat/completions path now.
+    const sanitizedOptions = isClaudeNative
+      ? options
+      : { ...options, body: this.sanitizeMessagesForChatCompletions(options.body) };
 
     const result = await super.execute({ ...sanitizedOptions, proxyOptions: options.proxyOptions || null });
 
     // Only escalate to /responses for models that endpoint can actually serve.
     // Gemini/Claude would otherwise loop into a misleading "does not support
     // Responses API" 400 instead of surfacing the real /chat/completions error (#1062).
-    if (result.response.status === HTTP_STATUS.BAD_REQUEST && this.supportsResponsesEndpoint(model)) {
+    // Claude native path already has its own endpoint — never escalate.
+    if (!isClaudeNative && result.response.status === HTTP_STATUS.BAD_REQUEST && this.supportsResponsesEndpoint(model)) {
       const errorBody = await result.response.clone().text();
 
       if (errorBody.includes("not accessible via the /chat/completions endpoint") || errorBody.includes("The requested model is not supported")) {
diff --git a/open-sse/executors/huggingchat.js b/open-sse/executors/huggingchat.js
index 5672674..9d16126 100644
--- a/open-sse/executors/huggingchat.js
+++ b/open-sse/executors/huggingchat.js
@@ -3,7 +3,9 @@ import { PROVIDERS } from "../config/providers.js";
 import { SSE_DONE, SSE_HEADERS_NO_BUFFER } from "../utils/sseConstants.js";
 import { sseChunk } from "../utils/sse.js";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
-import { tlsFetch } from "../utils/tlsClient.js";
+// NOTE: tlsFetch is intentionally NOT imported here. HuggingFace sits behind
+// CloudFront + AWS WAF which fingerprints TLS — the wreq-js impersonation
+// signature triggers the WAF (causing HTML login redirects). Use native fetch.
 
 // HuggingChatExecutor — HuggingChat (huggingface.co/chat) Web Provider.
 //
@@ -33,7 +35,11 @@ const CONVERSATION_URL = `${HUGGINGFACE_BASE}/chat/conversation`;
 const API_CONVERSATIONS_URL = `${HUGGINGFACE_BASE}/chat/api/v2/conversations`;
 
 const DEFAULT_COOKIE_NAME = "hf-chat";
-const DEFAULT_MODEL = "baidu/ERNIE-4.5-VL-424B-A47B-Base-PT";
+// Default to "omni" — HF's auto-router. Individual model ids get retired by
+// HuggingFace without notice, which causes /chat/conversation to fall back to
+// serving the HTML SPA shell instead of the JSON API response (200+HTML, no
+// conversationId). The omni router is stable and always available.
+const DEFAULT_MODEL = "omni";
 
 const USER_AGENT =
   "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/149.0.0.0 Safari/537.36";
@@ -206,7 +212,12 @@ function extractInitialParentMessageId(value) {
 }
 
 async function fetchInitialParentMessageId(conversationId, headers, proxyOptions, signal) {
-  const res = await tlsFetch(
+  // Use proxyAwareFetch (native Node fetch) — NOT tlsFetch. HuggingFace sits
+  // behind CloudFront + AWS WAF which fingerprints the TLS layer; the wreq-js
+  // impersonation signature actually TRIGGERS the WAF (it doesn't match the
+  // real navigator/UA behaviour), causing HF to redirect to the HTML login SPA
+  // instead of returning JSON. Native fetch passes cleanly. Mirrors OmniRoute.
+  const res = await proxyAwareFetch(
     `${API_CONVERSATIONS_URL}/${conversationId}`,
     { method: "GET", headers, signal },
     proxyOptions
@@ -546,10 +557,18 @@ export class HuggingChatExecutor extends BaseExecutor {
     // -- Step 1: Create conversation ----------------------------------------
     let conversationId;
     try {
-      const createBody = { model: resolvedModel };
-      if (systemPrompt) createBody.preprompt = systemPrompt;
+      // Build the create-conversation body. HuggingFace's SPA always sends
+      // `preprompt` (even as "") — omitting it has been observed to make the
+      // upstream fall back to serving the HTML SPA shell instead of the JSON
+      // API response. Always include the field.
+      const createBody = {
+        model: resolvedModel,
+        preprompt: systemPrompt || "",
+      };
 
-      const createRes = await tlsFetch(
+      // Use proxyAwareFetch (native Node fetch) — see fetchInitialParentMessageId
+      // note above. tlsFetch triggers CloudFront WAF → HTML login redirect.
+      const createRes = await proxyAwareFetch(
         CONVERSATION_URL,
         {
           method: "POST",
@@ -655,7 +674,9 @@ export class HuggingChatExecutor extends BaseExecutor {
 
     let upstreamResponse;
     try {
-      upstreamResponse = await tlsFetch(
+      // Use proxyAwareFetch (native Node fetch) — see note above. tlsFetch
+      // triggers CloudFront WAF → HTML login redirect.
+      upstreamResponse = await proxyAwareFetch(
         messageUrl,
         {
           method: "POST",
diff --git a/open-sse/executors/opencode-go.js b/open-sse/executors/opencode-go.js
index 7bf47ed..f106783 100644
--- a/open-sse/executors/opencode-go.js
+++ b/open-sse/executors/opencode-go.js
@@ -15,15 +15,57 @@ const MESSAGES_FORMAT_MODELS = new Set([
 
 const BASE = "https://opencode.ai/zen/go/v1";
 
+// Effort-tier aliases — models on opencode-go that support per-effort suffixes.
+// Each entry maps the canonical base id to the set of effort tiers the upstream
+// supports. parseEffortLevel() parses the suffix (e.g. "glm-5.2-high" →
+// baseModel "glm-5.2", effort "high"), transformRequest rewrites body.model to
+// the canonical id and injects reasoning_effort if not already set by client.
+//
+// Tier support varies per upstream:
+//   - deepseek-v4-pro: all four tiers (low/medium/high/max)
+//   - glm-5.2:        high/max only (Z.AI maps these through the reasoning
+//                     plane; low/medium are not supported on OpenAI transport)
+//   - mimo-v2.5:      high/max only (Xiaomi MiMo does not document low/medium)
+//
+// Port of OmniRoute commit 1843b34 (PR #6987, issue #6922).
+const EFFORT_LEVELS = ["low", "medium", "high", "max"];
+const EFFORT_TIERS = {
+  "deepseek-v4-pro": EFFORT_LEVELS,
+  "glm-5.2": ["high", "max"],
+  "mimo-v2.5": ["high", "max"],
+};
+
+/**
+ * Parse a model string with an effort-level suffix.
+ * e.g. "deepseek-v4-pro-low" → { baseModel: "deepseek-v4-pro", effort: "low" }
+ *      "glm-5.2-high"         → { baseModel: "glm-5.2", effort: "high" }
+ * Returns null if the model doesn't match any known effort-tier pattern.
+ */
+export function parseEffortLevel(model) {
+  const m = String(model || "");
+  for (const [baseModel, levels] of Object.entries(EFFORT_TIERS)) {
+    for (const level of levels) {
+      if (m === `${baseModel}-${level}`) {
+        return { baseModel, effort: level };
+      }
+    }
+  }
+  return null;
+}
+
 export class OpenCodeGoExecutor extends BaseExecutor {
   constructor() {
     super("opencode-go", PROVIDERS["opencode-go"]);
   }
 
-  // buildUrl runs before buildHeaders in BaseExecutor.execute, cache model here
+  // buildUrl runs before buildHeaders in BaseExecutor.execute, cache the
+  // CANONICAL model here (strip effort-tier suffix if present) so buildHeaders
+  // checks the right id against MESSAGES_FORMAT_MODELS.
   buildUrl(model) {
-    this._lastModel = model;
-    return MESSAGES_FORMAT_MODELS.has(model)
+    const parsed = parseEffortLevel(model);
+    const canonical = parsed ? parsed.baseModel : model;
+    this._lastModel = canonical;
+    return MESSAGES_FORMAT_MODELS.has(canonical)
       ? `${BASE}/messages`
       : `${BASE}/chat/completions`;
   }
@@ -43,7 +85,26 @@ export class OpenCodeGoExecutor extends BaseExecutor {
     return headers;
   }
 
-  transformRequest(model, body) {
-    return injectReasoningContent({ provider: this.provider, model, body });
+  transformRequest(model, body, stream, credentials) {
+    const transformed = { ...(body && typeof body === "object" ? body : {}) };
+
+    // Effort-tier alias: rewrite body.model to canonical id and inject
+    // reasoning_effort (only if the client hasn't set one explicitly).
+    const parsed = parseEffortLevel(model);
+    if (parsed) {
+      transformed.model = parsed.baseModel;
+      if (transformed.reasoning_effort === undefined) {
+        transformed.reasoning_effort = parsed.effort;
+      }
+      // Pass the canonical model to injectReasoningContent so capability
+      // lookups (which key on base ids, not aliases) resolve correctly.
+      return injectReasoningContent({
+        provider: this.provider,
+        model: parsed.baseModel,
+        body: transformed,
+      });
+    }
+
+    return injectReasoningContent({ provider: this.provider, model, body: transformed });
   }
 }
diff --git a/open-sse/executors/qwen-web.js b/open-sse/executors/qwen-web.js
index 8415e6e..4ee91cd 100644
--- a/open-sse/executors/qwen-web.js
+++ b/open-sse/executors/qwen-web.js
@@ -58,6 +58,11 @@ const MODEL_ALIASES = {
 
 const DEFAULT_MODEL = "qwen3.7-max";
 
+// Some Qwen models reject requests with `thinking_enabled: false`. The model
+// name doesn't always contain "think"/"reason", so we maintain an explicit
+// allowlist — these models MUST get thinking_enabled=true.
+const REQUIRED_THINKING_MODELS = new Set(["qwen3.8-max-preview"]);
+
 function mapModel(modelId) {
   return MODEL_ALIASES[modelId] || modelId;
 }
@@ -161,7 +166,11 @@ function foldMessages(messages) {
 
