@blockrun/clawrouter 0.3.41 → 0.4.2

package/README.md

@@ -12,7 +12,7 @@ One wallet, 30+ models, zero API keys.
 [![TypeScript](https://img.shields.io/badge/TypeScript-5.7-blue.svg)](https://typescriptlang.org)
 [![Node](https://img.shields.io/badge/node-%E2%89%A520-brightgreen.svg)](https://nodejs.org)
 
-[Docs](https://blockrun.ai/docs) · [Models](https://blockrun.ai/models) · [Telegram](https://t.me/blockrunAI) · [X](https://x.com/BlockRunAI)
+[Docs](https://blockrun.ai/docs) · [Models](https://blockrun.ai/models) · [Configuration](docs/configuration.md) · [Architecture](docs/architecture.md) · [Telegram](https://t.me/blockrunAI) · [X](https://x.com/BlockRunAI)
 
 </div>
 
@@ -283,35 +283,31 @@ If you explicitly want to use a different wallet:
 
 Routing is **client-side** — open source and inspectable.
 
-### Source Structure
-
-```
-src/
-├── index.ts          # Plugin entry point
-├── provider.ts       # OpenClaw provider registration
-├── proxy.ts          # Local HTTP proxy + x402 payment
-├── models.ts         # 30+ model definitions with pricing
-├── auth.ts           # Wallet key resolution
-├── logger.ts         # JSON usage logging
-├── dedup.ts          # Response deduplication (prevents double-charge)
-├── payment-cache.ts  # Pre-auth optimization (skips 402 round trip)
-├── x402.ts           # EIP-712 USDC payment signing
-└── router/
-    ├── index.ts      # route() entry point
-    ├── rules.ts      # 14-dimension weighted scoring
-    ├── selector.ts   # Tier → model selection
-    ├── config.ts     # Default routing config
-    └── types.ts      # TypeScript types
-```
+**Deep dive:** [docs/architecture.md](docs/architecture.md) — request flow, payment system, optimizations
 
 ---
 
 ## Configuration
 
-### Override Tier Models
+For basic usage, no configuration is needed. For advanced options:
+
+| Setting               | Default  | Description                  |
+| --------------------- | -------- | ---------------------------- |
+| `BLOCKRUN_PROXY_PORT` | `8402`   | Proxy port (env var)         |
+| `BLOCKRUN_WALLET_KEY` | auto     | Wallet private key (env var) |
+| `routing.tiers`       | see docs | Override tier→model mappings |
+| `routing.scoring`     | see docs | Custom keyword weights       |
+
+**Quick example:**
+
+```bash
+# Use different port
+export BLOCKRUN_PROXY_PORT=8403
+openclaw gateway restart
+```
 
 ```yaml
-# openclaw.yaml
+# openclaw.yaml — override models
 plugins:
   - id: "@blockrun/clawrouter"
     config:
@@ -319,18 +315,9 @@ plugins:
         tiers:
           COMPLEX:
             primary: "openai/gpt-4o"
-          SIMPLE:
-            primary: "google/gemini-2.5-flash"
 ```
 
-### Override Scoring Weights
-
-```yaml
-routing:
-  scoring:
-    reasoningKeywords: ["proof", "theorem", "formal verification"]
-    codeKeywords: ["function", "class", "async", "await"]
-```
+**Full reference:** [docs/configuration.md](docs/configuration.md)
 
 ---
 
@@ -478,6 +465,18 @@ See [`openclaw.security.json`](openclaw.security.json) for detailed security doc
 
