@blockrun/llm 3.24.0 → 3.24.2

package/dist/index.cjs

@@ -571,7 +571,69 @@ function withCost(body, response, bookedUsd, scheme) {
 // src/client.ts
 var import_accounts2 = require("viem/accounts");
 
-// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+814fd3b6bc788_xq5fkcckbb2p7dtp6dodey4yri/node_modules/@blockrun/router-core/dist/index.js
+// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+65a92c85ea1f1_ulf6q3mq2pzpwvcnu73z4krxvy/node_modules/@blockrun/router-core/dist/index.js
+var UNSEGMENTED_SCRIPT = /[؀-ۿ　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/;
+var WORD_CHAR = /[\p{L}\p{N}]/u;
+var ASCII_PHRASE = /^[a-z0-9' -]+$/;
+var SEGMENTED_WORD_CHAR = "\\p{Script=Latin}\\p{Script=Cyrillic}\\p{Script=Greek}\\p{N}";
+var BOUNDARY_BEFORE = `(?<![${SEGMENTED_WORD_CHAR}])`;
+var BOUNDARY_AFTER = `(?![${SEGMENTED_WORD_CHAR}])`;
+var BOUNDED_LAST_WORD_LENGTH = 6;
+var SHORT_SUFFIXES = "(?:s|ed|ing)?";
+var SIBILANT_PLURAL = "es";
+var MID_SUFFIXES = "(?:s|es|ed|d|ing|er|ers|ly|ment|ments|ation|ations|e|en|n|a|as|o|os|em|et|ung|ungen)?";
+var DROPPED_E_SUFFIXES = "(?:ing|ed|es|er|ers|ation|ations)";
+var DOUBLED_CONSONANT_SUFFIXES = "(?:ing|ed|er|ers)";
+var matcherCache = /* @__PURE__ */ new Map();
+function escapeRegExp(value) {
+  return value.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+function keywordMatcher(keyword) {
+  const cached = matcherCache.get(keyword);
+  if (cached) return cached;
+  const kw = keyword.toLowerCase();
+  let matcher;
+  if (UNSEGMENTED_SCRIPT.test(kw) || kw.length === 0) {
+    matcher = { needle: kw, pattern: null };
+  } else {
+    const lead = WORD_CHAR.test(kw[0]) ? BOUNDARY_BEFORE : "";
+    const escaped = escapeRegExp(kw);
+    let body = escaped;
+    let trail = "";
+    let needle = kw;
+    const lastWord = kw.slice(kw.lastIndexOf(" ") + 1);
+    if (WORD_CHAR.test(kw[kw.length - 1]) && ASCII_PHRASE.test(kw) && lastWord.length <= BOUNDED_LAST_WORD_LENGTH) {
+      const forms = [];
+      if (/\d$/.test(kw)) {
+        forms.push(escaped);
+      } else if (lastWord.length <= 3) {
+        forms.push(`${escaped}${SHORT_SUFFIXES}`);
+        if (/(?:x|s|z|ch|sh)$/.test(kw)) forms.push(`${escaped}${SIBILANT_PLURAL}`);
+      } else {
+        forms.push(`${escaped}${MID_SUFFIXES}`);
+        if (kw.endsWith("e")) {
+          forms.push(`${escaped.slice(0, -1)}${DROPPED_E_SUFFIXES}`);
+          needle = kw.slice(0, -1);
+        } else if (kw.endsWith("y")) {
+          forms.push(`${escaped.slice(0, -1)}(?:ies|ied)`);
+          needle = kw.slice(0, -1);
+        } else if (/[aeiou][bdgmnprt]$/.test(kw)) {
+          forms.push(`${escaped}${kw[kw.length - 1]}${DOUBLED_CONSONANT_SUFFIXES}`);
+        }
+      }
+      body = `(?:${forms.join("|")})`;
+      trail = BOUNDARY_AFTER;
+    }
+    matcher = { needle, pattern: new RegExp(`${lead}${body}${trail}`, "u") };
+  }
+  matcherCache.set(keyword, matcher);
+  return matcher;
+}
+function matchesKeyword(text, keyword) {
+  const { needle, pattern } = keywordMatcher(keyword);
+  if (!text.includes(needle)) return false;
+  return pattern === null ? true : pattern.test(text);
+}
 function scoreTokenCount(estimatedTokens, thresholds) {
   if (estimatedTokens < thresholds.simple) {
     return { name: "tokenCount", score: -1, signal: `short (${estimatedTokens} tokens)` };
@@ -582,7 +644,7 @@ function scoreTokenCount(estimatedTokens, thresholds) {
   return { name: "tokenCount", score: 0, signal: null };
 }
 function scoreKeywordMatch(text, keywords, name, signalLabel, thresholds, scores) {
-  const matches = keywords.filter((kw) => text.includes(kw.toLowerCase()));
+  const matches = keywords.filter((kw) => matchesKeyword(text, kw));
   if (matches.length >= thresholds.high) {
     return {
       name,
@@ -618,7 +680,7 @@ function scoreAgenticTask(text, keywords) {
   let matchCount = 0;
   const signals = [];
   for (const keyword of keywords) {
-    if (text.includes(keyword.toLowerCase())) {
+    if (matchesKeyword(text, keyword)) {
       matchCount++;