 function buildMessagePayload(chatId, modelId, prompt, requestedModel) {
   const fid = uuid();
-  const enableThinking = /think|reason|r1/i.test(requestedModel);
+  // Thinking must be enabled when (a) the user picked a thinking-flavored
+  // model name, OR (b) the model is in REQUIRED_THINKING_MODELS (those models
+  // reject thinking_enabled:false with "Invalid input or attachment").
+  const enableThinking =
+    REQUIRED_THINKING_MODELS.has(modelId) || /think|reason|r1/i.test(requestedModel);
   const featureConfig = {
     thinking_enabled: enableThinking,
     output_schema: "phase",
@@ -169,6 +178,9 @@ function buildMessagePayload(chatId, modelId, prompt, requestedModel) {
     research_mode: "normal",
     auto_search: false,
   };
+  // Payload shape mirrors the proven OmniRoute reference implementation.
+  // Adding extra fields (version, id, extra, edited, error, turn_id, instructions)
+  // has been observed to trigger "invalid_input" rejections — keep it minimal.
   return {
     stream: true,
     incremental_output: true,
@@ -209,11 +221,41 @@ function parseSseDelta(line) {
   } catch {
     return null;
   }
+  // v2.1 error envelope: top-level `error` object arrives as its own event
+  // (e.g. {"error":{"code":"invalid_input","details":"..."},"response_id":...}).
+  // Surface it so the streaming transform can show the message instead of
+  // silently ending with empty content.
+  if (parsed?.error) {
+    const details = parsed.error.details || parsed.error.message || parsed.error.code || "upstream error";
+    return { kind: "error", text: details };
+  }
+  // Lifecycle events (response.created, response.completed, etc.) carry no
+  // content — ignore. The v2.1 protocol wraps them as {"<event_name>": {...}}.
+  if (parsed?.response_created || parsed?.response_completed) return null;
+
   const delta = parsed?.choices?.[0]?.delta;
   if (!delta) return null;
   const phase = delta.phase;
   const content = typeof delta.content === "string" ? delta.content : "";
-  if (phase === "think" || phase === "thinking_summary") {
+
+  // thinking_summary phase (qwen3.8+ / v2.1): the actual reasoning text lives
+  // in delta.extra.summary_thought.content[] (array of strings), NOT in
+  // delta.content (which is empty ""). Same for summary_title. Concatenate
+  // the array entries so the reasoning stream isn't dropped on the floor.
+  if (phase === "thinking_summary") {
+    const extra = delta.extra || {};
+    const parts = [];
+    if (Array.isArray(extra.summary_title?.content)) {
+      parts.push(...extra.summary_title.content);
+    }
+    if (Array.isArray(extra.summary_thought?.content)) {
+      parts.push(...extra.summary_thought.content);
+    }
+    const text = parts.filter(Boolean).join("\n");
+    return text ? { kind: "think", text } : null;
+  }
+
+  if (phase === "think") {
     return { kind: "think", text: content };
   }
   // answer phase or a null/absent phase both carry assistant content.
@@ -269,7 +311,11 @@ function buildClientStream(upstreamBody, modelId, signal) {
               emittedRole = true;
               controller.enqueue(push({ role: "assistant", content: "" }));
             }
-            if (delta.kind === "answer") {
+            if (delta.kind === "error") {
+              // v2.1 upstream error mid-stream — surface as content so the user
+              // sees why the response stopped (e.g. "Invalid input or attachment").
+              controller.enqueue(push({ content: `\n\n[Qwen error: ${delta.text}]` }));
+            } else if (delta.kind === "answer") {
               controller.enqueue(push({ content: delta.text }));
             } else if (delta.kind === "think") {
               controller.enqueue(push({ reasoning_content: delta.text }));
@@ -297,7 +343,8 @@ async function collectStream(upstreamBody, signal) {
   const decoder = new TextDecoder();
   let content = "";
   let reasoning = "";
-  if (!reader) return { content, reasoning };
+  let error = "";
+  if (!reader) return { content, reasoning, error };
 
   let buffer = "";
   try {
@@ -313,12 +360,13 @@ async function collectStream(upstreamBody, signal) {
         if (!delta) continue;
         if (delta.kind === "answer") content += delta.text;
         else if (delta.kind === "think") reasoning += delta.text;
+        else if (delta.kind === "error") error = delta.text;
       }
     }
   } catch {
     /* upstream closed mid-stream — return what we have */
   }
-  return { content, reasoning };
+  return { content, reasoning, error };
 }
 
 // ── Executor ────────────────────────────────────────────────────────────
@@ -478,8 +526,19 @@ export class QwenWebExecutor extends BaseExecutor {
     }
 
     if (!stream) {
-      const { content, reasoning } = await collectStream(upstream.body, signal);
-      const message = { role: "assistant", content: content || "" };
+      const { content, reasoning, error } = await collectStream(upstream.body, signal);
+      // If the upstream sent an error envelope mid-stream (v2.1 format) AND no
+      // content was produced, treat it as a failed request so the user sees the
+      // real cause instead of an empty 200.
+      if (error && !content) {
+        return {
+          response: errorResponse(502, `Qwen error: ${error}`, "UPSTREAM_ERROR"),
+          url: completionUrl,
+          headers: buildHeaders(token, cookieHeader, chatId),
+          transformedBody: msgPayload,
+        };
+      }
+      const message = { role: "assistant", content: content || (error ? `[Qwen error: ${error}]` : "") };
       if (reasoning) message.reasoning_content = reasoning;
       return {
         response: new Response(
diff --git a/open-sse/providers/capabilities.js b/open-sse/providers/capabilities.js
index 803a32d..9c9908c 100644
--- a/open-sse/providers/capabilities.js
+++ b/open-sse/providers/capabilities.js
@@ -181,6 +181,8 @@ export const PATTERN_CAPABILITIES = [
   { pattern: "*gpt-5*codex*",   caps: { reasoning: true, search: true, thinkingFormat: "openai", contextWindow: 400000, maxOutput: 128000 } },
   { pattern: "*gpt-5*",         caps: { vision: true, reasoning: true, search: true, thinkingFormat: "openai", contextWindow: 400000, maxOutput: 128000 } },
   { pattern: "*gpt-4o*",        caps: { vision: true, search: true, contextWindow: 128000, maxOutput: 16384 } },
+  // MAI-Code-1-Flash (Microsoft via GitHub Copilot) — code-generation model.
+  { pattern: "*mai-code*",      caps: { reasoning: true, thinkingFormat: "openai", contextWindow: 256000, maxOutput: 128000 } },
   { pattern: "*gpt-4.1*",       caps: { vision: true, contextWindow: 1000000, maxOutput: 32768 } },
   { pattern: "*gpt-4-turbo*",   caps: { vision: true, contextWindow: 128000 } },
   { pattern: "*gpt-4*",         caps: { contextWindow: 128000 } },
diff --git a/open-sse/providers/pricing.js b/open-sse/providers/pricing.js
index 3b02abc..c600f5e 100644
--- a/open-sse/providers/pricing.js
+++ b/open-sse/providers/pricing.js
@@ -138,6 +138,42 @@ export const PROVIDER_PRICING = {
   gh: {
     "gpt-5.3-codex": { input: 1.75, output: 14.00, cached: 0.175, reasoning: 14.00, cache_creation: 1.75 },
   },
+  // Forge Workspace (forge) — tier-based pricing via /v1/models-rates
+  forge: {
+    "gpt-5.6-sol": { input: 5.00, output: 30.00 },
+    "gpt-5.6-terra": { input: 2.50, output: 15.00 },
+    "gpt-5.6-luna": { input: 1.00, output: 6.00 },
+    "gpt-5.5": { input: 5.00, output: 40.00 },
+    "gpt-5.3-codex": { input: 1.75, output: 14.00 },
+    "claude-opus-4-5-20251101": { input: 5.00, output: 25.00 },
+    "claude-sonnet-4-5-20250929": { input: 3.00, output: 15.00 },
+    "claude-haiku-4-5-20251001": { input: 1.00, output: 5.00 },
+    "claude-sonnet-4-6": { input: 3.00, output: 15.00 },
+    "claude-sonnet-4-6-thinking": { input: 3.00, output: 15.00 },
+    "claude-sonnet-5": { input: 2.00, output: 10.00 },
+    "claude-fable-5": { input: 10.00, output: 50.00 },
+    "grok-4.5": { input: 2.00, output: 6.00 },
+    "grok-4.3": { input: 1.25, output: 2.50 },
+    "grok-build-0.1": { input: 1.00, output: 2.00 },
+    "deepseek-r1": { input: 0.401, output: 1.605 },
+    "deepseek-v4-flash": { input: 0.098, output: 0.196 },
+    "deepseek-v4-pro": { input: 0.304, output: 0.609 },
+    "deepseek-v3.2": { input: 0.20, output: 0.301 },
+    "deepseek-v3.1": { input: 0.401, output: 1.204 },
+    "deepseek-v3": { input: 0.20, output: 0.802 },
+    "kimi-k3": { input: 3.00, output: 15.00 },
+    "kimi-k2.7-code": { input: 0.625, output: 2.599 },
+    "kimi-k2.6": { input: 0.625, output: 2.599 },
+    "kimi-k2.5": { input: 0.401, output: 2.107 },
+    "gemini-3-pro-preview": { input: 2.00, output: 12.00 },
+    "gemini-3.5-flash": { input: 1.50, output: 9.00 },
+    "tencent/hy3": { input: 0.20, output: 0.80 },
+    "mimo-v2.5": { input: 0.10, output: 0.20 },
+    "mimo-v2.5-pro": { input: 0.30, output: 0.60 },
+    "MiniMax-M3": { input: 0.21, output: 0.84 },
+    "MiniMax-M2.5": { input: 0.21, output: 0.84 },
+    "glm-5.2": { input: 0.77, output: 2.695 },
+  },
 };
 