 ### Port 8402 already in use
 
+As of v0.4.1, ClawRouter automatically detects and reuses an existing proxy on the configured port instead of failing with `EADDRINUSE`. You should no longer see this error.
+
+If you need to use a different port:
+
+```bash
+# Set custom port via environment variable
+export BLOCKRUN_PROXY_PORT=8403
+openclaw gateway restart
+```
+
+To manually check/kill the process:
+
 ```bash
 lsof -i :8402
 # Kill the process or restart OpenClaw

package/dist/index.d.ts

@@ -363,6 +363,10 @@ declare class BalanceMonitor {
  *   - Usage logging: log every request as JSON line to ~/.openclaw/blockrun/logs/
  */
 
+/**
+ * Get the proxy port from environment variable or default.
+ */
+declare function getProxyPort(): number;
 /** Callback info for low balance warning */
 type LowBalanceInfo = {
     balanceUSD: string;
@@ -382,6 +386,8 @@ type ProxyOptions = {
     routingConfig?: Partial<RoutingConfig>;
     /** Request timeout in ms (default: 180000 = 3 minutes). Covers on-chain tx + LLM response. */
     requestTimeoutMs?: number;
+    /** Skip balance checks (for testing only). Default: false */
+    skipBalanceCheck?: boolean;
     onReady?: (port: number) => void;
     onError?: (error: Error) => void;
     onPayment?: (info: {
@@ -405,6 +411,9 @@ type ProxyHandle = {
 /**
  * Start the local x402 proxy server.
  *
+ * If a proxy is already running on the target port, reuses it instead of failing.
+ * Port can be configured via BLOCKRUN_PROXY_PORT environment variable.
+ *
  * Returns a handle with the assigned port, base URL, and a close function.
  */
 declare function startProxy(options: ProxyOptions): Promise<ProxyHandle>;
@@ -686,4 +695,4 @@ declare function isRetryable(errorOrResponse: Error | Response, config?: Partial
 
 declare const plugin: OpenClawPluginDefinition;
 
-export { BALANCE_THRESHOLDS, BLOCKRUN_MODELS, type BalanceInfo, BalanceMonitor, type CachedPaymentParams, type CachedResponse, DEFAULT_RETRY_CONFIG, DEFAULT_ROUTING_CONFIG, EmptyWalletError, InsufficientFundsError, type InsufficientFundsInfo, type LowBalanceInfo, OPENCLAW_MODELS, PaymentCache, type PaymentFetchResult, type PreAuthParams, type ProxyHandle, type ProxyOptions, RequestDeduplicator, type RetryConfig, type RoutingConfig, type RoutingDecision, RpcError, type SufficiencyResult, type Tier, type UsageEntry, blockrunProvider, buildProviderModels, createPaymentFetch, plugin as default, fetchWithRetry, isBalanceError, isEmptyWalletError, isInsufficientFundsError, isRetryable, isRpcError, logUsage, route, startProxy };
+export { BALANCE_THRESHOLDS, BLOCKRUN_MODELS, type BalanceInfo, BalanceMonitor, type CachedPaymentParams, type CachedResponse, DEFAULT_RETRY_CONFIG, DEFAULT_ROUTING_CONFIG, EmptyWalletError, InsufficientFundsError, type InsufficientFundsInfo, type LowBalanceInfo, OPENCLAW_MODELS, PaymentCache, type PaymentFetchResult, type PreAuthParams, type ProxyHandle, type ProxyOptions, RequestDeduplicator, type RetryConfig, type RoutingConfig, type RoutingDecision, RpcError, type SufficiencyResult, type Tier, type UsageEntry, blockrunProvider, buildProviderModels, createPaymentFetch, plugin as default, fetchWithRetry, getProxyPort, isBalanceError, isEmptyWalletError, isInsufficientFundsError, isRetryable, isRpcError, logUsage, route, startProxy };

package/dist/index.js

@@ -712,6 +712,10 @@ function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPric
     savings
   };
 }
+function getFallbackChain(tier, tierConfigs) {
+  const config = tierConfigs[tier];
+  return [config.primary, ...config.fallback];
+}
 
 // src/router/config.ts