       if (signals.length < 3) {
         signals.push(keyword);
@@ -765,9 +827,7 @@ function classifyByRules(prompt, systemPrompt, estimatedTokens, config) {
     const w = weights[d.name] ?? 0;
     weightedScore += d.score * w;
   }
-  const reasoningMatches = config.reasoningKeywords.filter(
-    (kw) => userText.includes(kw.toLowerCase())
-  );
+  const reasoningMatches = config.reasoningKeywords.filter((kw) => matchesKeyword(userText, kw));
   if (reasoningMatches.length >= 2) {
     const confidence2 = calibrateConfidence(
       Math.max(weightedScore, 0.3),
@@ -832,7 +892,7 @@ function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPric
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
   const baselineOutput = maxOutputTokens / 1e6 * opusOutputPrice;
   const baselineCost = baselineInput + baselineOutput;
-  const savings = routingProfile === "premium" ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
+  const savings = routingProfile === "premium" || pricing === void 0 ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
   return {
     model,
     tier,
@@ -883,6 +943,21 @@ function filterCandidatesByCapacity(models, estimatedInputTokens, requestedOutpu
   });
   return filtered;
 }
+var CJK_CHARS = /[　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/g;
+function estimateTokens(text) {
+  if (text.length === 0) return 0;
+  const cjk = text.match(CJK_CHARS)?.length ?? 0;
+  return cjk + Math.ceil((text.length - cjk) / 4);
+}
+function scanLimit(promptTruncationChars) {
+  return Math.max(1, Math.min(8e3, promptTruncationChars));
+}
+function sampleForScan(value, limit) {
+  if (value.length <= limit) return value;
+  const prefixLength = Math.ceil(limit / 2);
+  return `${value.slice(0, prefixLength)}
+${value.slice(-(limit - prefixLength))}`;
+}
 function applyPromotions(tierConfigs, promotions, profile, now = /* @__PURE__ */ new Date()) {
   if (!promotions || promotions.length === 0) return tierConfigs;
   let result = tierConfigs;
@@ -925,16 +1000,10 @@ var RulesStrategy = class {
   route(prompt, systemPrompt, maxOutputTokens, options) {
     const { config, modelPricing } = options;
     const fullText = `${systemPrompt ?? ""} ${prompt}`;
-    const estimatedTokens = Math.ceil(fullText.length / 4);
-    const scanLimit = Math.max(1, Math.min(8e3, config.classifier.promptTruncationChars));
-    const sample = (value) => {
-      if (value.length <= scanLimit) return value;
-      const prefixLength = Math.ceil(scanLimit / 2);
-      return `${value.slice(0, prefixLength)}
-${value.slice(-(scanLimit - prefixLength))}`;
-    };
-    const scannedPrompt = sample(prompt);
-    const scannedSystemPrompt = systemPrompt ? sample(systemPrompt) : void 0;
+    const estimatedTokens = estimateTokens(fullText);
+    const limit = scanLimit(config.classifier.promptTruncationChars);
+    const scannedPrompt = sampleForScan(prompt, limit);
+    const scannedSystemPrompt = systemPrompt ? sampleForScan(systemPrompt, limit) : void 0;
     const ruleResult = classifyByRules(
       scannedPrompt,
       scannedSystemPrompt,
@@ -988,7 +1057,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
       );
       return { ...decision2, tierConfigs, profile };
     }
-    const hasStructuredOutput = options.requiresStructuredOutput === true || (scannedSystemPrompt ? /json|structured|schema/i.test(scannedSystemPrompt) : false);
+    const hasStructuredOutput = options.requiresStructuredOutput === true;
     let tier;
     let confidence;
     const method = "rules";
@@ -1044,6 +1113,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-fable-5.1": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "anthropic/claude-haiku-4.5": {
     // override: The public catalog's `categories` omit "vision" for this Anthropic model even though the gateway accepts image input for it (the prior hand-maintained snapshot had it, and Anthropic's model card lists it). Without this the vision filter would silently drop it — reported against the catalog; remove once the categories carry vision.