 /**
diff --git a/open-sse/providers/registry/alibaba-cn.js b/open-sse/providers/registry/alibaba-cn.js
new file mode 100644
index 0000000..4057f65
--- /dev/null
+++ b/open-sse/providers/registry/alibaba-cn.js
@@ -0,0 +1,50 @@
+// Alibaba (China) — DashScope China endpoint at dashscope.aliyuncs.com.
+// Same model catalog as the international `alibaba` provider but pointed at
+// the China-region host. Port of OmniRoute commit 00677044 (Part 3/3).
+export default {
+  id: "alibaba-cn",
+  priority: 161,
+  alias: "alibaba-cn",
+  aliases: ["alicn"],
+  uiAlias: "alicn",
+  display: {
+    name: "Alibaba (China)",
+    icon: "cloud",
+    color: "#FF6A00",
+    textIcon: "AC",
+    website: "https://dashscope.console.aliyun.com",
+    notice: {
+      signupUrl: "https://dashscope.console.aliyun.com",
+      apiKeyUrl: "https://dashscope.console.aliyun.com/apiKey",
+      text: "Alibaba Cloud Model Studio (DashScope China region). Create an API key at dashscope.console.aliyun.com. OpenAI-compatible.",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1/models",
+    auth: {
+      combined: true,
+      header: "Authorization",
+      scheme: "bearer",
+    },
+  },
+  models: [
+    { id: "qwen3.7-max", name: "Qwen3.7 Max" },
+    { id: "qwen3.7-plus", name: "Qwen3.7 Plus" },
+    { id: "qwen3.6-plus", name: "Qwen3.6 Plus" },
+    { id: "qwen3.6-27b", name: "Qwen3.6 27B" },
+    { id: "qwen3.6-35b-a3b", name: "Qwen3.6 35B A3B" },
+    { id: "qwen3.5-plus", name: "Qwen3.5 Plus" },
+    { id: "qwen3.5-122b-a10b", name: "Qwen3.5 122B A10B" },
+    { id: "qwen3.5-397b-a17b", name: "Qwen3.5 397B A17B" },
+    { id: "glm-5.2", name: "GLM 5.2" },
+    { id: "glm-5.2-fast-preview", name: "GLM 5.2 Fast Preview" },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
+    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+  ],
+  passthroughModels: true,
+};
diff --git a/open-sse/providers/registry/alibaba.js b/open-sse/providers/registry/alibaba.js
new file mode 100644
index 0000000..1549f77
--- /dev/null
+++ b/open-sse/providers/registry/alibaba.js
@@ -0,0 +1,52 @@
+// Alibaba (Model Studio / DashScope International) — API-key OpenAI-compatible
+// gateway. Same model catalog as `qwen-cloud` but under the canonical
+// "alibaba" alias for users who think in terms of the Alibaba Cloud brand.
+//
+// Port of OmniRoute commit 00677044 (Part 3/3).
+export default {
+  id: "alibaba",
+  priority: 160,
+  alias: "alibaba",
+  aliases: ["ali"],
+  uiAlias: "ali",
+  display: {
+    name: "Alibaba (Model Studio)",
+    icon: "cloud",
+    color: "#FF6A00",
+    textIcon: "AL",
+    website: "https://dashscope.console.aliyun.com",
+    notice: {
+      signupUrl: "https://dashscope.console.aliyun.com",
+      apiKeyUrl: "https://dashscope.console.aliyun.com/apiKey",
+      text: "Alibaba Cloud Model Studio (DashScope International). Create an API key at dashscope.console.aliyun.com. OpenAI-compatible.",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/models",
+    auth: {
+      combined: true,
+      header: "Authorization",
+      scheme: "bearer",
+    },
+  },
+  models: [
+    { id: "qwen3.7-max", name: "Qwen3.7 Max" },
+    { id: "qwen3.7-plus", name: "Qwen3.7 Plus" },
+    { id: "qwen3.6-plus", name: "Qwen3.6 Plus" },
+    { id: "qwen3.6-27b", name: "Qwen3.6 27B" },
+    { id: "qwen3.6-35b-a3b", name: "Qwen3.6 35B A3B" },
+    { id: "qwen3.5-plus", name: "Qwen3.5 Plus" },
+    { id: "qwen3.5-122b-a10b", name: "Qwen3.5 122B A10B" },
+    { id: "qwen3.5-397b-a17b", name: "Qwen3.5 397B A17B" },
+    { id: "glm-5.2", name: "GLM 5.2" },
+    { id: "glm-5.2-fast-preview", name: "GLM 5.2 Fast Preview" },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
+    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+  ],
+  passthroughModels: true,
+};
diff --git a/open-sse/providers/registry/forge.js b/open-sse/providers/registry/forge.js
new file mode 100644
index 0000000..fadd4ca
--- /dev/null
+++ b/open-sse/providers/registry/forge.js
@@ -0,0 +1,97 @@
+// Forge Workspace (Forge AI) — AI model gateway (forge-ai.space).
+//
+// OpenAI-compatible API with tier-based model access:
+//   - FREE tier: 32 models (GPT-5.6, Claude, DeepSeek, Kimi, Gemini, Grok, etc.)
+//   - PRO tier: +1 model (Claude Fable 5)
+//
+// The standard /v1/models endpoint auto-filters by the API key's tier, so
+// live model discovery naturally shows only accessible models.
+// The /v1/models-rates endpoint provides full tier + pricing metadata.
+//
+// No custom executor needed — DefaultExecutor handles OpenAI-compatible APIs.
+export default {
+  id: "forge",
+  priority: 310,
+  alias: "forge",
+  aliases: ["fg", "forgeai"],
+  uiAlias: "fg",
+  display: {
+    name: "Forge Workspace",
+    icon: "local_fire_department",
+    color: "#FF6B35",
+    textIcon: "FG",
+    website: "https://www.forge-ai.space/",
+    notice: {
+      signupUrl: "https://www.forge-ai.space/?ref=ref-a556c7a8",
+      apiKeyUrl: "https://www.forge-ai.space/?ref=ref-a556c7a8",
+      text: "Forge Workspace is an AI model gateway with 33+ models across FREE and PRO tiers. Sign up for free credits at forge-ai.space, then paste your API key. OpenAI-compatible — works with any OpenAI-format client.",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://forge-gateway-api.fly.dev/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://forge-gateway-api.fly.dev/v1/models",
+  },
+  // Seed catalog — offline fallback when live /v1/models fetch fails.
+  // Models grouped by upstream provider for discoverability.
+  // Tier metadata: FREE models available to all keys, PRO requires upgrade.
+  // The /v1/models endpoint auto-filters by the API key's tier at runtime.
+  models: [
+    // OpenAI (FREE)
+    { id: "gpt-5.6-sol", name: "GPT-5.6 Sol", contextWindow: 1050000, tier: "FREE" },
+    { id: "gpt-5.6-terra", name: "GPT-5.6 Terra", contextWindow: 1050000, tier: "FREE" },
+    { id: "gpt-5.6-luna", name: "GPT-5.6 Luna", contextWindow: 1050000, tier: "FREE" },
+    { id: "gpt-5.5", name: "GPT-5.5", contextWindow: 128000, tier: "FREE" },
+    { id: "gpt-5.3-codex", name: "GPT-5.3 Codex", contextWindow: 400000, tier: "FREE" },
+    // Anthropic (FREE)
+    { id: "claude-sonnet-4-6", name: "Claude Sonnet 4.6", contextWindow: 1000000, tier: "FREE" },
+    { id: "claude-sonnet-4-6-thinking", name: "Claude Sonnet 4.6 Thinking", contextWindow: 1000000, tier: "FREE" },
+    { id: "claude-sonnet-5", name: "Claude Sonnet 5", contextWindow: 1000000, tier: "FREE" },
+    { id: "claude-opus-4-5-20251101", name: "Claude Opus 4.5", contextWindow: 200000, tier: "FREE" },
+    { id: "claude-sonnet-4-5-20250929", name: "Claude Sonnet 4.5", contextWindow: 200000, tier: "FREE" },
+    { id: "claude-haiku-4-5-20251001", name: "Claude Haiku 4.5", contextWindow: 200000, tier: "FREE" },
+    // Anthropic (PRO)
+    { id: "claude-fable-5", name: "Claude Fable 5", contextWindow: 1000000, tier: "PRO" },
+    // xAI (FREE)
+    { id: "grok-4.5", name: "Grok 4.5", contextWindow: 500000, tier: "FREE" },
+    { id: "grok-4.3", name: "Grok 4.3", contextWindow: 1000000, tier: "FREE" },
+    { id: "grok-build-0.1", name: "Grok Build 0.1", contextWindow: 256000, tier: "FREE" },
+    // DeepSeek (FREE)
+    { id: "deepseek-r1", name: "DeepSeek R1", contextWindow: 128000, tier: "FREE" },
+    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash", contextWindow: 1000000, tier: "FREE" },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro", contextWindow: 1000000, tier: "FREE" },
+    { id: "deepseek-v3.2", name: "DeepSeek V3.2", contextWindow: 163000, tier: "FREE" },
+    { id: "deepseek-v3.1", name: "DeepSeek V3.1", contextWindow: 128000, tier: "FREE" },
+    { id: "deepseek-v3", name: "DeepSeek V3", contextWindow: 128000, tier: "FREE" },
+    // Kimi (FREE)
+    { id: "kimi-k3", name: "Kimi K3", contextWindow: 1000000, tier: "FREE" },
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code", contextWindow: 1000000, tier: "FREE" },
+    { id: "kimi-k2.6", name: "Kimi K2.6", contextWindow: 1000000, tier: "FREE" },
+    { id: "kimi-k2.5", name: "Kimi K2.5", contextWindow: 1000000, tier: "FREE" },
+    // Gemini (FREE)
+    { id: "gemini-3-pro-preview", name: "Gemini 3 Pro Preview", contextWindow: 1000000, tier: "FREE" },
+    { id: "gemini-3.5-flash", name: "Gemini 3.5 Flash", contextWindow: 1000000, tier: "FREE" },
+    // Others (FREE)
+    { id: "tencent/hy3", name: "Tencent HY3", contextWindow: 1000000, tier: "FREE" },
+    { id: "mimo-v2.5", name: "MiMo V2.5", contextWindow: 1000000, tier: "FREE" },
+    { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro", contextWindow: 1000000, tier: "FREE" },
+    { id: "MiniMax-M3", name: "MiniMax M3", contextWindow: 1000000, tier: "FREE" },
+    { id: "MiniMax-M2.5", name: "MiniMax M2.5", contextWindow: 1000000, tier: "FREE" },
+    { id: "glm-5.2", name: "GLM 5.2", contextWindow: 1000000, tier: "FREE" },
+  ],
+  // Live model discovery: /v1/models auto-filters by API key tier.
+  // FREE key → 32 models, PRO key → 33 models (+ Claude Fable 5).
+  modelsFetcher: {
+    url: "https://forge-gateway-api.fly.dev/v1/models",
+    type: "openai",
+  },
+  // Forge supports reasoning via "thinking": true in the request body.