 var DEFAULT_ROUTING_CONFIG = {
@@ -763,10 +767,13 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u0444\u0443\u043D\u043A\u0446\u0438\u044F",
       "\u043A\u043B\u0430\u0441\u0441",
       "\u0438\u043C\u043F\u043E\u0440\u0442",
+      "\u043E\u043F\u0440\u0435\u0434\u0435\u043B",
       "\u0437\u0430\u043F\u0440\u043E\u0441",
       "\u0430\u0441\u0438\u043D\u0445\u0440\u043E\u043D\u043D\u044B\u0439",
+      "\u043E\u0436\u0438\u0434\u0430\u0442\u044C",
       "\u043A\u043E\u043D\u0441\u0442\u0430\u043D\u0442\u0430",
-      "\u043F\u0435\u0440\u0435\u043C\u0435\u043D\u043D\u0430\u044F"
+      "\u043F\u0435\u0440\u0435\u043C\u0435\u043D\u043D\u0430\u044F",
+      "\u0432\u0435\u0440\u043D\u0443\u0442\u044C"
     ],
     reasoningKeywords: [
       // English
@@ -796,10 +803,15 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u8AD6\u7406\u7684",
       // Russian
       "\u0434\u043E\u043A\u0430\u0437\u0430\u0442\u044C",
+      "\u0434\u043E\u043A\u0430\u0436\u0438",
+      "\u0434\u043E\u043A\u0430\u0437\u0430\u0442\u0435\u043B\u044C\u0441\u0442\u0432",
       "\u0442\u0435\u043E\u0440\u0435\u043C\u0430",
       "\u0432\u044B\u0432\u0435\u0441\u0442\u0438",
       "\u0448\u0430\u0433 \u0437\u0430 \u0448\u0430\u0433\u043E\u043C",
+      "\u043F\u043E\u0448\u0430\u0433\u043E\u0432\u043E",
+      "\u043F\u043E\u044D\u0442\u0430\u043F\u043D\u043E",
       "\u0446\u0435\u043F\u043E\u0447\u043A\u0430 \u0440\u0430\u0441\u0441\u0443\u0436\u0434\u0435\u043D\u0438\u0439",
+      "\u0440\u0430\u0441\u0441\u0443\u0436\u0434\u0435\u043D\u0438",
       "\u0444\u043E\u0440\u043C\u0430\u043B\u044C\u043D\u043E",
       "\u043C\u0430\u0442\u0435\u043C\u0430\u0442\u0438\u0447\u0435\u0441\u043A\u0438",
       "\u043B\u043E\u0433\u0438\u0447\u0435\u0441\u043A\u0438"
@@ -837,11 +849,14 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u0447\u0442\u043E \u0442\u0430\u043A\u043E\u0435",
       "\u043E\u043F\u0440\u0435\u0434\u0435\u043B\u0435\u043D\u0438\u0435",
       "\u043F\u0435\u0440\u0435\u0432\u0435\u0441\u0442\u0438",
+      "\u043F\u0435\u0440\u0435\u0432\u0435\u0434\u0438",
       "\u043F\u0440\u0438\u0432\u0435\u0442",
       "\u0434\u0430 \u0438\u043B\u0438 \u043D\u0435\u0442",
       "\u0441\u0442\u043E\u043B\u0438\u0446\u0430",
+      "\u0441\u043A\u043E\u043B\u044C\u043A\u043E \u043B\u0435\u0442",
       "\u043A\u0442\u043E \u0442\u0430\u043A\u043E\u0439",
-      "\u043A\u043E\u0433\u0434\u0430"
+      "\u043A\u043E\u0433\u0434\u0430",
+      "\u043E\u0431\u044A\u044F\u0441\u043D\u0438"
     ],
     technicalKeywords: [
       // English
@@ -871,6 +886,8 @@ var DEFAULT_ROUTING_CONFIG = {
       // Russian
       "\u0430\u043B\u0433\u043E\u0440\u0438\u0442\u043C",
       "\u043E\u043F\u0442\u0438\u043C\u0438\u0437\u0438\u0440\u043E\u0432\u0430\u0442\u044C",
+      "\u043E\u043F\u0442\u0438\u043C\u0438\u0437\u0430\u0446\u0438",
+      "\u043E\u043F\u0442\u0438\u043C\u0438\u0437\u0438\u0440\u0443\u0439",
       "\u0430\u0440\u0445\u0438\u0442\u0435\u043A\u0442\u0443\u0440\u0430",
       "\u0440\u0430\u0441\u043F\u0440\u0435\u0434\u0435\u043B\u0451\u043D\u043D\u044B\u0439",
       "\u043C\u0438\u043A\u0440\u043E\u0441\u0435\u0440\u0432\u0438\u0441",
@@ -903,11 +920,14 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u60F3\u50CF",
       // Russian
       "\u0438\u0441\u0442\u043E\u0440\u0438\u044F",
+      "\u0440\u0430\u0441\u0441\u043A\u0430\u0437",
       "\u0441\u0442\u0438\u0445\u043E\u0442\u0432\u043E\u0440\u0435\u043D\u0438\u0435",
       "\u0441\u043E\u0447\u0438\u043D\u0438\u0442\u044C",
+      "\u0441\u043E\u0447\u0438\u043D\u0438",
