@blockrun/clawrouter 0.12.230 → 0.12.232

package/README.md

@@ -434,6 +434,7 @@ No Surf account, no API key — settles directly to Surf's Base treasury in USDC
 | google/gemini-2.5-pro       |     $1.25 |     $10.00 |    $0.0056 | 1M      | reasoning, vision, tools                  |
 | zai/glm-5.2                 |     $1.40 |      $4.40 |    $0.0029 | 1M      | reasoning, coding, tools (flagship)       |
 | zai/glm-5.1                 |     $1.40 |      $4.40 |    $0.0029 | 200K    | reasoning, tools (promo ended 2026-06-05) |
+| qwen/qwen3.7-max            |    $1.475 |     $4.425 |    $0.0030 | 1M      | reasoning, agentic, tools (Qwen flagship) |
 | xai/grok-4.3                |     $1.50 |      $4.00 |    $0.0028 | 1M      | reasoning, vision, agentic, tools         |
 | xai/grok-4.5                |     $2.50 |      $9.00 |    $0.0058 | 500K    | reasoning, vision, agentic, tools         |
 | xai/grok-build-0.1          |     $1.50 |      $3.00 |    $0.0023 | 256K    | agentic coding, tools                     |

package/dist/cli.js

@@ -31393,6 +31393,22 @@ var init_esm4 = __esm({
   }
 });
 