+  // thinkingConfig enables the thinking level picker in the UI for all models.
+  thinkingConfig: {
+    options: ["auto", "none", "max"],
+    defaultMode: "auto",
+  },
+  passthroughModels: true,
+};
diff --git a/open-sse/providers/registry/github.js b/open-sse/providers/registry/github.js
index 95169eb..9e1bbb2 100644
--- a/open-sse/providers/registry/github.js
+++ b/open-sse/providers/registry/github.js
@@ -18,6 +18,12 @@ export default {
   transport: {
     baseUrl: "https://api.githubcopilot.com/chat/completions",
     responsesUrl: "https://api.githubcopilot.com/responses",
+    // Anthropic-native shim: the only Copilot endpoint that surfaces prompt-cache
+    // token counts (cached_tokens) for Claude models, and avoids round-tripping
+    // tool_use/tool_result/thinking content blocks through the OpenAI shape.
+    // Routed via each claude-* model's targetFormat: "claude" below (see
+    // GithubExecutor buildUrl/buildHeaders). Port of decolua/9router#2608.
+    messagesUrl: "https://api.githubcopilot.com/v1/messages",
     headers: {
       "copilot-integration-id": "vscode-chat",
       "editor-version": "vscode/1.110.0",
@@ -40,24 +46,39 @@ export default {
       url: "https://api.github.com/copilot_internal/user",
     },
   },
+  // Models synced with OmniRoute commit fea1d54 (2026-07). All claude-* entries
+  // carry targetFormat: "claude" so chatCore translates them to Anthropic-native
+  // shape and GithubExecutor routes them to /v1/messages (see messagesUrl above).
+  // contextLength → contextWindow, maxOutputTokens → maxOutput (ExtremeRouter
+  // camelCase convention).
   models: [
-    { id: "gpt-5.2", name: "GPT-5.2" },
-    { id: "gpt-5.2-codex", name: "GPT-5.2 Codex" },
-    { id: "gpt-5.3-codex", name: "GPT-5.3 Codex" },
-    { id: "gpt-5.4", name: "GPT-5.4" },
-    { id: "gpt-5.4-mini", name: "GPT-5.4 Mini" },
-    { id: "claude-haiku-4.5", name: "Claude Haiku 4.5" },
-    { id: "claude-opus-4.5", name: "Claude Opus 4.5" },
-    { id: "claude-sonnet-4.5", name: "Claude Sonnet 4.5" },
-    { id: "claude-sonnet-4.6", name: "Claude Sonnet 4.6" },
-    { id: "claude-opus-4.6", name: "Claude Opus 4.6" },
-    { id: "claude-opus-4.7", name: "Claude Opus 4.7" },
-    { id: "gemini-2.5-pro", name: "Gemini 2.5 Pro" },
-    { id: "gemini-3-flash-preview", name: "Gemini 3 Flash" },
-    { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro" },
-    { id: "grok-code-fast-1", name: "Grok Code Fast 1" },
-    { id: "oswe-vscode-prime", name: "Raptor Mini" },
-    { id: "goldeneye-free-auto", name: "GoldenEye" },
+    // ── Claude (native /v1/messages path) ─────────────────────────────────
+    { id: "claude-fable-5", name: "Claude Fable 5", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000 },
+    { id: "claude-opus-4.8-fast", name: "Claude Opus 4.8 (fast mode)", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000, unsupportedParams: ["temperature", "top_p", "top_k"] },
+    { id: "claude-opus-4.8", name: "Claude Opus 4.8", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000, unsupportedParams: ["temperature", "top_p", "top_k"] },
+    { id: "claude-opus-4.7", name: "Claude Opus 4.7", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000 },
+    { id: "claude-sonnet-4.6", name: "Claude Sonnet 4.6", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000 },
+    { id: "claude-opus-4.5", name: "Claude Opus 4.5", targetFormat: "claude", contextWindow: 200000, maxOutput: 32000 },
+    { id: "claude-sonnet-5", name: "Claude Sonnet 5", targetFormat: "claude", contextWindow: 1000000, maxOutput: 64000 },
+    { id: "claude-sonnet-4.5", name: "Claude Sonnet 4.5", targetFormat: "claude", contextWindow: 200000, maxOutput: 32000 },
+    { id: "claude-haiku-4.5", name: "Claude Haiku 4.5", targetFormat: "claude", contextWindow: 200000, maxOutput: 32000 },
+    // ── Gemini (#2911: must use chat/completions, not Responses API) ───────
+    { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro", contextWindow: 1000000, maxOutput: 64000 },
+    { id: "gemini-3.5-flash", name: "Gemini 3.5 Flash", contextWindow: 1000000, maxOutput: 64000 },
+    // ── GPT (Responses API) ────────────────────────────────────────────────
+    { id: "gpt-5.5", name: "GPT-5.5", targetFormat: "openai-responses", contextWindow: 1050000, maxOutput: 128000 },
+    { id: "gpt-5.4", name: "GPT-5.4", targetFormat: "openai-responses", contextWindow: 1050000, maxOutput: 128000 },
+    { id: "gpt-5.4-mini", name: "GPT-5.4 Mini", targetFormat: "openai-responses", contextWindow: 400000, maxOutput: 128000 },
+    { id: "gpt-5.3-codex", name: "GPT-5.3 Codex", targetFormat: "openai-responses", contextWindow: 400000, maxOutput: 128000 },
+    { id: "gpt-5-mini", name: "GPT-5 Mini", targetFormat: "openai-responses", contextWindow: 264000, maxOutput: 64000 },
+    { id: "gpt-4o-2024-11-20", name: "GPT-4o", contextWindow: 128000, maxOutput: 16384 },
+    { id: "gpt-4o-mini", name: "GPT-4o Mini", contextWindow: 128000, maxOutput: 4096 },
+    { id: "gpt-4-0125-preview", name: "GPT 4 Turbo", contextWindow: 128000, maxOutput: 4096 },
+    // ── Other ─────────────────────────────────────────────────────────────
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code", contextWindow: 256000, maxOutput: 32000 },
+    { id: "mai-code-1-flash", name: "MAI-Code-1-Flash", targetFormat: "openai-responses", contextWindow: 256000, maxOutput: 128000 },
+    { id: "oswe-vscode-prime", name: "Raptor Mini", targetFormat: "openai-responses", contextWindow: 264000, maxOutput: 64000 },
+    // ── Embeddings (separate kind, handled by embeddingConfig above) ───────
     { id: "text-embedding-3-small", name: "Text Embedding 3 Small (GitHub)", kind: "embedding" },
     { id: "text-embedding-3-large", name: "Text Embedding 3 Large (GitHub)", kind: "embedding" },
   ],
diff --git a/open-sse/providers/registry/hcnsec.js b/open-sse/providers/registry/hcnsec.js
new file mode 100644
index 0000000..89d5ee0
--- /dev/null
+++ b/open-sse/providers/registry/hcnsec.js
@@ -0,0 +1,47 @@
+// Huancheng Public API (hcnsec) — Xinjiang Huancheng Cybersecurity public LLM
+// API platform. OpenAI-compatible /v1 endpoints behind a single API key.
+//
+// Free credits available with daily check-ins. Models discovered live via
+// /v1/models at runtime (passthroughModels + empty seed catalog).
+//
+// Port of OmniRoute commit 437ca488 (PR #6410).
+export default {
+  id: "hcnsec",
+  priority: 330,
+  alias: "hcnsec",
+  aliases: ["hc"],
+  uiAlias: "hc",
+  display: {
+    name: "Huancheng Public API",
+    icon: "security",
+    color: "#0EA5E9",
+    textIcon: "HC",
+    website: "https://api.hcnsec.cn",
+    notice: {
+      signupUrl: "https://api.hcnsec.cn/sign-up?aff=ZKgv",
+      apiKeyUrl: "https://api.hcnsec.cn/sign-up?aff=ZKgv",
+      text: "Xinjiang Huancheng Cybersecurity public LLM API platform. Free credits with daily check-ins. Create an API key at api.hcnsec.cn, then paste it here. OpenAI-compatible — works with any OpenAI-format client.",
+    },
+  },
+  category: "freeTier",
+  hasFree: true,
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://api.hcnsec.cn/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://api.hcnsec.cn/v1/models",
+    auth: {
+      combined: true,
+      header: "Authorization",
+      scheme: "bearer",
+    },
+  },
+  // Live discovery only — /v1/models exposes whatever the key has access to.
+  // Empty seed catalog; passthroughModels allows any model id.
+  models: [],
+  passthroughModels: true,
+  modelsFetcher: {
+    url: "https://api.hcnsec.cn/v1/models",
+    type: "openai",
+  },
+};
diff --git a/open-sse/providers/registry/huggingchat.js b/open-sse/providers/registry/huggingchat.js
index f1f21b9..b313324 100644
--- a/open-sse/providers/registry/huggingchat.js
+++ b/open-sse/providers/registry/huggingchat.js
@@ -45,6 +45,11 @@ export default {
   // the upstream model id verbatim. Vision/tool/reasoning capabilities are not
   // surfaced here (plain text chat only).
   models: [
+    // "omni" is HuggingChat's internal auto-router — it picks the best model
+    // dynamically. This is what the SPA sends by default when no specific model
+    // is selected, and is the most reliable entry (individual model ids can be
+    // retired by HF at any time, causing 200+HTML fallbacks).
+    { id: "omni", name: "Omni (Auto-Router)" },
     { id: "baidu/ERNIE-4.5-VL-424B-A47B-Base-PT", name: "ERNIE 4.5 VL 424B A47B Base PT" },
     { id: "CohereLabs/c4ai-command-r7b-12-2024", name: "Command R7B 12-2024" },
     { id: "CohereLabs/command-a-reasoning-08-2025", name: "Command A Reasoning 08-2025" },
diff --git a/open-sse/providers/registry/index.js b/open-sse/providers/registry/index.js
index 5fbce04..c01f57e 100644
--- a/open-sse/providers/registry/index.js
+++ b/open-sse/providers/registry/index.js
@@ -130,6 +130,13 @@ import p127 from "./perplexity-agent.js";
 import p128 from "./featherless.js";
 import p129 from "./moonshot.js";
 import p130 from "./qwencloud.js";
+import p131 from "./forge.js";
+import p132 from "./tokenrouter.js";
+import p133 from "./qwen-cloud.js";
+import p134 from "./qwen-cloud-token-plan.js";
+import p135 from "./alibaba.js";
+import p136 from "./alibaba-cn.js";
+import p137 from "./hcnsec.js";
 