       "\u043C\u043E\u0437\u0433\u043E\u0432\u043E\u0439 \u0448\u0442\u0443\u0440\u043C",
       "\u0442\u0432\u043E\u0440\u0447\u0435\u0441\u043A\u0438\u0439",
       "\u043F\u0440\u0435\u0434\u0441\u0442\u0430\u0432\u0438\u0442\u044C",
+      "\u043F\u0440\u0438\u0434\u0443\u043C\u0430\u0439",
       "\u043D\u0430\u043F\u0438\u0448\u0438"
     ],
     // New dimension keyword lists (multilingual)
@@ -944,13 +964,21 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u8A2D\u5B9A",
       // Russian
       "\u043F\u043E\u0441\u0442\u0440\u043E\u0438\u0442\u044C",
+      "\u043F\u043E\u0441\u0442\u0440\u043E\u0439",
       "\u0441\u043E\u0437\u0434\u0430\u0442\u044C",
+      "\u0441\u043E\u0437\u0434\u0430\u0439",
       "\u0440\u0435\u0430\u043B\u0438\u0437\u043E\u0432\u0430\u0442\u044C",
+      "\u0440\u0435\u0430\u043B\u0438\u0437\u0443\u0439",
       "\u0441\u043F\u0440\u043E\u0435\u043A\u0442\u0438\u0440\u043E\u0432\u0430\u0442\u044C",
       "\u0440\u0430\u0437\u0440\u0430\u0431\u043E\u0442\u0430\u0442\u044C",
+      "\u0440\u0430\u0437\u0440\u0430\u0431\u043E\u0442\u0430\u0439",
+      "\u0441\u043A\u043E\u043D\u0441\u0442\u0440\u0443\u0438\u0440\u043E\u0432\u0430\u0442\u044C",
       "\u0441\u0433\u0435\u043D\u0435\u0440\u0438\u0440\u043E\u0432\u0430\u0442\u044C",
+      "\u0441\u0433\u0435\u043D\u0435\u0440\u0438\u0440\u0443\u0439",
       "\u0440\u0430\u0437\u0432\u0435\u0440\u043D\u0443\u0442\u044C",
-      "\u043D\u0430\u0441\u0442\u0440\u043E\u0438\u0442\u044C"
+      "\u0440\u0430\u0437\u0432\u0435\u0440\u043D\u0438",
+      "\u043D\u0430\u0441\u0442\u0440\u043E\u0438\u0442\u044C",
+      "\u043D\u0430\u0441\u0442\u0440\u043E\u0439"
     ],
     constraintIndicators: [
       // English
@@ -981,7 +1009,9 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u4E88\u7B97",
       // Russian
       "\u043D\u0435 \u0431\u043E\u043B\u0435\u0435",
+      "\u043D\u0435 \u043C\u0435\u043D\u0435\u0435",
       "\u043A\u0430\u043A \u043C\u0438\u043D\u0438\u043C\u0443\u043C",
+      "\u0432 \u043F\u0440\u0435\u0434\u0435\u043B\u0430\u0445",
       "\u043C\u0430\u043A\u0441\u0438\u043C\u0443\u043C",
       "\u043C\u0438\u043D\u0438\u043C\u0443\u043C",
       "\u043E\u0433\u0440\u0430\u043D\u0438\u0447\u0435\u043D\u0438\u0435",
@@ -1044,6 +1074,7 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u0441\u043B\u0435\u0434\u0443\u044E\u0449\u0438\u0439",
       "\u0434\u043E\u043A\u0443\u043C\u0435\u043D\u0442\u0430\u0446\u0438\u044F",
       "\u043A\u043E\u0434",
+      "\u0440\u0430\u043D\u0435\u0435",
       "\u0432\u043B\u043E\u0436\u0435\u043D\u0438\u0435"
     ],
     negationKeywords: [
@@ -1071,11 +1102,14 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u9664\u304F",
       // Russian
       "\u043D\u0435 \u0434\u0435\u043B\u0430\u0439",
+      "\u043D\u0435 \u043D\u0430\u0434\u043E",
+      "\u043D\u0435\u043B\u044C\u0437\u044F",
       "\u0438\u0437\u0431\u0435\u0433\u0430\u0442\u044C",
       "\u043D\u0438\u043A\u043E\u0433\u0434\u0430",
       "\u0431\u0435\u0437",
       "\u043A\u0440\u043E\u043C\u0435",
-      "\u0438\u0441\u043A\u043B\u044E\u0447\u0438\u0442\u044C"
+      "\u0438\u0441\u043A\u043B\u044E\u0447\u0438\u0442\u044C",
+      "\u0431\u043E\u043B\u044C\u0448\u0435 \u043D\u0435"
     ],
     domainSpecificKeywords: [
       // English
@@ -1112,7 +1146,8 @@ var DEFAULT_ROUTING_CONFIG = {
       "\u043F\u0440\u043E\u0442\u0435\u043E\u043C\u0438\u043A\u0430",
       "\u0442\u043E\u043F\u043E\u043B\u043E\u0433\u0438\u0447\u0435\u0441\u043A\u0438\u0439",
       "\u0433\u043E\u043C\u043E\u043C\u043E\u0440\u0444\u043D\u044B\u0439",