+// src/max-tokens.ts
+function resolveMaxTokens(parsed, fallback2 = DEFAULT_MAX_TOKENS) {
+  const usable = (v) => typeof v === "number" && Number.isFinite(v) && v > 0 ? v : void 0;
+  const candidates = [usable(parsed.max_tokens), usable(parsed.max_completion_tokens)].filter(
+    (n) => n !== void 0
+  );
+  return candidates.length > 0 ? Math.max(...candidates) : fallback2;
+}
+var DEFAULT_MAX_TOKENS;
+var init_max_tokens = __esm({
+  "src/max-tokens.ts"() {
+    "use strict";
+    DEFAULT_MAX_TOKENS = 4096;
+  }
+});
+
 // src/payment-preauth.ts
 function createPayFetchWithPreAuth(baseFetch, client, ttlMs = DEFAULT_TTL_MS, options) {
   const httpClient = new x402HTTPClient(client);
@@ -31410,7 +31426,7 @@ function createPayFetchWithPreAuth(baseFetch, client, ttlMs = DEFAULT_TTL_MS, op
           bodyLength = bodyStr.length;
           const parsed = JSON.parse(bodyStr);
           requestModel = parsed.model ?? "";
-          maxTokens = Number(parsed.max_tokens) || 0;
+          maxTokens = resolveMaxTokens(parsed, 0);
         }
       } catch {
       }
@@ -31481,6 +31497,7 @@ var init_payment_preauth = __esm({
   "src/payment-preauth.ts"() {
     "use strict";
     init_esm4();
+    init_max_tokens();
     DEFAULT_TTL_MS = 36e5;
   }
 });
@@ -34296,6 +34313,7 @@ var init_top_models = __esm({
       "minimax/minimax-m3",
       "minimax/minimax-m2.7",
       "moonshot/kimi-k3",
+      "qwen/qwen3.7-max",
       "deepseek/deepseek-v4-pro",
       "deepseek/deepseek-chat",
       "deepseek/deepseek-reasoner",
@@ -34522,6 +34540,17 @@ var init_models = __esm({
       "kimi-k2.6": "moonshot/kimi-k2.6",
       "kimi-k2.5": "moonshot/kimi-k2.5",
       "nvidia/kimi-k2.5": "moonshot/kimi-k2.5",
+      // Qwen — Qwen3.7 Max is Alibaba's current flagship for reasoning, coding,
+      // and agentic tool use. EXPLICIT PINS ONLY: bare `qwen` deliberately stays
+      // unbound. Every other qwen* shorthand below points at a FREE model
+      // (qwen-coder, qwen-thinking, qwen3-next, qwen3.5-122b), so binding the
+      // shortest name to a $1.475/$4.425 flagship would silently charge callers
+      // who typed it expecting the free tier — same rule that keeps generic
+      // `kimi` on K2.7. (`grok` WAS promoted to 4.5, but only after the cost
+      // tradeoff was argued explicitly; there's no such case for qwen yet.)
+      "qwen3.7-max": "qwen/qwen3.7-max",
+      "qwen-3.7-max": "qwen/qwen3.7-max",
+      "qwen3-7-max": "qwen/qwen3.7-max",
       // Google
       // gemini-3-pro-preview delisted by Google 2026-06-06 — mirror the gateway
       // redirect to its successor so pinned callers land on 3.1-pro, not an error.
@@ -35331,6 +35360,27 @@ var init_models = __esm({
         agentic: true,
         toolCalling: true
       },
+      // Qwen3.7 Max — BlockRun's paid Qwen flagship, served through its
+      // OpenRouter credit pool. The gateway applies its standard 5% margin at
+      // settlement, so keep the catalog rates at the upstream $1.475/$4.425 COGS.
+      // NOTE: `reasoning: true` also enrolls this id in REASONING_MODEL_IDS
+      // (proxy.ts), which raises its per-model timeout from 60s to 180s.
+      // `toolCalling: true` is LIVE-VERIFIED (2026-07-20): a real request through
+      // the gateway returned a structured tool_calls array (name + valid JSON
+      // arguments, finish_reason "tool_calls"), not the textual leak that Kimi K3
+      // (#213), Gemini (#189) and GPT (#193) produce. Don't downgrade on a hunch.
+      {
+        id: "qwen/qwen3.7-max",
+        name: "Qwen3.7 Max",
+        version: "3.7-max",
+        inputPrice: 1.475,
+        outputPrice: 4.425,
+        contextWindow: 1e6,
+        maxOutput: 65536,
+        reasoning: true,
+        agentic: true,
+        toolCalling: true
+      },
       // Kimi K2.6 — superseded by K2.7 (2026-06-13), hidden on BlockRun but still routable.
       {
         id: "moonshot/kimi-k2.6",
@@ -86058,7 +86108,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   let hasVision = false;
   let isStreaming = false;
   let modelId = "";
-  let maxTokens = 4096;
+  let maxTokens = DEFAULT_MAX_TOKENS;
   let routingProfile = null;
   let stickyExplicitModel;
   let balanceFallbackNotice;
@@ -86077,7 +86127,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       const parsed = JSON.parse(body.toString());
       isStreaming = parsed.stream === true;
       modelId = parsed.model || "";
-      maxTokens = parsed.max_tokens || 4096;
+      maxTokens = resolveMaxTokens(parsed);
       let bodyModified = false;
       const parsedMessages = Array.isArray(parsed.messages) ? parsed.messages : [];
       const lastUserMsg = [...parsedMessages].reverse().find((m) => m.role === "user");
@@ -88074,6 +88124,7 @@ var init_proxy = __esm({
     init_router();
     init_rules();
     init_models();
+    init_max_tokens();
     init_logger();
     init_stats();
     init_dedup();
@@ -88091,6 +88142,7 @@ var init_proxy = __esm({
     init_textual_tool_calls();
     init_response_store();
     init_share_formatters();
+    init_max_tokens();
     paymentStore = new AsyncLocalStorage();
     BLOCKRUN_API = "https://blockrun.ai/api";
     BLOCKRUN_SOLANA_API = "https://sol.blockrun.ai/api";