 export default [
   p0,
@@ -262,5 +269,12 @@ export default [
   p127,
   p128,
   p129,
-  p130
+  p130,
+  p131,
+  p132,
+  p133,
+  p134,
+  p135,
+  p136,
+  p137,
 ];
diff --git a/open-sse/providers/registry/opencode-go.js b/open-sse/providers/registry/opencode-go.js
index c980cc8..85d4de7 100644
--- a/open-sse/providers/registry/opencode-go.js
+++ b/open-sse/providers/registry/opencode-go.js
@@ -23,13 +23,28 @@ export default {
     headers: {},
   },
   models: [
+    // GLM-5.2 — base model + effort-tier aliases (#6922).
+    // OpenCodeGoExecutor.parseEffortLevel() rewrites the alias to the canonical
+    // id and injects reasoning_effort, mirroring the deepseek-v4-pro-* pattern.
+    // GLM-5.2 supports high/max only (Z.AI reasoning plane; low/medium rejected).
     { id: "glm-5.2", name: "GLM 5.2" },
+    { id: "glm-5.2-high", name: "GLM 5.2 (High Effort)" },
+    { id: "glm-5.2-max", name: "GLM 5.2 (Max Effort)" },
     { id: "glm-5.1", name: "GLM 5.1" },
     { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
     { id: "kimi-k2.6", name: "Kimi K2.6" },
+    // DeepSeek V4 Pro — all four effort tiers (low/medium/high/max).
     { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
+    { id: "deepseek-v4-pro-low", name: "DeepSeek V4 Pro (Low Effort)" },
+    { id: "deepseek-v4-pro-medium", name: "DeepSeek V4 Pro (Medium Effort)" },
+    { id: "deepseek-v4-pro-high", name: "DeepSeek V4 Pro (High Effort)" },
+    { id: "deepseek-v4-pro-max", name: "DeepSeek V4 Pro (Max Effort)" },
     { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
+    // MiMo-V2.5 — base model + effort-tier aliases (#6922).
+    // Supports high/max only (Xiaomi MiMo does not document low/medium tiers).
     { id: "mimo-v2.5", name: "MiMo V2.5" },
+    { id: "mimo-v2.5-high", name: "MiMo V2.5 (High Effort)" },
+    { id: "mimo-v2.5-max", name: "MiMo V2.5 (Max Effort)" },
     { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro" },
     { id: "minimax-m3", name: "MiniMax M3", targetFormat: "claude" },
     { id: "minimax-m2.7", name: "MiniMax M2.7", targetFormat: "claude" },
diff --git a/open-sse/providers/registry/qwen-cloud-token-plan.js b/open-sse/providers/registry/qwen-cloud-token-plan.js
new file mode 100644
index 0000000..2f58e5f
--- /dev/null
+++ b/open-sse/providers/registry/qwen-cloud-token-plan.js
@@ -0,0 +1,76 @@
+// Qwen Cloud (Token Plan, ap-southeast-1) — Alibaba token-plan endpoint with
+// qwen3.8-max-preview, hosted at ap-southeast-1.maas.aliyuncs.com.
+//
+// This is the regional Alibaba MaaS endpoint used by token-plan subscriptions.
+// Distinct from `qwen-cloud` (DashScope international) and `qwencloud` (cookie
+// web). Port of OmniRoute commit 00677044 (Part 3/3).
+export default {
+  id: "qwen-cloud-token-plan",
+  priority: 166,
+  alias: "qct",
+  aliases: ["qwen-cloud-token-plan"],
+  uiAlias: "qct",
+  display: {
+    name: "Qwen Cloud (Token Plan)",
+    icon: "cloud",
+    color: "#615CED",
+    textIcon: "QT",
+    website: "https://dashscope.console.aliyun.com",
+    notice: {
+      signupUrl: "https://dashscope.console.aliyun.com",
+      apiKeyUrl: "https://dashscope.console.aliyun.com/apiKey",
+      text: "Alibaba token-plan endpoint (ap-southeast-1) for Qwen3.8 Max Preview and other frontier models. Paste your API key here. OpenAI-compatible.",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl:
+      "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions",
+    format: "openai",
+    auth: {
+      combined: true,
+      header: "Authorization",
+      scheme: "bearer",
+    },
+  },
+  models: [
+    {
+      id: "qwen3.8-max-preview",
+      name: "Qwen3.8 Max Preview",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.7-max",
+      name: "Qwen3.7 Max",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.7-plus",
+      name: "Qwen3.7 Plus",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.6-flash",
+      name: "Qwen3.6 Flash",
+      contextWindow: 1000000,
+      maxOutput: 32768,
+    },
+    {
+      id: "glm-5.2",
+      name: "GLM 5.2",
+      contextWindow: 1000000,
+      maxOutput: 16384,
+    },
+    {
+      id: "deepseek-v4-pro",
+      name: "DeepSeek V4 Pro",
+      contextWindow: 163840,
+      maxOutput: 32768,
+    },
+  ],
+  passthroughModels: true,
+};
diff --git a/open-sse/providers/registry/qwen-cloud.js b/open-sse/providers/registry/qwen-cloud.js
new file mode 100644
index 0000000..e73d716
--- /dev/null
+++ b/open-sse/providers/registry/qwen-cloud.js
@@ -0,0 +1,55 @@
+// Qwen Cloud (Alibaba DashScope International) — API-key OpenAI-compatible
+// gateway at dashscope-intl.aliyuncs.com.
+//
+// Distinct from `qwencloud` (the cookie-based chat.qwen.ai web provider) and
+// from `qwen-web` (cookie-based chat.qwen.ai v2 API). This is the official
+// DashScope international API-key endpoint, OpenAI-compatible.
+//
+// Port of OmniRoute commit 00677044 (Part 3/3).
+export default {
+  id: "qwen-cloud",
+  priority: 165,
+  alias: "qwc",
+  aliases: ["qwen-cloud"],
+  uiAlias: "qwc",
+  display: {
+    name: "Qwen Cloud (DashScope)",
+    icon: "cloud",
+    color: "#615CED",
+    textIcon: "QC",
+    website: "https://dashscope.console.aliyun.com",
+    notice: {
+      signupUrl: "https://dashscope.console.aliyun.com",
+      apiKeyUrl: "https://dashscope.console.aliyun.com/apiKey",
+      text: "Alibaba DashScope international API. Create an API key at dashscope.console.aliyun.com, then paste it here. OpenAI-compatible — works with any OpenAI-format client.",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/models",
+    auth: {
+      combined: true,
+      header: "Authorization",
+      scheme: "bearer",
+    },
+  },
+  models: [
+    { id: "qwen3.7-max-2026-06-08", name: "Qwen3.7 Max (2026-06-08)" },
+    { id: "qwen3.7-plus", name: "Qwen3.7 Plus" },
+    { id: "qwen3.6-plus", name: "Qwen3.6 Plus" },
+    { id: "qwen3.6-27b", name: "Qwen3.6 27B" },
+    { id: "qwen3.6-35b-a3b", name: "Qwen3.6 35B A3B" },
+    { id: "qwen3.5-plus-2026-04-20", name: "Qwen3.5 Plus (2026-04-20)" },
+    { id: "qwen3.5-122b-a10b", name: "Qwen3.5 122B A10B" },
+    { id: "qwen3.5-397b-a17b", name: "Qwen3.5 397B A17B" },
+    { id: "glm-5.2", name: "GLM 5.2" },
+    { id: "glm-5.2-fast-preview", name: "GLM 5.2 Fast Preview" },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
+    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+  ],
+  passthroughModels: true,
+};
diff --git a/open-sse/providers/registry/qwen-web.js b/open-sse/providers/registry/qwen-web.js
index 663eaca..58a326f 100644
--- a/open-sse/providers/registry/qwen-web.js
+++ b/open-sse/providers/registry/qwen-web.js
@@ -35,10 +35,34 @@ export default {
     format: "qwen-web",
     authType: "cookie",
   },
+  // Model entries include capability flags (contextWindow, maxOutput,
+  // supportsReasoning/vision/toolCalling) so the model picker + capability
+  // resolver surface them correctly. Port of OmniRoute commit ccdbc89 (Part 2/3).
   models: [
-    { id: "qwen3.7-max", name: "Qwen3.7 Max" },
-    { id: "qwen3.7-plus", name: "Qwen3.7 Plus" },
-    { id: "qwen3.6-plus", name: "Qwen3.6 Plus" },
+    {
+      id: "qwen3.8-max-preview",
+      name: "Qwen3.8 Max Preview",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.7-max",
+      name: "Qwen3.7 Max",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.7-plus",
+      name: "Qwen3.7 Plus",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
+    {
+      id: "qwen3.6-plus",
+      name: "Qwen3.6 Plus",
+      contextWindow: 1000000,
+      maxOutput: 65536,
+    },
   ],
   passthroughModels: true,
 };
diff --git a/open-sse/providers/registry/tokenrouter.js b/open-sse/providers/registry/tokenrouter.js
new file mode 100644
index 0000000..2cadab2
--- /dev/null
+++ b/open-sse/providers/registry/tokenrouter.js
@@ -0,0 +1,55 @@
+// TokenRouter — AI model gateway (tokenrouter.com).
+//
+// OpenAI-compatible /v1/chat/completions behind a single `sk-` API key.
+// Hosts Claude, GPT, Gemini, Kimi, DeepSeek, GLM and many other models — the
+// catalog is exposed via /v1/models (live discovery).
+//
+// Quota tracking is supported via the management API. The user pastes BOTH:
+//   - chat API key (sk-...) → used for /v1 chat completions
+//   - management key         → stored in providerSpecificData.mgmtKey, used for
+//                              /api/management/self/wallet (Quota Tracker)
+//
+// No custom executor needed — DefaultExecutor handles OpenAI-compatible APIs.
+export default {
+  id: "tokenrouter",
+  priority: 320,
+  alias: "tokenrouter",
+  aliases: ["tr"],
+  uiAlias: "tr",
+  display: {
+    name: "TokenRouter",
+    icon: "router",
+    color: "#10B981",
+    textIcon: "TR",
+    website: "https://api.tokenrouter.com",
+    notice: {
+      signupUrl: "https://api.tokenrouter.com",
+      apiKeyUrl: "https://api.tokenrouter.com",
+      text: "TokenRouter is an AI model gateway with Claude, GPT, Gemini, Kimi, DeepSeek, GLM and more. Create an API key at api.tokenrouter.com, then paste it here. OpenAI-compatible — works with any OpenAI-format client. For Quota Tracker, also paste your management key (separate from the chat sk- key).",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  // BOTH flags are required: `usage` gates USAGE_SUPPORTED_PROVIDERS (the
+  // general list shown in the Quota UI), `usageApikey` gates
+  // USAGE_APIKEY_PROVIDERS (api-key providers allowed past the authType check
+  // in /api/providers/client route.js isUsageEligible).
+  features: { usage: true, usageApikey: true },
+  transport: {
+    baseUrl: "https://api.tokenrouter.com/v1/chat/completions",
+    format: "openai",
+    validateUrl: "https://api.tokenrouter.com/v1/models",
+    // Quota Tracker endpoints (require the management key, NOT the chat key).
+    usage: {
+      walletUrl: "https://api.tokenrouter.com/api/management/self/wallet",
+    },
+  },
+  // Live discovery only — /v1/models auto-filters by the API key's tier.
+  // Empty seed list; passthroughModels allows any model id.
+  models: [],
+  passthroughModels: true,
+  modelsFetcher: {
+    url: "https://api.tokenrouter.com/v1/models",
+    type: "openai",
+  },
+};
diff --git a/open-sse/providers/registry/xai.js b/open-sse/providers/registry/xai.js
index 4ad8831..3e56098 100644
--- a/open-sse/providers/registry/xai.js
+++ b/open-sse/providers/registry/xai.js
@@ -34,7 +34,11 @@ export default {
       subscriptionUrl: "https://api.x.ai/v1/user?include=subscription",
     },
   },
-  features: { usage: true },
+  // Both flags required: `usage` gates USAGE_SUPPORTED_PROVIDERS (general
+  // list shown in Quota UI), `usageApikey` gates USAGE_APIKEY_PROVIDERS
+  // (api-key connections allowed past the authType check). xAI billing API
+  // works with both OAuth tokens and API keys.
+  features: { usage: true, usageApikey: true },
   models: [
     // Reasoning LLMs (Responses API + Chat Completions)
     { id: "grok-4.5", name: "Grok 4.5" },
diff --git a/open-sse/services/usage.js b/open-sse/services/usage.js
index d3872e4..c36f6ec 100644
--- a/open-sse/services/usage.js
+++ b/open-sse/services/usage.js
@@ -20,6 +20,7 @@ import {
   getQoderUsage,
 } from "./usage/misc.js";
 import { getXaiUsage } from "./usage/xai.js";
+import { getTokenRouterUsage } from "./usage/tokenrouter.js";
 