-      "\u0441 \u043D\u0443\u043B\u0435\u0432\u044B\u043C \u0440\u0430\u0437\u0433\u043B\u0430\u0448\u0435\u043D\u0438\u0435\u043C"
+      "\u0441 \u043D\u0443\u043B\u0435\u0432\u044B\u043C \u0440\u0430\u0437\u0433\u043B\u0430\u0448\u0435\u043D\u0438\u0435\u043C",
+      "\u043D\u0430 \u043E\u0441\u043D\u043E\u0432\u0435 \u0440\u0435\u0448\u0451\u0442\u043E\u043A"
     ],
     // Dimension weights (sum to 1.0)
     dimensionWeights: {
@@ -1547,6 +1582,119 @@ var AUTO_MODEL_SHORT = "auto";
 var HEARTBEAT_INTERVAL_MS = 2e3;
 var DEFAULT_REQUEST_TIMEOUT_MS = 18e4;
 var DEFAULT_PORT = 8402;
+var MAX_FALLBACK_ATTEMPTS = 3;
+var HEALTH_CHECK_TIMEOUT_MS = 2e3;
+function getProxyPort() {
+  const envPort = process.env.BLOCKRUN_PROXY_PORT;
+  if (envPort) {
+    const parsed = parseInt(envPort, 10);
+    if (!isNaN(parsed) && parsed > 0 && parsed < 65536) {
+      return parsed;
+    }
+  }
+  return DEFAULT_PORT;
+}
+async function checkExistingProxy(port) {
+  const controller = new AbortController();
+  const timeoutId = setTimeout(() => controller.abort(), HEALTH_CHECK_TIMEOUT_MS);
+  try {
+    const response = await fetch(`http://127.0.0.1:${port}/health`, {
+      signal: controller.signal
+    });
+    clearTimeout(timeoutId);
+    if (response.ok) {
+      const data = await response.json();
+      if (data.status === "ok" && data.wallet) {
+        return data.wallet;
+      }
+    }
+    return void 0;
+  } catch {
+    clearTimeout(timeoutId);
+    return void 0;
+  }
+}
+var PROVIDER_ERROR_PATTERNS = [
+  /billing/i,
+  /insufficient.*balance/i,
+  /credits/i,
+  /quota.*exceeded/i,
+  /rate.*limit/i,
+  /model.*unavailable/i,
+  /model.*not.*available/i,
+  /service.*unavailable/i,
+  /capacity/i,
+  /overloaded/i,
+  /temporarily.*unavailable/i,
+  /api.*key.*invalid/i,
+  /authentication.*failed/i
+];
+var FALLBACK_STATUS_CODES = [
+  400,
+  // Bad request - sometimes used for billing errors
+  401,
+  // Unauthorized - provider API key issues
+  402,
+  // Payment required - but from upstream, not x402
+  403,
+  // Forbidden - provider restrictions
+  429,
+  // Rate limited
+  500,
+  // Internal server error
+  502,
+  // Bad gateway
+  503,
+  // Service unavailable
+  504
+  // Gateway timeout
+];
+function isProviderError(status, body) {
+  if (!FALLBACK_STATUS_CODES.includes(status)) {
+    return false;
+  }
+  if (status >= 500) {
+    return true;
+  }
+  return PROVIDER_ERROR_PATTERNS.some((pattern) => pattern.test(body));
+}
+function normalizeMessagesForGoogle(messages) {
+  if (!messages || messages.length === 0) return messages;
+  let firstNonSystemIdx = -1;
+  for (let i = 0; i < messages.length; i++) {
+    if (messages[i].role !== "system") {
+      firstNonSystemIdx = i;
+      break;
+    }
+  }
+  if (firstNonSystemIdx === -1) return messages;
+  const firstRole = messages[firstNonSystemIdx].role;
+  if (firstRole === "user") return messages;
+  if (firstRole === "assistant" || firstRole === "model") {
+    const normalized = [...messages];
+    normalized.splice(firstNonSystemIdx, 0, {
+      role: "user",
+      content: "(continuing conversation)"
+    });
+    return normalized;
+  }
+  return messages;
+}
+function isGoogleModel(modelId) {
+  return modelId.startsWith("google/") || modelId.startsWith("gemini");
+}
+var KIMI_BLOCK_RE = /<[｜|][^<>]*begin[^<>]*[｜|]>[\s\S]*?<[｜|][^<>]*end[^<>]*[｜|]>/gi;
+var KIMI_TOKEN_RE = /<[｜|][^<>]*[｜|]>/g;
+var THINKING_TAG_RE = /<\s*\/?\s*(?:think(?:ing)?|thought|antthinking)\b[^>]*>/gi;
+var THINKING_BLOCK_RE = /<\s*(?:think(?:ing)?|thought|antthinking)\b[^>]*>[\s\S]*?<\s*\/\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