package/dist/index.js

@@ -86,6 +86,7 @@ var init_top_models = __esm({
       "minimax/minimax-m3",
       "minimax/minimax-m2.7",
       "moonshot/kimi-k3",
+      "qwen/qwen3.7-max",
       "deepseek/deepseek-v4-pro",
       "deepseek/deepseek-chat",
       "deepseek/deepseek-reasoner",
@@ -312,6 +313,17 @@ var init_models = __esm({
       "kimi-k2.6": "moonshot/kimi-k2.6",
       "kimi-k2.5": "moonshot/kimi-k2.5",
       "nvidia/kimi-k2.5": "moonshot/kimi-k2.5",
+      // Qwen — Qwen3.7 Max is Alibaba's current flagship for reasoning, coding,
+      // and agentic tool use. EXPLICIT PINS ONLY: bare `qwen` deliberately stays
+      // unbound. Every other qwen* shorthand below points at a FREE model
+      // (qwen-coder, qwen-thinking, qwen3-next, qwen3.5-122b), so binding the
+      // shortest name to a $1.475/$4.425 flagship would silently charge callers
+      // who typed it expecting the free tier — same rule that keeps generic
+      // `kimi` on K2.7. (`grok` WAS promoted to 4.5, but only after the cost
+      // tradeoff was argued explicitly; there's no such case for qwen yet.)
+      "qwen3.7-max": "qwen/qwen3.7-max",
+      "qwen-3.7-max": "qwen/qwen3.7-max",
+      "qwen3-7-max": "qwen/qwen3.7-max",
       // Google
       // gemini-3-pro-preview delisted by Google 2026-06-06 — mirror the gateway
       // redirect to its successor so pinned callers land on 3.1-pro, not an error.
@@ -1121,6 +1133,27 @@ var init_models = __esm({
         agentic: true,
         toolCalling: true
       },
+      // Qwen3.7 Max — BlockRun's paid Qwen flagship, served through its
+      // OpenRouter credit pool. The gateway applies its standard 5% margin at
+      // settlement, so keep the catalog rates at the upstream $1.475/$4.425 COGS.
+      // NOTE: `reasoning: true` also enrolls this id in REASONING_MODEL_IDS
+      // (proxy.ts), which raises its per-model timeout from 60s to 180s.
+      // `toolCalling: true` is LIVE-VERIFIED (2026-07-20): a real request through
+      // the gateway returned a structured tool_calls array (name + valid JSON
+      // arguments, finish_reason "tool_calls"), not the textual leak that Kimi K3
+      // (#213), Gemini (#189) and GPT (#193) produce. Don't downgrade on a hunch.
+      {
+        id: "qwen/qwen3.7-max",
+        name: "Qwen3.7 Max",
+        version: "3.7-max",
+        inputPrice: 1.475,
+        outputPrice: 4.425,
+        contextWindow: 1e6,
+        maxOutput: 65536,
+        reasoning: true,
+        agentic: true,
+        toolCalling: true
+      },
       // Kimi K2.6 — superseded by K2.7 (2026-06-13), hidden on BlockRun but still routable.
       {
         id: "moonshot/kimi-k2.6",
@@ -32884,6 +32917,22 @@ var init_esm4 = __esm({
   }
 });
 
+// src/max-tokens.ts
+function resolveMaxTokens(parsed, fallback2 = DEFAULT_MAX_TOKENS) {
+  const usable = (v) => typeof v === "number" && Number.isFinite(v) && v > 0 ? v : void 0;
+  const candidates = [usable(parsed.max_tokens), usable(parsed.max_completion_tokens)].filter(
+    (n) => n !== void 0
+  );
+  return candidates.length > 0 ? Math.max(...candidates) : fallback2;
+}
+var DEFAULT_MAX_TOKENS;
+var init_max_tokens = __esm({
+  "src/max-tokens.ts"() {
+    "use strict";
+    DEFAULT_MAX_TOKENS = 4096;
+  }
+});
+
 // src/payment-preauth.ts
 function createPayFetchWithPreAuth(baseFetch, client, ttlMs = DEFAULT_TTL_MS, options) {
   const httpClient = new x402HTTPClient(client);
@@ -32901,7 +32950,7 @@ function createPayFetchWithPreAuth(baseFetch, client, ttlMs = DEFAULT_TTL_MS, op
           bodyLength = bodyStr.length;
           const parsed = JSON.parse(bodyStr);
           requestModel = parsed.model ?? "";
-          maxTokens = Number(parsed.max_tokens) || 0;
+          maxTokens = resolveMaxTokens(parsed, 0);
         }
       } catch {
       }
@@ -32972,6 +33021,7 @@ var init_payment_preauth = __esm({
   "src/payment-preauth.ts"() {
     "use strict";
     init_esm4();
+    init_max_tokens();
     DEFAULT_TTL_MS = 36e5;
   }
 });
@@ -85878,7 +85928,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   let hasVision = false;
   let isStreaming = false;
   let modelId = "";
-  let maxTokens = 4096;
+  let maxTokens = DEFAULT_MAX_TOKENS;
   let routingProfile = null;
   let stickyExplicitModel;
   let balanceFallbackNotice;
@@ -85897,7 +85947,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       const parsed = JSON.parse(body.toString());
       isStreaming = parsed.stream === true;
       modelId = parsed.model || "";
-      maxTokens = parsed.max_tokens || 4096;
+      maxTokens = resolveMaxTokens(parsed);
       let bodyModified = false;
       const parsedMessages = Array.isArray(parsed.messages) ? parsed.messages : [];
       const lastUserMsg = [...parsedMessages].reverse().find((m) => m.role === "user");
@@ -87894,6 +87944,7 @@ var init_proxy = __esm({
     init_router();
     init_rules();
     init_models();
+    init_max_tokens();
     init_logger();
     init_stats();
     init_dedup();
@@ -87911,6 +87962,7 @@ var init_proxy = __esm({
     init_textual_tool_calls();
     init_response_store();
     init_share_formatters();
+    init_max_tokens();
     paymentStore = new AsyncLocalStorage();
     BLOCKRUN_API = "https://blockrun.ai/api";
     BLOCKRUN_SOLANA_API = "https://sol.blockrun.ai/api";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blockrun/clawrouter",
-  "version": "0.12.230",
+  "version": "0.12.232",
   "description": "Smart LLM router — save 85% on inference costs. 55+ models (8 free), one wallet, x402 micropayments.",
   "type": "module",
   "main": "dist/index.js",