 /**
  * Get usage data for a provider connection
@@ -45,6 +46,7 @@ const USAGE_HANDLERS = {
   "vercel-ai-gateway": (c) => getVercelAiGatewayUsage(c.apiKey, c.proxyOptions),
   "codebuddy-cn": (c) => getCodeBuddyCnUsage(c.accessToken, c.apiKey, c.providerSpecificData, c.proxyOptions),
   xai: (c) => getXaiUsage(c, c.proxyOptions),
+  tokenrouter: (c) => getTokenRouterUsage(c, c.providerSpecificData, c.proxyOptions),
 };
 
 export async function getUsageForProvider(connection, proxyOptions = null) {
diff --git a/open-sse/services/usage/tokenrouter.js b/open-sse/services/usage/tokenrouter.js
new file mode 100644
index 0000000..f9ee956
--- /dev/null
+++ b/open-sse/services/usage/tokenrouter.js
@@ -0,0 +1,116 @@
+// TokenRouter usage handler for Quota Tracker.
+//
+// Pulls wallet balance from the management API:
+//   GET /api/management/self/wallet
+//   Authorization: Bearer <management key>
+//
+// Returns:
+//   {
+//     topUpBalance: 120.50,         // remaining topped-up balance
+//     voucherEfficientAmount: 30.00,// valid voucher balance
+//     toppedUpSpent: 80.25,         // amount consumed from top-ups
+//     voucherSpent: 10.00           // amount consumed from vouchers
+//   }
+//
+// The management key is SEPARATE from the chat API key — it lives in
+// connection.providerSpecificData.mgmtKey. Without it, we return null and the
+// Quota Tracker simply shows nothing for this connection (no error).
+
+import { U, toFiniteNumber } from "./shared.js";
+import { proxyAwareFetch } from "../../utils/proxyFetch.js";
+
+const USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36";
+
+export async function getTokenRouterUsage(credentials, providerSpecificData = {}, proxyOptions = null) {
+  // The management key is intentionally NOT the chat API key. It must be set
+  // separately on the connection (providerSpecificData.mgmtKey).
+  const mgmtKey = providerSpecificData?.mgmtKey || credentials?.providerSpecificData?.mgmtKey;
+  if (!mgmtKey) {
+    // Return an empty (but non-null) result with a message explaining what's
+    // missing. Returning `null` would make the Quota UI crash on `data.plan`.
+    return {
+      quotas: {},
+      plan: null,
+      message: "No management key set. Edit the connection and add your TokenRouter management key (separate from the chat API key) to enable wallet tracking.",
+    };
+  }
+
+  const cfg = U("tokenrouter");
+  const walletUrl = cfg.walletUrl;
+  if (!walletUrl) return null;
+
+  const headers = {
+    Authorization: `Bearer ${mgmtKey}`,
+    Accept: "application/json",
+    "User-Agent": USER_AGENT,
+  };
+
+  let res;
+  try {
+    res = await proxyAwareFetch(walletUrl, { method: "GET", headers }, proxyOptions);
+  } catch {
+    return { quotas: {}, plan: null, message: "Could not reach TokenRouter wallet API (network error)." };
+  }
+  if (!res?.ok) {
+    const status = res?.status ?? 0;
+    const hint = status === 401 || status === 403
+      ? "Authentication failed — check that your management key is correct and hasn't expired."
+      : `Wallet API returned HTTP ${status}.`;
+    return { quotas: {}, plan: null, message: hint };
+  }
+
+  const body = await res.json().catch(() => null);
+  if (!body?.success || !body.data) {
+    return { quotas: {}, plan: null, message: "Wallet API returned an unexpected response shape." };
+  }
+
+  const w = body.data;
+  // Aggregate topped-up + voucher into a single "wallet" quota row.
+  const topUpBalance = toFiniteNumber(w.topUpBalance);
+  const voucherBalance = toFiniteNumber(w.voucherEfficientAmount);
+  const toppedUpSpent = toFiniteNumber(w.toppedUpSpent);
+  const voucherSpent = toFiniteNumber(w.voucherSpent);
+
+  const totalBalance = topUpBalance + voucherBalance;
+  const totalSpent = toppedUpSpent + voucherSpent;
+  const total = totalBalance + totalSpent;
+
+  const quotas = {
+    Wallet: {
+      used: totalSpent,
+      total,
+      remaining: totalBalance,
+      remainingPercentage: total > 0 ? Math.round((totalBalance / total) * 100) : null,
+      resetAt: null,
+    },
+  };
+
+  // Break down top-up vs voucher as separate rows when both are present, so the
+  // user can see where their credit comes from.
+  if (topUpBalance > 0 || toppedUpSpent > 0) {
+    const topUpTotal = topUpBalance + toppedUpSpent;
+    quotas["Top-up"] = {
+      used: toppedUpSpent,
+      total: topUpTotal,
+      remaining: topUpBalance,
+      remainingPercentage: topUpTotal > 0 ? Math.round((topUpBalance / topUpTotal) * 100) : null,
+      resetAt: null,
+    };
+  }
+  if (voucherBalance > 0 || voucherSpent > 0) {
+    const voucherTotal = voucherBalance + voucherSpent;
+    quotas["Voucher"] = {
+      used: voucherSpent,
+      total: voucherTotal,
+      remaining: voucherBalance,
+      remainingPercentage: voucherTotal > 0 ? Math.round((voucherBalance / voucherTotal) * 100) : null,
+      resetAt: null,
+    };
+  }
+
+  return {
+    quotas,
+    plan: null,
+    credits: { remaining: totalBalance, used: totalSpent, total },
+  };
+}
diff --git a/open-sse/services/usage/xai.js b/open-sse/services/usage/xai.js
index 8a368de..94f8259 100644
--- a/open-sse/services/usage/xai.js
+++ b/open-sse/services/usage/xai.js
@@ -1,98 +1,62 @@
 // xAI (Grok) usage handler for Quota Tracker.
 //
-// Uses two read-only endpoints with the OAuth access token:
-//   1. GET /v1/billing?format=credits — on-demand credits (remaining, used, total)
-//   2. GET /v1/user?include=subscription — active subscription plan name
+// STATUS: DISABLED — xAI removed the public billing/subscription endpoints from
+// the inference API (api.x.ai). Both endpoints below now return HTTP 404:
+//   - GET /v1/billing?format=credits  -> 404 (removed)
+//   - GET /v1/user?include=subscription -> 404 (removed)
 //
-// Only OAuth connections are eligible (the token scope includes grok-cli:access).
-// API-key xAI connections are excluded by the USAGE_SUPPORTED_PROVIDERS gate.
+// xAI migrated billing & usage tracking to a SEPARATE Management API on a
+// different host (https://management-api.x.ai) that requires a dedicated
+// Management API key — distinct from both the chat API key and the OAuth
+// access token. The Management API is only available to account owners via
+// console.x.ai → Settings → Management Keys.
 //
-// Reference: github.com/decolua/9router PR #2672
-
-import { U, parseResetTime, toFiniteNumber } from "./shared.js";
-import { proxyAwareFetch } from "../../utils/proxyFetch.js";
-
-const USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36";
-
-export async function getXaiUsage(credentials, proxyOptions = null) {
-  const token = credentials?.accessToken;
-  if (!token) return null;
-
-  const cfg = U("xai");
-  const headers = {
-    Authorization: `Bearer ${token}`,
-    Accept: "application/json",
-    "User-Agent": USER_AGENT,
-  };
-
-  // Fetch billing + subscription in parallel
-  const [billingRes, subRes] = await Promise.allSettled([
-    cfg.billingUrl
-      ? proxyAwareFetch(cfg.billingUrl, { method: "GET", headers }, proxyOptions)
-      : Promise.reject(new Error("no billingUrl")),
-    cfg.subscriptionUrl
-      ? proxyAwareFetch(cfg.subscriptionUrl, { method: "GET", headers }, proxyOptions)
-      : Promise.reject(new Error("no subscriptionUrl")),
-  ]);
-
-  let credits = null;
-  let plan = null;
-
-  // Parse billing response (credits data)
-  if (billingRes.status === "fulfilled" && billingRes.value?.ok) {
-    const billingData = await billingRes.value.json().catch(() => null);
-    if (billingData) {
-      const remaining = toFiniteNumber(billingData.credits ?? billingData.remaining_credits);
-      const used = toFiniteNumber(billingData.used_credits ?? billingData.used);
-      const total = toFiniteNumber(billingData.total_credits ?? billingData.total);
-
-      if (remaining > 0 || total > 0 || used > 0) {
-        credits = {
-          remaining,
-          used,
-          total: total || (remaining + used),
-          remainingPercentage: total > 0 ? Math.round((remaining / total) * 100) : null,
-          resetAt: parseResetTime(billingData.reset_at ?? billingData.reset_date) || null,
-        };
-      }
-    }
-  }
-
-  // Parse subscription response (plan name)
-  if (subRes.status === "fulfilled" && subRes.value?.ok) {
-    const subData = await subRes.value.json().catch(() => null);
-    if (subData) {
-      const sub = subData.subscription || subData;
-      plan = sub.plan || sub.tier || sub.name || null;
-    }
-  }
-
-  // Build quotas object
-  const quotas = {};
-
-  if (credits) {
-    quotas["Credits"] = {
-      used: credits.used,
-      total: credits.total,
-      remaining: credits.remaining,
-      remainingPercentage: credits.remainingPercentage,
-      resetAt: credits.resetAt,
-    };
-  }
-
-  // If no numeric credit data but we have a plan, show it as info-only quota
-  if (Object.keys(quotas).length === 0 && plan) {
-    quotas[plan] = {
-      used: 0,
-      total: 0,
-      unlimited: true,
-      remainingPercentage: null,
+// As of 2026-07, there is no public REST endpoint reachable with the
+// credentials stored in an ExtremeRouter connection. Until xAI documents a
+// public billing endpoint OR ExtremeRouter adds a Management Key field (same
+// two-key design as TokenRouter), this handler returns an informational
+// message so the Quota Tracker card explains WHY it is empty instead of
+// appearing blank.
+//
+// Reference: github.com/decolua/9router PR #2672 (original implementation
+// against the now-defunct endpoints).
+
+import { U } from "./shared.js";
+
+// Kept for forward-compatibility: if a Management API key is ever wired up
+// (providerSpecificData.mgmtKey, mirroring TokenRouter), this handler can be
+// re-enabled against https://management-api.x.ai/billing without touching the
+// registration in services/usage.js.
+export async function getXaiUsage(credentials, _proxyOptions = null) {
+  // Sanity reference — confirms the provider config still exists. Intentionally
+  // unused for network calls while the handler is in the informational state.
+  void U;
+
+  const hasMgmtKey = Boolean(
+    credentials?.providerSpecificData?.mgmtKey ||
+      credentials?.providerDataWithProjectId?.mgmtKey
+  );
+
+  if (hasMgmtKey) {
+    // Forward-compat hook: if a Management Key is present we still cannot reach
+    // a documented billing endpoint today, but we acknowledge the credential
+    // so the message can guide the user.
+    return {
+      quotas: {},
+      plan: null,
+      credits: null,
+      message:
+        "xAI Management API key detected, but the billing endpoint is not yet implemented. " +
+        "Usage tracking for xAI is pending upstream documentation.",
     };
   }
 
   return {
-    quotas,
-    plan,
-    credits,
+    quotas: {},
+    plan: null,
+    credits: null,
+    message:
+      "xAI removed the public billing API. Usage tracking is only available in the " +
+      "xAI Console at console.x.ai → Billing.",
   };
 }
diff --git a/package.json b/package.json
index 303751a..6b1a749 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@rsalmn/extremerouter-app",
-  "version": "0.7.3",
+  "version": "0.7.4",
   "description": "ExtremeRouter web dashboard",
   "private": true,
   "license": "MIT",
@@ -42,6 +42,7 @@
     "@xyflow/react": "^12.10.1",
     "bcryptjs": "^3.0.3",
     "confbox": "^0.2.4",
+    "dompurify": "^3.4.12",
     "express": "^5.2.1",
     "fs": "^0.0.1-security",
     "http-proxy-middleware": "^3.0.5",
@@ -74,6 +75,7 @@
     "@tailwindcss/postcss": "^4.1.18",
     "eslint": "^9",
     "eslint-config-next": "16.1.6",
+    "jsdom": "^29.1.1",
     "postcss": "^8.5.6",
     "tailwindcss": "^4"
   }
diff --git a/public/providers/alibaba-cn.svg b/public/providers/alibaba-cn.svg
new file mode 100644
index 0000000..fcb6174
--- /dev/null
+++ b/public/providers/alibaba-cn.svg
@@ -0,0 +1,11 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#1A1A2E"/>
+  <!-- Alibaba orange smile/curl shape -->
+  <path d="M16 22 C16 22 24 14 32 14 C40 14 48 22 48 22" fill="none" stroke="#FF6A00" stroke-width="4" stroke-linecap="round"/>
+  <path d="M14 38 C14 38 24 32 32 32 C40 32 50 38 50 38" fill="none" stroke="#FF6A00" stroke-width="4" stroke-linecap="round"/>
+  <!-- Eyes -->
+  <circle cx="24" cy="28" r="2.5" fill="#FF6A00"/>
+  <circle cx="40" cy="28" r="2.5" fill="#FF6A00"/>
+  <!-- CN badge -->
+  <text x="32" y="54" text-anchor="middle" font-family="system-ui,sans-serif" font-size="10" font-weight="bold" fill="#FF6A00">CN</text>
+</svg>
diff --git a/public/providers/alibaba.svg b/public/providers/alibaba.svg
new file mode 100644
index 0000000..db6b3c6
--- /dev/null
+++ b/public/providers/alibaba.svg
@@ -0,0 +1,11 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#1A1A2E"/>
+  <!-- Alibaba orange smile/curl shape -->
+  <path d="M16 22 C16 22 24 14 32 14 C40 14 48 22 48 22" fill="none" stroke="#FF6A00" stroke-width="4" stroke-linecap="round"/>
+  <path d="M14 38 C14 38 24 32 32 32 C40 32 50 38 50 38" fill="none" stroke="#FF6A00" stroke-width="4" stroke-linecap="round"/>
+  <!-- Eyes -->
+  <circle cx="24" cy="28" r="2.5" fill="#FF6A00"/>
+  <circle cx="40" cy="28" r="2.5" fill="#FF6A00"/>
+  <!-- "A" letter -->
+  <text x="32" y="54" text-anchor="middle" font-family="system-ui,sans-serif" font-size="10" font-weight="bold" fill="#FF6A00">ALIBABA</text>
+</svg>
diff --git a/public/providers/forge.svg b/public/providers/forge.svg
new file mode 100644
index 0000000..72cffe1
--- /dev/null
+++ b/public/providers/forge.svg
@@ -0,0 +1,9 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#1A1A2E"/>
+  <!-- Flame / fire shape representing "Forge" -->
+  <path d="M32 8 C32 8 22 20 22 32 C22 38 26 42 32 44 C38 42 42 38 42 32 C42 20 32 8 32 8Z" fill="none" stroke="#FF6B35" stroke-width="3" stroke-linecap="round" stroke-linejoin="round"/>
+  <path d="M32 18 C32 18 26 26 26 34 C26 38 28 40 32 41 C36 40 38 38 38 34 C38 26 32 18 32 18Z" fill="#FF6B35" opacity="0.8"/>
+  <!-- Anvil base -->
+  <path d="M18 48 L26 44 L38 44 L46 48 L46 52 L18 52Z" fill="#FF6B35" opacity="0.6"/>
+  <rect x="18" y="52" width="28" height="4" rx="2" fill="#FF6B35" opacity="0.4"/>
+</svg>
diff --git a/public/providers/hcnsec.svg b/public/providers/hcnsec.svg
new file mode 100644
index 0000000..0ed47ce
--- /dev/null
+++ b/public/providers/hcnsec.svg
@@ -0,0 +1,10 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#0A1929"/>
+  <!-- Shield (cybersecurity branding) -->
+  <path d="M32 8 L52 16 L52 32 C52 44 42 52 32 56 C22 52 12 44 12 32 L12 16 Z" fill="none" stroke="#0EA5E9" stroke-width="3" stroke-linejoin="round"/>
+  <!-- HC monogram -->
+  <text x="32" y="38" text-anchor="middle" font-family="system-ui,sans-serif" font-size="16" font-weight="bold" fill="#0EA5E9">HC</text>
+  <!-- Lock detail at bottom of shield -->
+  <rect x="28" y="44" width="8" height="6" rx="1" fill="#0EA5E9" opacity="0.7"/>
+  <path d="M29 44 V41 C29 39.5 30 38 32 38 C34 38 35 39.5 35 41 V44" fill="none" stroke="#0EA5E9" stroke-width="1.5" opacity="0.7"/>
+</svg>
diff --git a/public/providers/qwen-cloud-token-plan.svg b/public/providers/qwen-cloud-token-plan.svg
new file mode 100644
index 0000000..6a72fc4
--- /dev/null
+++ b/public/providers/qwen-cloud-token-plan.svg
@@ -0,0 +1,8 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#0F1A2E"/>
+  <!-- Cloud shape -->
+  <path d="M20 40 C14 40 10 36 10 31 C10 26 14 22 19 22 C20 16 25 12 31 12 C38 12 43 17 44 23 C49 23 53 27 53 32 C53 37 49 40 44 40 Z" fill="none" stroke="#615CED" stroke-width="3" stroke-linejoin="round"/>
+  <!-- Token coin overlay -->
+  <circle cx="44" cy="42" r="10" fill="#FFB800" stroke="#0F1A2E" stroke-width="2"/>
+  <text x="44" y="47" text-anchor="middle" font-family="system-ui,sans-serif" font-size="11" font-weight="bold" fill="#0F1A2E">T</text>
+</svg>
diff --git a/public/providers/qwen-cloud.svg b/public/providers/qwen-cloud.svg
new file mode 100644
index 0000000..6e966ee
--- /dev/null
+++ b/public/providers/qwen-cloud.svg
@@ -0,0 +1,8 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#0F1A2E"/>
+  <!-- Cloud shape (Qwen Cloud branding) -->
+  <path d="M20 40 C14 40 10 36 10 31 C10 26 14 22 19 22 C20 16 25 12 31 12 C38 12 43 17 44 23 C49 23 53 27 53 32 C53 37 49 40 44 40 Z" fill="none" stroke="#615CED" stroke-width="3" stroke-linejoin="round"/>
+  <!-- "Q" letter inside cloud -->
+  <circle cx="32" cy="30" r="6" fill="none" stroke="#615CED" stroke-width="2.5"/>
+  <line x1="36" y1="34" x2="40" y2="38" stroke="#615CED" stroke-width="2.5" stroke-linecap="round"/>
+</svg>
diff --git a/public/providers/tokenrouter.svg b/public/providers/tokenrouter.svg
new file mode 100644
index 0000000..0877166
--- /dev/null
+++ b/public/providers/tokenrouter.svg
@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" width="64" height="64">
+  <rect width="64" height="64" rx="14" fill="#0F1F1A"/>
+  <!-- Router box with token slot -->
+  <rect x="14" y="22" width="36" height="22" rx="4" fill="none" stroke="#10B981" stroke-width="3"/>
+  <!-- Token slot (top) -->
+  <rect x="26" y="16" width="12" height="6" rx="1" fill="#10B981"/>
+  <!-- LED indicators -->
+  <circle cx="22" cy="33" r="2.5" fill="#10B981"/>
+  <circle cx="32" cy="33" r="2.5" fill="#10B981" opacity="0.6"/>
+  <circle cx="42" cy="33" r="2.5" fill="#10B981" opacity="0.3"/>
+  <!-- Signal waves below -->
+  <path d="M20 50 Q24 48 24 50" fill="none" stroke="#10B981" stroke-width="2" stroke-linecap="round"/>
+  <path d="M28 52 Q32 48 32 52" fill="none" stroke="#10B981" stroke-width="2" stroke-linecap="round" opacity="0.6"/>
+  <path d="M36 50 Q40 48 40 50" fill="none" stroke="#10B981" stroke-width="2" stroke-linecap="round" opacity="0.4"/>
+</svg>
diff --git a/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.js b/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.js
index 7f781ec..031caad 100644
--- a/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.js
+++ b/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.js
@@ -1318,10 +1318,18 @@ export default function APIPageClient({ machineId }) {
 