     contextWindow: 2e5,
@@ -1075,6 +1150,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-opus-5.5": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "anthropic/claude-sonnet-4.5": {
     contextWindow: 2e5,
     maxOutputTokens: 64e3,
@@ -1094,6 +1175,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-sonnet-5.5": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "cohere/north-mini-code": {
     // supportsTools: not probed — fails closed
     contextWindow: 256e3,
@@ -1113,6 +1200,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "deepseek/deepseek-v4-flash-vision-exp": {
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: false,
+    supportsVision: true
+  },
   "deepseek/deepseek-v4-pro": {
     contextWindow: 1048576,
     maxOutputTokens: 65536,
@@ -1173,6 +1266,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "google/gemini-3.8-flash": {
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: true,
+    supportsVision: true
+  },
   "minimax/minimax-m2.7": {
     contextWindow: 204800,
     maxOutputTokens: 16384,
@@ -1198,7 +1297,7 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: false,
     supportsVision: true
   },
-  "nvidia/nemotron-3-nano-30b": {
+  "nvidia/muse-glimmer-30b": {
     // supportsTools: not probed — fails closed
     contextWindow: 131072,
     maxOutputTokens: 16384,
@@ -1206,10 +1305,11 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: false
   },
   "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning": {
+    // override: The catalog tags this model "vision", but a correctly sized probe does not hold up (ClawRouter, 2026-08-31): a 64x64 solid-red PNG was named correctly 1 of 4 times on Base, and on Solana the image was silently dropped and a text model answered "white". An HTTP 200 with a confident wrong answer gives the caller nothing to branch on, so image turns must not be routed here. It is ecoTiers.SIMPLE.fallback[0] since nemotron-3-nano-30b was delisted (2026-09-08); remove once a probe of this size comes back right on both chains.
     contextWindow: 256e3,
     maxOutputTokens: 16384,
     supportsTools: false,
-    supportsVision: true
+    supportsVision: false
   },
   "nvidia/nemotron-3-ultra-550b": {
     // supportsTools: not probed — fails closed
@@ -1267,6 +1367,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "openai/gpt-5.1": {
+    contextWindow: 4e5,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "openai/gpt-5.2": {
     contextWindow: 4e5,
     maxOutputTokens: 128e3,
@@ -1361,6 +1467,24 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "openai/gpt-6-astra": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "openai/gpt-6-luna": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "openai/gpt-6-sol": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "openai/o1": {
     contextWindow: 2e5,
     maxOutputTokens: 1e5,
@@ -1385,6 +1509,13 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "poolside/laguna-s-2.1": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: false
+  },
   "poolside/laguna-xs-2.1": {
     // supportsTools: not probed — fails closed
     contextWindow: 131072,
@@ -1410,10 +1541,24 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
-  "tencent/hy3": {
-    contextWindow: 262144,
-    maxOutputTokens: 128e3,
-    supportsTools: true,