+function stripThinkingTokens(content) {
+  if (!content) return content;
+  let cleaned = content.replace(KIMI_BLOCK_RE, "");
+  cleaned = cleaned.replace(KIMI_TOKEN_RE, "");
+  cleaned = cleaned.replace(THINKING_BLOCK_RE, "");
+  cleaned = cleaned.replace(THINKING_TAG_RE, "");
+  return cleaned;
+}
 function buildModelPricing() {
   const map = /* @__PURE__ */ new Map();
   for (const m of BLOCKRUN_MODELS) {
@@ -1577,6 +1725,27 @@ function estimateAmount(modelId, bodyLength, maxTokens) {
 }
 async function startProxy(options) {
   const apiBase = options.apiBase ?? BLOCKRUN_API;
+  const listenPort = options.port ?? getProxyPort();
+  const existingWallet = await checkExistingProxy(listenPort);
+  if (existingWallet) {
+    const account2 = privateKeyToAccount2(options.walletKey);
+    const balanceMonitor2 = new BalanceMonitor(account2.address);
+    const baseUrl = `http://127.0.0.1:${listenPort}`;
+    if (existingWallet !== account2.address) {
+      console.warn(
+        `[ClawRouter] Existing proxy on port ${listenPort} uses wallet ${existingWallet}, but current config uses ${account2.address}. Reusing existing proxy.`
+      );
+    }
+    options.onReady?.(listenPort);
+    return {
+      port: listenPort,
+      baseUrl,
+      walletAddress: existingWallet,
+      balanceMonitor: balanceMonitor2,
+      close: async () => {
+      }
+    };
+  }
   const account = privateKeyToAccount2(options.walletKey);
   const { fetch: payFetch } = createPaymentFetch(options.walletKey);
   const balanceMonitor = new BalanceMonitor(account.address);
@@ -1646,7 +1815,6 @@ async function startProxy(options) {
       }
     }
   });
-  const listenPort = options.port ?? DEFAULT_PORT;
   return new Promise((resolve, reject) => {
     server.on("error", reject);
     server.listen(listenPort, "127.0.0.1", () => {
@@ -1666,6 +1834,52 @@ async function startProxy(options) {
     });
   });
 }
+async function tryModelRequest(upstreamUrl, method, headers, body, modelId, maxTokens, payFetch, balanceMonitor, signal) {
+  let requestBody = body;
+  try {
+    const parsed = JSON.parse(body.toString());
+    parsed.model = modelId;
+    if (isGoogleModel(modelId) && Array.isArray(parsed.messages)) {
+      parsed.messages = normalizeMessagesForGoogle(parsed.messages);
+    }
+    requestBody = Buffer.from(JSON.stringify(parsed));
+  } catch {
+  }
+  const estimated = estimateAmount(modelId, requestBody.length, maxTokens);
+  const preAuth = estimated ? { estimatedAmount: estimated } : void 0;
+  try {
+    const response = await payFetch(
+      upstreamUrl,
+      {
+        method,
+        headers,
+        body: requestBody.length > 0 ? new Uint8Array(requestBody) : void 0,
+        signal
+      },
+      preAuth
+    );
+    if (response.status !== 200) {
+      const errorBody = await response.text();
+      const isProviderErr = isProviderError(response.status, errorBody);
+      return {
+        success: false,
+        errorBody,
+        errorStatus: response.status,
+        isProviderError: isProviderErr
+      };
+    }
+    return { success: true, response };
+  } catch (err) {
+    const errorMsg = err instanceof Error ? err.message : String(err);
+    return {
+      success: false,
+      errorBody: errorMsg,
+      errorStatus: 500,
+      isProviderError: true
+      // Network errors are retryable
+    };
+  }
+}
 async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, deduplicator, balanceMonitor) {
   const startTime = Date.now();
   const upstreamUrl = `${apiBase}${req.url}`;