 
 function ModelAccessModal({ apiKey, onClose, onSaved }) {
-  const [allowedModels, setAllowedModels] = useState(null);
+  // IMPORTANT: keep the early return AFTER all hooks. We initialize
+  // allowedModels directly from apiKey so the very first render already has
+  // a stable array (never null), and we resync via useEffect when apiKey
+  // changes. Returning null before useState/useEffect would violate the
+  // Rules of Hooks (conditional hook execution) and crash React.
+  const [allowedModels, setAllowedModels] = useState(
+    () => (Array.isArray(apiKey?.allowedModels) ? [...apiKey.allowedModels] : []),
+  );
   const [modelInput, setModelInput] = useState("");
   const [saving, setSaving] = useState(false);
 
+  // Reset the local draft whenever a different apiKey is opened.
   useEffect(() => {
     if (apiKey) {
       setAllowedModels(Array.isArray(apiKey.allowedModels) ? [...apiKey.allowedModels] : []);
@@ -1334,7 +1342,7 @@ function ModelAccessModal({ apiKey, onClose, onSaved }) {
   const addModel = () => {
     const m = modelInput.trim();
     if (!m) return;
-    if (!allowedModels.includes(m)) setAllowedModels((prev) => [...prev, m]);
+    setAllowedModels((prev) => (prev.includes(m) ? prev : [...prev, m]));
     setModelInput("");
   };
 
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/AddApiKeyModal.js b/src/app/(dashboard)/dashboard/providers/[id]/AddApiKeyModal.js
index 81b86a3..4000f22 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/AddApiKeyModal.js
+++ b/src/app/(dashboard)/dashboard/providers/[id]/AddApiKeyModal.js
@@ -12,6 +12,7 @@ export default function AddApiKeyModal({ isOpen, provider, providerName, isCompa
   const isOllamaLocal = provider === "ollama-local";
   const isCookie = authType === "cookie";
   const isXaiApiKey = provider === "xai" && !isCookie;
+  const isTokenRouter = provider === "tokenrouter";
   const credentialLabel = isCookie ? "Cookie Value" : "API Key";
   const credentialPlaceholder = isCookie
     ? (provider === "grok-web"
@@ -34,6 +35,8 @@ export default function AddApiKeyModal({ isOpen, provider, providerName, isCompa
     proxyPoolId: NONE_PROXY_POOL_VALUE,
     ollamaHostUrl: "",
   });
+  // TokenRouter management key — separate from chat API key, used by Quota Tracker.
+  const [tokenRouterMgmtKey, setTokenRouterMgmtKey] = useState("");
   const [azureData, setAzureData] = useState({
     azureEndpoint: "",
     apiVersion: "2024-10-01-preview",
@@ -64,6 +67,9 @@ export default function AddApiKeyModal({ isOpen, provider, providerName, isCompa
     if (isCloudflareAi) {
       return { accountId: cloudflareData.accountId };
     }
+    if (isTokenRouter && tokenRouterMgmtKey.trim()) {
+      return { mgmtKey: tokenRouterMgmtKey.trim() };
+    }
     if (providerRegions && region) {
       return { region };
     }
@@ -302,6 +308,23 @@ export default function AddApiKeyModal({ isOpen, provider, providerName, isCompa
             Enter the model ID exactly as your compatible endpoint expects it. This model will be saved as the connection default.
           </p>
         )}
+        {isTokenRouter && (
+          <div className="bg-sidebar/50 p-4 rounded-lg border border-accent/20">
+            <h3 className="font-semibold mb-3 text-sm">TokenRouter Management Key (optional)</h3>
+            <Input
+              label="Management Key"
+              type="password"
+              value={tokenRouterMgmtKey}
+              onChange={(e) => setTokenRouterMgmtKey(e.target.value)}
+              placeholder="your_management_key..."
+            />
+            <p className="text-xs text-text-muted mt-2">
+              The management key is separate from the chat API key (sk-...) and is required for
+              Quota Tracker to read your wallet balance. Without it, chat still works but quota
+              tracking is disabled. Get it from the TokenRouter dashboard.
+            </p>
+          </div>
+        )}
         {isCloudflareAi && (
           <div className="bg-sidebar/50 p-4 rounded-lg border border-accent/20">
             <h3 className="font-semibold mb-3 text-sm">Cloudflare Workers AI</h3>
diff --git a/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/index.js b/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/index.js
index 2d43465..23d2c3b 100644
--- a/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/index.js
+++ b/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/index.js
@@ -261,13 +261,18 @@ export default function ProviderLimits() {
       const data = await response.json();
       console.log(`[ProviderLimits] Got quota for ${provider}:`, data);
 
+      // Some usage handlers legitimately return `null` (e.g. TokenRouter when
+      // the user hasn't configured a management key). Guard against null so the
+      // UI doesn't crash reading `data.plan` — show an empty result instead.
+      const safeData = data ?? {};
+
       // Parse quota data using provider-specific parser
-      const parsedQuotas = parseQuotaData(provider, data);
+      const parsedQuotas = parseQuotaData(provider, safeData);
 
       const quotaEntry = {
         quotas: parsedQuotas,
-        plan: data.plan || null,
-        message: data.message || null,
+        plan: safeData.plan || null,
+        message: safeData.message || null,
         raw: data,
       };
 
diff --git a/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/utils.js b/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/utils.js
index 97ad0fa..b90b816 100644
--- a/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/utils.js
+++ b/src/app/(dashboard)/dashboard/quota/components/ProviderLimits/utils.js
@@ -481,6 +481,24 @@ export function parseQuotaData(provider, data) {
         }
         break;
       }
+      case "tokenrouter": {
+        // TokenRouter — wallet (top-up + voucher) from management API.
+        if (data.quotas) {
+          Object.entries(data.quotas).forEach(([name, quota]) => {
+            normalizedQuotas.push({
+              name,
+              used: quota.used || 0,
+              total: quota.total || 0,
+              remaining: quota.remaining,
+              remainingPercentage: quota.remainingPercentage,
+              resetAt: quota.resetAt || null,
+              recurring: false,
+              unlimited: false,
+            });
+          });
+        }
+        break;
+      }
       default:
         // Generic fallback for unknown providers
         if (data.quotas) {
diff --git a/src/dashboardGuard.js b/src/dashboardGuard.js
index 83e4364..07fa5eb 100644
--- a/src/dashboardGuard.js
+++ b/src/dashboardGuard.js
@@ -204,6 +204,27 @@ export async function proxy(request) {
   // Deny-by-default for /api/* — public allow-list bypasses, everything else requires auth.
   if (pathname.startsWith("/api/")) {
     if (isPublicApi(pathname)) return NextResponse.next();
+
+    // C1 FIX: Mutation routes (POST/PUT/PATCH/DELETE) ALWAYS require real
+    // authentication (JWT session or CLI token), even when requireLogin is
+    // disabled. Without this, an attacker who reaches the network when
+    // requireLogin=false can PATCH /api/settings to set requireApiKey=false,
+    // effectively disabling LLM API key auth for all external callers.
+    //
+    // Read-only routes (GET/HEAD) keep the existing bypass: if requireLogin
+    // is disabled, the dashboard is accessible without login — this is safe
+    // because no state is changed and no secrets are exposed.
+    const isMutation = !["GET", "HEAD", "OPTIONS"].includes(request.method);
+    if (isMutation) {
+      if (await hasValidCliToken(request) || await hasValidToken(request))
+        return NextResponse.next();
+      return NextResponse.json(
+        { error: "Authentication required for this action" },
+        { status: 401 }
+      );
+    }
+
+    // Read-only GET/HEAD: allow if requireLogin is disabled.
     if (await hasValidCliToken(request) || await isAuthenticated(request))
       return NextResponse.next();
     return NextResponse.json({ error: "Unauthorized" }, { status: 401 });
diff --git a/src/shared/utils/providerIcon.js b/src/shared/utils/providerIcon.js
index 30b85ac..d7c83f8 100644
--- a/src/shared/utils/providerIcon.js
+++ b/src/shared/utils/providerIcon.js
@@ -22,7 +22,8 @@ export const SVG_ICON_IDS = new Set([
   "claude-web", "pollinations", "poe-web", "v0-vercel-web", "qwen-web",
   "kimi-web", "huggingchat", "api-airforce", "openvecta", "freebuff-web",
   "zenmux-free", "perplexity-agent", "featherless", "moonshot", "qwencloud",
-  "devin",
+  "devin", "forge", "tokenrouter",
+  "qwen-cloud", "qwen-cloud-token-plan", "alibaba", "alibaba-cn", "hcnsec",
 ]);
 