package/skills/clawrouter/SKILL.md

@@ -135,7 +135,7 @@ Rules handle ~~80% of requests in <1ms. Only ambiguous queries hit the LLM class
 
 ## Available Models
 
-55+ models including: gpt-5.6-terra [balanced, stable default], gpt-5.6-sol [flagship], gpt-5.6-luna [cost-efficient], gpt-5.5, gpt-5.4, gpt-4o, o3, claude-fable-5, claude-opus-4.8, claude-opus-4.7, claude-opus-4.5, claude-sonnet-5, claude-sonnet-4.6, gemini-3.1-pro, gemini-2.5-flash, deepseek-v4-pro, deepseek-chat, grok-4.5, grok-4.3, grok-build-0.1, kimi-k3 [1M ctx flagship], kimi-k2.7, and 8 free NVIDIA models (mistral-large-3-675b [default, 675B flagship], deepseek-v4-flash [1M context], seed-oss-36b [coding], nemotron-3-nano-omni-30b-a3b-reasoning [vision], mistral-nemotron, step-3.7-flash, nemotron-nano-9b-v2, nemotron-nano-12b-v2-vl [vision]).
+55+ models including: gpt-5.6-terra [balanced, stable default], gpt-5.6-sol [flagship], gpt-5.6-luna [cost-efficient], gpt-5.5, gpt-5.4, gpt-4o, o3, claude-fable-5, claude-opus-4.8, claude-opus-4.7, claude-opus-4.5, claude-sonnet-5, claude-sonnet-4.6, gemini-3.1-pro, gemini-2.5-flash, deepseek-v4-pro, deepseek-chat, grok-4.5, grok-4.3, grok-build-0.1, kimi-k3 [1M ctx flagship], kimi-k2.7, qwen3.7-max [Qwen flagship, 1M ctx], and 8 free NVIDIA models (mistral-large-3-675b [default, 675B flagship], deepseek-v4-flash [1M context], seed-oss-36b [coding], nemotron-3-nano-omni-30b-a3b-reasoning [vision], mistral-nemotron, step-3.7-flash, nemotron-nano-9b-v2, nemotron-nano-12b-v2-vl [vision]).
 
 ## Built-in Agent Tools
 

package/src/top-models.json

@@ -32,6 +32,7 @@
   "minimax/minimax-m3",
   "minimax/minimax-m2.7",
   "moonshot/kimi-k3",
+  "qwen/qwen3.7-max",
   "deepseek/deepseek-v4-pro",
   "deepseek/deepseek-chat",
   "deepseek/deepseek-reasoner",