+  "qwen/qwen3.8-flash": {
+    contextWindow: 1e6,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: true
+  },
+  "qwen/qwen3.8-max": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1e6,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: false
+  },
+  "tencent/hy4-preview": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 64e3,
+    supportsTools: false,
     supportsVision: false
   },
   "xai/grok-4.3": {
@@ -1428,12 +1573,30 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "xai/grok-4.6": {
+    contextWindow: 5e5,
+    maxOutputTokens: 16384,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "xai/grok-4.7": {
+    contextWindow: 5e5,
+    maxOutputTokens: 16384,
+    supportsTools: true,
+    supportsVision: true
+  },
   "xai/grok-build-0.1": {
     contextWindow: 256e3,
     maxOutputTokens: 16384,
     supportsTools: true,
     supportsVision: false
   },
+  "xiaomi/mimo-v2.5": {
+    contextWindow: 1048576,
+    maxOutputTokens: 131072,
+    supportsTools: true,
+    supportsVision: true
+  },
   "xiaomi/mimo-v2.5-pro": {
     contextWindow: 1048576,
     maxOutputTokens: 131072,
@@ -2147,16 +2310,10 @@ function likelyNeedsParallelToolCalls(prompt, needsTools, toolCount, toolNames)
 }
 function classifyTask(prompt, systemPrompt, options) {
   const fullText = `${systemPrompt ?? ""} ${prompt}`;
-  const estimatedInputTokens = Math.ceil(fullText.length / 4);
-  const scanLimit = Math.max(1, Math.min(8e3, options.config.classifier.promptTruncationChars));
-  const sample = (value) => {
-    if (value.length <= scanLimit) return value;
-    const prefixLength = Math.ceil(scanLimit / 2);
-    return `${value.slice(0, prefixLength)}
-${value.slice(-(scanLimit - prefixLength))}`;
-  };
-  const scannedPrompt = sample(prompt);
-  const scannedSystemPrompt = sample(systemPrompt ?? "");
+  const estimatedInputTokens = estimateTokens(fullText);
+  const limit = scanLimit(options.config.classifier.promptTruncationChars);
+  const scannedPrompt = sampleForScan(prompt, limit);
+  const scannedSystemPrompt = sampleForScan(systemPrompt ?? "", limit);
   const scannedFullText = `${scannedSystemPrompt} ${scannedPrompt}`;
   const text = scannedPrompt.toLowerCase();
   const explicitCodeSignal = /```|\b(?:typescript|javascript|python|rust|java|sql|stack trace|traceback|exception)\b|\.(?:ts|tsx|js|py|go|rs)\b/i.test(
@@ -2168,7 +2325,9 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   const nativeCodeSignal = /\b(?:programmed|written|implemented?|code)\s+(?:in|using)\s+(?:c\+\+|c|rust|go)\b/i.test(
     scannedPrompt
   );
-  const hasCode = explicitCodeSignal || codeConstructSignal || nativeCodeSignal;
+  const errorClassSignal = /\b[A-Z][A-Za-z]*(?:Error|Exception)\b/.test(scannedPrompt);
+  const hasCode = explicitCodeSignal || codeConstructSignal || nativeCodeSignal || errorClassSignal;
+  const codeArtifact = /```|\b(?:stack trace|traceback)\b|\.(?:ts|tsx|js|py|go|rs)\b/i.test(scannedPrompt) || codeConstructSignal || nativeCodeSignal || errorClassSignal;
   const toolsAvailable = options.hasTools ?? false;
   const needsTools = options.requiresTools ?? (toolsAvailable && inferToolRequirement(scannedPrompt, scannedSystemPrompt));
   const likelyParallelToolCalls = likelyNeedsParallelToolCalls(
@@ -2240,10 +2399,10 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   const needsVision = options.hasVision ?? false;