 /**
diff --git a/src/shared/utils/sanitizeHtml.js b/src/shared/utils/sanitizeHtml.js
index 7e62fd8..81bfcfd 100644
--- a/src/shared/utils/sanitizeHtml.js
+++ b/src/shared/utils/sanitizeHtml.js
@@ -1,34 +1,53 @@
-// Minimal HTML sanitizer for rendering model output as markdown.
-//
-// This is DEFENSE-IN-DEPTH, not a primary security control. The primary
-// defense is that `marked` produces a constrained AST (code blocks are
-// escaped text, not raw HTML by default). This sanitizer catches residual
-// dangerous markup that could slip through (e.g. raw HTML embedded in
-// markdown, or a future marked config change that allows raw HTML).
-//
-// It is deliberately conservative and regex-based to avoid pulling in a
-// heavyweight DOM-parser dependency. It strips:
-//   - <script>, <iframe>, <object>, <embed>, <form>, <style> elements entirely
-//   - on* event-handler attributes (onclick, onerror, ...)
-//   - javascript: / vbscript: / data:text/html URLs in href/src
-//
-// For untrusted model output this is acceptable; if you ever need to render
-// arbitrary untrusted HTML from a non-markdown source, use a real DOM parser
-// (DOMPurify) instead.
-
-const DANGEROUS_TAGS = /<\/?(script|iframe|object|embed|form|style|link|meta|base|applet)\b[^>]*>/gi;
-const EVENT_HANDLER_ATTRS = /\son\w+\s*=\s*(?:"[^"]*"|'[^']*'|[^\s>]+)/gi;
-const DANGEROUS_URL_SCHEMES = /(href|src|xlink:href)\s*=\s*("(?:javascript|vbscript|data:text\/html)[^"]*"|'(?:javascript|vbscript|data:text\/html)[^']*')/gi;
+import DOMPurify from "dompurify";
 
 /**
- * Sanitize an HTML string produced by a markdown parser.
- * @param {string} html
- * @returns {string} sanitized HTML, safe for dangerouslySetInnerHTML
+ * Sanitize HTML produced from markdown before injecting via dangerouslySetInnerHTML.
+ *
+ * Uses DOMPurify — a battle-tested, browser-native DOM parser sanitizer.
+ * The previous regex-based approach could be bypassed via nested tags
+ * (`<scr<script>ipt>`), unquoted attributes, and entity-encoded handlers.
+ * DOMPurify parses into a real DOM tree and rebuilds a clean serialization,
+ * making those evasion classes structurally impossible.
+ *
+ * Output from an LLM is untrusted content: prompt injection can force a model
+ * to emit hostile markup. This runs in an admin dashboard where XSS could
+ * perform actions on behalf of the operator.
+ *
+ * Hardening applied:
+ * - FORBID_TAGS: style (CSS exfil), form/input/textarea/button/select
+ *   (credential harvesting UI), svg (can carry script even in data: URIs).
+ * - FORBID_ATTR: style (inline CSS can url()-exfiltrate).
+ * - ALLOW_DATA_ATTR false is default; data: URIs restricted below.
+ * - Links get rel="noopener noreferrer" + target="_blank" via hook so
+ *   model-rendered links can't tabnab or leak the dashboard URL as referrer.
+ *
+ * @param {string} html - Raw HTML string (typically from a markdown renderer).
+ * @returns {string} Sanitized HTML safe for dangerouslySetInnerHTML.
  */
+DOMPurify.addHook("afterSanitizeAttributes", (node) => {
+  if (node.tagName === "A") {
+    node.setAttribute("target", "_blank");
+    node.setAttribute("rel", "noopener noreferrer");
+  }
+  // Deterministic data: URI policy — regex-based ALLOWED_URI_REGEXP is leaky
+  // (its `[^a-z]` clause passes almost anything). Instead, strip any data:
+  // attribute value that isn't an inert raster image. SVG deliberately
+  // excluded: it can carry script/onload even inside a data: URI.
+  for (const attr of ["src", "href", "xlink:href", "action", "formaction"]) {
+    const val = node.getAttribute(attr);
+    if (val && /^\s*data:/i.test(val) && !/^\s*data:image\/(?:png|jpe?g|gif|webp);/i.test(val)) {
+      node.removeAttribute(attr);
+    }
+  }
+});
+
+const SANITIZE_CONFIG = {
+  FORBID_TAGS: ["style", "form", "input", "textarea", "button", "select", "svg"],
+  FORBID_ATTR: ["style"],
+  ALLOW_UNKNOWN_PROTOCOLS: false,
+};
+
 export function sanitizeHtml(html) {
-  if (typeof html !== "string" || html.length === 0) return "";
-  return html
-    .replace(DANGEROUS_TAGS, "")
-    .replace(EVENT_HANDLER_ATTRS, "")
-    .replace(DANGEROUS_URL_SCHEMES, "");
+  if (typeof html !== "string" || !html) return "";
+  return DOMPurify.sanitize(html, SANITIZE_CONFIG);
 }
diff --git a/tests/unit/sanitize-html.test.js b/tests/unit/sanitize-html.test.js
new file mode 100644
index 0000000..460bf3e
--- /dev/null
+++ b/tests/unit/sanitize-html.test.js
@@ -0,0 +1,90 @@
+/**
+ * Adversarial self-check for the DOMPurify-based sanitizeHtml.
+ *
+ * Asserts that attack vectors which bypassed the previous regex-based
+ * sanitizer are now neutralized. Run with: node tests/unit/sanitize-html.test.js
+ * (or as a plain script — it throws on failure, prints PASS on success).
+ *
+ * Note: DOMPurify requires a DOM. Under Node we provide one via jsdom if
+ * available; the test harness (vitest with environment:'jsdom') supplies it
+ * automatically. When run as a bare script without jsdom we skip gracefully.
+ */
+
+let sanitizeHtml;
+let skipped = false;
+
+try {
+  // Prefer a real browser-like DOM if the harness provides one (vitest jsdom).
+  if (typeof window === "undefined" || typeof window.DOMParser === "undefined") {
+    let JSDOM;
+    try {
+      ({ JSDOM } = await import("jsdom"));
+    } catch {
+      skipped = true;
+      console.log("SKIP: jsdom not available — install jsdom or run under vitest with environment 'jsdom'.");
+    }
+    if (!skipped) {
+      const dom = new JSDOM("<!doctype html><html><body></body></html>");
+      globalThis.window = dom.window;
+      globalThis.document = dom.window.document;
+      globalThis.DOMParser = dom.window.DOMParser;
+    }
+  }
+  if (!skipped) {
+    ({ sanitizeHtml } = await import("../../src/shared/utils/sanitizeHtml.js"));
+  }
+} catch (err) {
+  console.error("SETUP FAIL:", err.message);
+  process.exit(1);
+}
+
+if (!skipped) {
+  const cases = [
+    // [name, input, predicate(out) => true if SAFE]
+    ["strips <script>", '<p>hi</p><script>alert(1)</script>', (o) => !o.includes("<script")],
+    // Nested-tag evasion: safe if no executable markup remains. The alert text
+    // surviving as inert, HTML-encoded plain text is correct behavior.
+    ["blocks nested-tag bypass <scr<script>ipt>", "<scr<script>ipt>alert(1)</scr</script>ipt>", (o) => !/<script/i.test(o) && !/<[a-z]+[^>]*on\w+=/i.test(o)],
+    ["strips onerror handler", '<img src=x onerror="alert(1)">', (o) => !/onerror/i.test(o)],
+    ["strips unquoted onerror", "<img src=x onerror=alert(1)>", (o) => !/onerror/i.test(o)],
+    ["strips entity-encoded handler", '<img src=x &#111;nerror="alert(1)">', (o) => !/onerror=/i.test(o)],
+    ["strips onclick on div", '<div onclick="alert(1)">x</div>', (o) => !/onclick/i.test(o)],
+    ["blocks javascript: href", '<a href="javascript:alert(1)">x</a>', (o) => !/javascript:/i.test(o)],
+    ["blocks entity-encoded javascript:", '<a href="&#106;avascript:alert(1)">x</a>', (o) => !/javascript:/i.test(o)],
+    ["blocks vbscript:", '<a href="vbscript:msgbox(1)">x</a>', (o) => !/vbscript:/i.test(o)],
+    ["blocks svg data: in img src", '<img src="data:image/svg+xml,<svg onload=alert(1)>">', (o) => !/onload/i.test(o) && !/svg/i.test(o)],
+    ["keeps raster data: images", '<img src="data:image/png;base64,iVBORw0KGgo=">', (o) => /data:image\/png/.test(o)],
+    ["strips <style> tag", '<style>body{background:url(http://evil)}</style><p>x</p>', (o) => !/<style/i.test(o)],
+    ["strips style attribute", '<p style="background:url(javascript:alert(1))">x</p>', (o) => !/style=/i.test(o)],
+    ["strips <form>", '<form action="http://evil"><input name=p></form>', (o) => !/<form/i.test(o) && !/<input/i.test(o)],
+    ["strips <iframe>", '<iframe src="http://evil"></iframe>', (o) => !/<iframe/i.test(o)],
+    ["strips <object>", '<object data="http://evil"></object>', (o) => !/<object/i.test(o)],
+    ["strips <svg onload>", '<svg onload="alert(1)"><circle/></svg>', (o) => !/onload/i.test(o) && !/alert\(1\)/.test(o)],
+    ["adds noopener to links", '<a href="https://ok.example">x</a>', (o) => /rel="noopener noreferrer"/.test(o) && /target="_blank"/.test(o)],
+    ["keeps safe markdown html", '<p>Hello <strong>world</strong></p>', (o) => o.includes("<strong>world</strong>")],
+    ["keeps code blocks", '<pre><code class="language-js">const a = 1;</code></pre>', (o) => o.includes("const a = 1;")],
+    ["empty string safe", "", (o) => o === ""],
+    ["non-string safe", null, (o) => o === ""],
+  ];
+
+  let failures = 0;
+  for (const [name, input, isSafe] of cases) {
+    let out;
+    try {
+      out = sanitizeHtml(input);
+    } catch (err) {
+      console.log(`FAIL  ${name} — threw: ${err.message}`);
+      failures++;
+      continue;
+    }
+    if (isSafe(out)) {
+      console.log(`PASS  ${name}`);
+    } else {
+      console.log(`FAIL  ${name}\n      in : ${JSON.stringify(input)}\n      out: ${JSON.stringify(out)}`);
+      failures++;
+    }
+  }
+
+  console.log(failures === 0 ? `\n✅ ALL ${cases.length} CASES PASS` : `\n❌ ${failures}/${cases.length} FAILURES`);
+  process.exit(failures === 0 ? 0 : 1);
+}