@@ -1740,7 +1954,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   }
   deduplicator.markInflight(dedupKey);
   let estimatedCostMicros;
-  if (modelId) {
+  if (modelId && !options.skipBalanceCheck) {
     const estimated = estimateAmount(modelId, body.length, maxTokens);
     if (estimated) {
       estimatedCostMicros = BigInt(estimated);
@@ -1805,10 +2019,6 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
     headers["content-type"] = "application/json";
   }
   headers["user-agent"] = USER_AGENT;
-  let preAuth;
-  if (estimatedCostMicros !== void 0) {
-    preAuth = { estimatedAmount: estimatedCostMicros.toString() };
-  }
   let completed = false;
   res.on("close", () => {
     if (heartbeatInterval) {
@@ -1823,26 +2033,72 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   const controller = new AbortController();
   const timeoutId = setTimeout(() => controller.abort(), timeoutMs);
   try {
-    const upstream = await payFetch(
-      upstreamUrl,
-      {
-        method: req.method ?? "POST",
+    let modelsToTry;
+    if (routingDecision) {
+      modelsToTry = getFallbackChain(routingDecision.tier, routerOpts.config.tiers);
+      modelsToTry = modelsToTry.slice(0, MAX_FALLBACK_ATTEMPTS);
+    } else {
+      modelsToTry = modelId ? [modelId] : [];
+    }
+    let upstream;
+    let lastError;
+    let actualModelUsed = modelId;
+    for (let i = 0; i < modelsToTry.length; i++) {
+      const tryModel = modelsToTry[i];
+      const isLastAttempt = i === modelsToTry.length - 1;
+      console.log(`[ClawRouter] Trying model ${i + 1}/${modelsToTry.length}: ${tryModel}`);
+      const result = await tryModelRequest(
+        upstreamUrl,
+        req.method ?? "POST",
         headers,
-        body: body.length > 0 ? body : void 0,
-        signal: controller.signal
-      },
-      preAuth
-    );
+        body,
+        tryModel,
+        maxTokens,
+        payFetch,
+        balanceMonitor,
+        controller.signal
+      );
+      if (result.success && result.response) {
+        upstream = result.response;
+        actualModelUsed = tryModel;
+        console.log(`[ClawRouter] Success with model: ${tryModel}`);
+        break;
+      }
+      lastError = {
+        body: result.errorBody || "Unknown error",
+        status: result.errorStatus || 500
+      };
+      if (result.isProviderError && !isLastAttempt) {
+        console.log(
+          `[ClawRouter] Provider error from ${tryModel}, trying fallback: ${result.errorBody?.slice(0, 100)}`
+        );
+        continue;
+      }
+      if (!result.isProviderError) {
+        console.log(
+          `[ClawRouter] Non-provider error from ${tryModel}, not retrying: ${result.errorBody?.slice(0, 100)}`
+        );
+      }
+      break;
+    }
     clearTimeout(timeoutId);
     if (heartbeatInterval) {
       clearInterval(heartbeatInterval);
       heartbeatInterval = void 0;
     }
-    const responseChunks = [];
-    if (headersSentEarly) {
-      if (upstream.status !== 200) {
-        const errBody = await upstream.text();
-        const errEvent = `data: ${JSON.stringify({ error: { message: errBody, type: "upstream_error", status: upstream.status } })}
+    if (routingDecision && actualModelUsed !== routingDecision.model) {
+      routingDecision = {
+        ...routingDecision,
+        model: actualModelUsed,
+        reasoning: `${routingDecision.reasoning} | fallback to ${actualModelUsed}`
+      };
+      options.onRouted?.(routingDecision);
+    }
+    if (!upstream) {
+      const errBody = lastError?.body || "All models in fallback chain failed";