   const needsStructuredOutput = options.requiresStructuredOutput ?? false;
   const latencySensitive = /\b(?:urgent|asap|fast|quick|low latency|real[- ]time)\b|尽快|马上|快速|低延迟/i.test(
-    scannedFullText
+    scannedPrompt
   );
   const highStakes = /\b(?:production|security|payment|legal|medical|financial|audit)\b|生产|安全|支付|法律|医疗|财务|审计/i.test(
-    scannedFullText
+    scannedPrompt
   );
   const terminalToolSignal = normalizedToolNames.some(
     (name) => /^(?:terminalexec|terminalinspect|terminalsendkeys)$/.test(name)
@@ -2279,7 +2438,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   );
   const terminalSafetySensitive = terminalToolSignal && (highStakes || terminalCredentialSignal);
   const implicitTerminalCode = needsTools && terminalToolSignal && agentRisk === "standard" && !highStakes && !complexTerminalOperation && numberedSteps < 3 && prompt.length <= 1e3 && simpleTerminalArtifact;
-  const language = /[\u3400-\u9fff]/.test(scannedFullText) ? "zh" : "other";
+  const language = /[\u3400-\u9fff]/.test(scannedPrompt) ? "zh" : "other";
   const multipleChoiceSignals = (scannedPrompt.match(/(?:^|\n)\s*[A-D][.)]\s+/gim) ?? []).length;
   const numericSignals = (scannedPrompt.match(/-?\d+(?:[.,]\d+)?/g) ?? []).length;
   const compactMathProblem = !hasCode && prompt.length < 2500 && numericSignals >= 2 && (/[+×÷=%$€£¥]|\b(?:total|each|per|times|half|twice|percent|how many|how much|calculate)\b/i.test(
@@ -2294,7 +2453,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   else if (needsTools) taskType = "tool_agent";
   else if (multipleChoiceSignals >= 3) taskType = "reasoning_mcq";
   else if (compactMathProblem) taskType = "reasoning_math";
-  else if (/\b(?:bug|debug|error|failure|failing|regression|crash|修复|报错|错误|调试)\b/i.test(text))
+  else if (errorClassSignal || /\b(?:bug|debug|error|failure|failing|regression|crash|修复|报错|错误|调试)\b/i.test(text))
     taskType = "debug";
   else if (hasCode || /\b(?:refactor|implement|patch|edit|rewrite|重构|实现|修改)\b/i.test(text))
     taskType = "code_edit";
@@ -2320,7 +2479,8 @@ ${value.slice(-(scanLimit - prefixLength))}`;
     agentRisk,
     terminalToolSignal,
     terminalSafetySensitive,
-    implicitTerminalCode
+    implicitTerminalCode,
+    codeArtifact
   };
 }
 function affinity(modelId, task, language = "other", agentDomain = "other", deepWebResearch = false, agentRisk = "standard", terminalToolSignal = false, terminalSafetySensitive = false) {
@@ -2676,6 +2836,19 @@ function evidenceCandidates(task) {
   }
   return [];
 }
+var VERIFIED_TOOL_FALLBACKS = [
+  "anthropic/claude-fable-5.1",
+  "anthropic/claude-opus-5.5",
+  "anthropic/claude-sonnet-5.5",
+  "google/gemini-3.8-flash",
+  "openai/gpt-5.1",
+  "openai/gpt-6-astra",
+  "openai/gpt-6-luna",
+  "openai/gpt-6-sol",
+  "xai/grok-4.6",
+  "xai/grok-4.7",
+  "xiaomi/mimo-v2.5"
+];
 function isEligible(modelId, features, maxOutputTokens, options) {
   const model = options.modelCapabilities?.[modelId] ?? DEFAULT_MODEL_CAPABILITIES[modelId];
   if (!model) return true;
@@ -2710,27 +2883,38 @@ function profileScore(modelId, options, now) {
 }
 var PortfolioStrategy = class {
   name = "portfolio";
+  // Rules set the capability tier; the strategy is stateless, so one instance serves every request.
+  rules = new RulesStrategy();
   route(prompt, systemPrompt, maxOutputTokens, options) {
     const features = classifyTask(prompt, systemPrompt, options);
-    const base = new RulesStrategy().route(prompt, systemPrompt, maxOutputTokens, {
+    const base = this.rules.route(prompt, systemPrompt, maxOutputTokens, {
       ...options,
       requiresTools: features.needsTools
     });
     const tierConfigs = base.tierConfigs;
     if (!tierConfigs) return base;
-    const targetTier = (features.taskType === "reasoning_mcq" || features.taskType === "reasoning_math") && (base.tier === "SIMPLE" || base.tier === "MEDIUM") ? "REASONING" : base.tier;
+    const targetTier = (features.taskType === "reasoning_mcq" || features.taskType === "reasoning_math") && (base.tier === "SIMPLE" || base.tier === "MEDIUM") ? "REASONING" : (features.taskType === "debug" || features.taskType === "code_edit") && features.codeArtifact && base.tier === "SIMPLE" ? "MEDIUM" : base.tier;
     const tierConfig = tierConfigs[targetTier];
     const configuredCandidates = tierConfig ? getFallbackChain(targetTier, tierConfigs) : [];
     const unavailable = new Set(options.unavailableModels ?? []);
+    const verifiedToolFallbacks = features.needsTools ? VERIFIED_TOOL_FALLBACKS.filter((model2) => options.modelPricing.has(model2)) : [];
+    if (options.routingProfile !== "premium") {
+      const fallbackCost = (model2) => estimatedCost(model2, options, features.estimatedInputTokens, maxOutputTokens);
+      verifiedToolFallbacks.sort((a, b) => fallbackCost(a) - fallbackCost(b));
+    }
     const chain = [
-      .../* @__PURE__ */ new Set([...configuredCandidates, ...evidenceCandidates(features.taskType)])
+      .../* @__PURE__ */ new Set([
+        ...configuredCandidates,
+        ...evidenceCandidates(features.taskType),
+        ...verifiedToolFallbacks
+      ])
     ].filter(
       (model2) => typeof model2 === "string" && model2.length > 0 && !unavailable.has(model2)
     );
-    const eligible = chain.filter(
-      (model2) => isEligible(model2, features, maxOutputTokens, options)
+    const eligible = new Set(
+      chain.filter((model2) => isEligible(model2, features, maxOutputTokens, options))
     );
-    const eligibleCandidates = eligible.length > 0 ? eligible : chain;
+    const eligibleCandidates = eligible.size > 0 ? [...eligible] : chain;
     if (eligibleCandidates.length === 0) return base;
     const profileName = options.routingProfile === "eco" ? "eco" : options.routingProfile === "premium" ? "premium" : "auto";
     const portfolio = options.config.portfolio ?? DEFAULT_PORTFOLIO_WEIGHTS;
@@ -2765,11 +2949,11 @@ var PortfolioStrategy = class {
     const rankedEntries = candidates.map((model2, index) => {
       const cost = estimatedCost(model2, options, features.estimatedInputTokens, maxOutputTokens);
       const costScore = Number.isFinite(cost) && maxCost > minCost ? 1 - (cost - minCost) / (maxCost - minCost) : 0.5;
-      const capabilityScore = isEligible(model2, features, maxOutputTokens, options) ? 1 : 0;
+      const capabilityScore = eligible.has(model2) ? 1 : 0;
       const profile = profileScore(model2, options, now);
       const observedQuality = profile?.quality === void 0 ? getAffinity(model2) : getAffinity(model2) * (1 - profile.freshness) + profile.quality * profile.freshness;
-      const observedSpeed = profile ? profile.speed * profile.freshness : 0.5;
-      const observedTailSpeed = profile ? profile.tailSpeed * profile.freshness : 0.5;
+      const observedSpeed = profile ? profile.speed * profile.freshness + 0.5 * (1 - profile.freshness) : 0.5;
+      const observedTailSpeed = profile ? profile.tailSpeed * profile.freshness + 0.5 * (1 - profile.freshness) : 0.5;
       const observedReliability = profile ? profile.reliability * profile.freshness + (1 - profile.freshness) : 1;
       const legacyScore = 1 - index / Math.max(1, candidates.length - 1);
       const qualityWeight = profileWeights.quality + (features.highStakes ? portfolio.highStakesBoost.quality : 0);
@@ -4026,8 +4210,13 @@ var DEFAULT_ROUTING_CONFIG = {
       primary: "nvidia/nemotron-3.5-lightning",
       // FREE — NVIDIA free tier flagship, 1M ctx
       fallback: [
-        "nvidia/nemotron-3-nano-30b",
-        // FREE — fastest free model (~121 tok/s)
+        // Was nvidia/nemotron-3-nano-30b until blockrun delisted it on 2026-09-08
+        // (NVIDIA deprovisioned it for the account — a structured per-account
+        // 404). nano-omni is blockrun's own redirect target for that id, so this
+        // follows the same rule as the head: the router and the gateway name the
+        // same model.