+      const errStatus = lastError?.status || 502;
+      if (headersSentEarly) {
+        const errEvent = `data: ${JSON.stringify({ error: { message: errBody, type: "provider_error", status: errStatus } })}
 
 `;
         res.write(errEvent);
@@ -1851,13 +2107,30 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
         const errBuf = Buffer.from(errEvent + "data: [DONE]\n\n");
         deduplicator.complete(dedupKey, {
           status: 200,
-          // we already sent 200
           headers: { "content-type": "text/event-stream" },
           body: errBuf,
           completedAt: Date.now()
         });
-        return;
+      } else {
+        res.writeHead(errStatus, { "Content-Type": "application/json" });
+        res.end(
+          JSON.stringify({
+            error: { message: errBody, type: "provider_error" }
+          })
+        );
+        deduplicator.complete(dedupKey, {
+          status: errStatus,
+          headers: { "content-type": "application/json" },
+          body: Buffer.from(
+            JSON.stringify({ error: { message: errBody, type: "provider_error" } })
+          ),
+          completedAt: Date.now()
+        });
       }
+      return;
+    }
+    const responseChunks = [];
+    if (headersSentEarly) {
       if (upstream.body) {
         const reader = upstream.body.getReader();
         const chunks = [];
@@ -1882,7 +2155,8 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
           };
           if (rsp.choices && Array.isArray(rsp.choices)) {
             for (const choice of rsp.choices) {
-              const content = choice.message?.content ?? choice.delta?.content ?? "";
+              const rawContent = choice.message?.content ?? choice.delta?.content ?? "";
+              const content = stripThinkingTokens(rawContent);
               const role = choice.message?.role ?? choice.delta?.role ?? "assistant";
               const index = choice.index ?? 0;
               const roleChunk = {
@@ -2104,13 +2378,18 @@ function injectModelsConfig(logger) {
     let needsWrite = false;
     if (!config.models) config.models = {};
     if (!config.models.providers) config.models.providers = {};
+    const proxyPort = getProxyPort();
+    const expectedBaseUrl = `http://127.0.0.1:${proxyPort}/v1`;
     if (!config.models.providers.blockrun) {
       config.models.providers.blockrun = {
-        baseUrl: "http://127.0.0.1:8402/v1",
+        baseUrl: expectedBaseUrl,
         api: "openai-completions",
         models: OPENCLAW_MODELS
       };
       needsWrite = true;
+    } else if (config.models.providers.blockrun.baseUrl !== expectedBaseUrl) {
+      config.models.providers.blockrun.baseUrl = expectedBaseUrl;
+      needsWrite = true;
     }
     if (!config.agents) config.agents = {};
     if (!config.agents.defaults) config.agents.defaults = {};
@@ -2259,6 +2538,7 @@ var plugin = {
     api.registerProvider(blockrunProvider);
     injectModelsConfig(api.logger);
     injectAuthProfile(api.logger);
+    const runtimePort = getProxyPort();
     if (!api.config.models) {
       api.config.models = { providers: {} };
     }
@@ -2266,7 +2546,7 @@ var plugin = {
       api.config.models.providers = {};
     }
     api.config.models.providers.blockrun = {
-      baseUrl: "http://127.0.0.1:8402/v1",
+      baseUrl: `http://127.0.0.1:${runtimePort}/v1`,
       api: "openai-completions",
       models: OPENCLAW_MODELS
     };
@@ -2320,6 +2600,7 @@ export {
   createPaymentFetch,
   index_default as default,
   fetchWithRetry,
+  getProxyPort,
   isBalanceError,
   isEmptyWalletError,
   isInsufficientFundsError,