+        "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+        // FREE — nano-30b's successor, 256K ctx
         // The free head keeps rotting with NVIDIA's hosting (deepseek-v4-flash
         // 410 2026-08-12, seed-oss-36b 410 2026-08-03, gpt-oss-120b/20b 400
         // 2026-08-21, and on 2026-08-30 FOUR of the five visible free models at
@@ -4856,7 +5045,7 @@ function getCostSummary() {
 }
 
 // src/version.ts
-var SDK_VERSION = "3.24.0";
+var SDK_VERSION = "3.24.2";
 var USER_AGENT = `blockrun-ts/${SDK_VERSION}`;
 
 // src/client.ts

package/dist/index.js

@@ -440,7 +440,69 @@ function withCost(body, response, bookedUsd, scheme) {
 // src/client.ts
 import { privateKeyToAccount } from "viem/accounts";
 
-// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+814fd3b6bc788_xq5fkcckbb2p7dtp6dodey4yri/node_modules/@blockrun/router-core/dist/index.js
+// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+65a92c85ea1f1_ulf6q3mq2pzpwvcnu73z4krxvy/node_modules/@blockrun/router-core/dist/index.js
+var UNSEGMENTED_SCRIPT = /[؀-ۿ　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/;
+var WORD_CHAR = /[\p{L}\p{N}]/u;
+var ASCII_PHRASE = /^[a-z0-9' -]+$/;
+var SEGMENTED_WORD_CHAR = "\\p{Script=Latin}\\p{Script=Cyrillic}\\p{Script=Greek}\\p{N}";
+var BOUNDARY_BEFORE = `(?<![${SEGMENTED_WORD_CHAR}])`;
+var BOUNDARY_AFTER = `(?![${SEGMENTED_WORD_CHAR}])`;
+var BOUNDED_LAST_WORD_LENGTH = 6;
+var SHORT_SUFFIXES = "(?:s|ed|ing)?";
+var SIBILANT_PLURAL = "es";
+var MID_SUFFIXES = "(?:s|es|ed|d|ing|er|ers|ly|ment|ments|ation|ations|e|en|n|a|as|o|os|em|et|ung|ungen)?";
+var DROPPED_E_SUFFIXES = "(?:ing|ed|es|er|ers|ation|ations)";
+var DOUBLED_CONSONANT_SUFFIXES = "(?:ing|ed|er|ers)";
+var matcherCache = /* @__PURE__ */ new Map();
+function escapeRegExp(value) {
+  return value.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+function keywordMatcher(keyword) {
+  const cached = matcherCache.get(keyword);
+  if (cached) return cached;
+  const kw = keyword.toLowerCase();
+  let matcher;
+  if (UNSEGMENTED_SCRIPT.test(kw) || kw.length === 0) {
+    matcher = { needle: kw, pattern: null };
+  } else {
+    const lead = WORD_CHAR.test(kw[0]) ? BOUNDARY_BEFORE : "";
+    const escaped = escapeRegExp(kw);
+    let body = escaped;
+    let trail = "";
+    let needle = kw;
+    const lastWord = kw.slice(kw.lastIndexOf(" ") + 1);
+    if (WORD_CHAR.test(kw[kw.length - 1]) && ASCII_PHRASE.test(kw) && lastWord.length <= BOUNDED_LAST_WORD_LENGTH) {
+      const forms = [];
+      if (/\d$/.test(kw)) {
+        forms.push(escaped);
+      } else if (lastWord.length <= 3) {
+        forms.push(`${escaped}${SHORT_SUFFIXES}`);
+        if (/(?:x|s|z|ch|sh)$/.test(kw)) forms.push(`${escaped}${SIBILANT_PLURAL}`);
+      } else {
+        forms.push(`${escaped}${MID_SUFFIXES}`);
+        if (kw.endsWith("e")) {
+          forms.push(`${escaped.slice(0, -1)}${DROPPED_E_SUFFIXES}`);
+          needle = kw.slice(0, -1);
+        } else if (kw.endsWith("y")) {
+          forms.push(`${escaped.slice(0, -1)}(?:ies|ied)`);
+          needle = kw.slice(0, -1);
+        } else if (/[aeiou][bdgmnprt]$/.test(kw)) {
+          forms.push(`${escaped}${kw[kw.length - 1]}${DOUBLED_CONSONANT_SUFFIXES}`);
+        }
+      }
+      body = `(?:${forms.join("|")})`;
+      trail = BOUNDARY_AFTER;
+    }
+    matcher = { needle, pattern: new RegExp(`${lead}${body}${trail}`, "u") };
+  }
+  matcherCache.set(keyword, matcher);
+  return matcher;
+}
+function matchesKeyword(text, keyword) {
+  const { needle, pattern } = keywordMatcher(keyword);
+  if (!text.includes(needle)) return false;
+  return pattern === null ? true : pattern.test(text);
+}
 function scoreTokenCount(estimatedTokens, thresholds) {
   if (estimatedTokens < thresholds.simple) {
     return { name: "tokenCount", score: -1, signal: `short (${estimatedTokens} tokens)` };
@@ -451,7 +513,7 @@ function scoreTokenCount(estimatedTokens, thresholds) {
   return { name: "tokenCount", score: 0, signal: null };
 }
 function scoreKeywordMatch(text, keywords, name, signalLabel, thresholds, scores) {
-  const matches = keywords.filter((kw) => text.includes(kw.toLowerCase()));
+  const matches = keywords.filter((kw) => matchesKeyword(text, kw));
   if (matches.length >= thresholds.high) {
     return {
       name,
@@ -487,7 +549,7 @@ function scoreAgenticTask(text, keywords) {
   let matchCount = 0;
   const signals = [];
   for (const keyword of keywords) {
-    if (text.includes(keyword.toLowerCase())) {
+    if (matchesKeyword(text, keyword)) {
       matchCount++;
       if (signals.length < 3) {
         signals.push(keyword);
@@ -634,9 +696,7 @@ function classifyByRules(prompt, systemPrompt, estimatedTokens, config) {
     const w = weights[d.name] ?? 0;
     weightedScore += d.score * w;
   }
-  const reasoningMatches = config.reasoningKeywords.filter(
-    (kw) => userText.includes(kw.toLowerCase())
-  );
+  const reasoningMatches = config.reasoningKeywords.filter((kw) => matchesKeyword(userText, kw));
   if (reasoningMatches.length >= 2) {
     const confidence2 = calibrateConfidence(
       Math.max(weightedScore, 0.3),
@@ -701,7 +761,7 @@ function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPric
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
   const baselineOutput = maxOutputTokens / 1e6 * opusOutputPrice;
   const baselineCost = baselineInput + baselineOutput;
-  const savings = routingProfile === "premium" ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
+  const savings = routingProfile === "premium" || pricing === void 0 ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
   return {
     model,
     tier,
@@ -752,6 +812,21 @@ function filterCandidatesByCapacity(models, estimatedInputTokens, requestedOutpu
   });
   return filtered;
 }
+var CJK_CHARS = /[　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/g;
+function estimateTokens(text) {
+  if (text.length === 0) return 0;
+  const cjk = text.match(CJK_CHARS)?.length ?? 0;
+  return cjk + Math.ceil((text.length - cjk) / 4);
+}
+function scanLimit(promptTruncationChars) {
+  return Math.max(1, Math.min(8e3, promptTruncationChars));
+}
+function sampleForScan(value, limit) {
+  if (value.length <= limit) return value;
+  const prefixLength = Math.ceil(limit / 2);
+  return `${value.slice(0, prefixLength)}
+${value.slice(-(limit - prefixLength))}`;
+}
 function applyPromotions(tierConfigs, promotions, profile, now = /* @__PURE__ */ new Date()) {
   if (!promotions || promotions.length === 0) return tierConfigs;
   let result = tierConfigs;
@@ -794,16 +869,10 @@ var RulesStrategy = class {
   route(prompt, systemPrompt, maxOutputTokens, options) {
     const { config, modelPricing } = options;
     const fullText = `${systemPrompt ?? ""} ${prompt}`;
-    const estimatedTokens = Math.ceil(fullText.length / 4);
-    const scanLimit = Math.max(1, Math.min(8e3, config.classifier.promptTruncationChars));
-    const sample = (value) => {
-      if (value.length <= scanLimit) return value;
-      const prefixLength = Math.ceil(scanLimit / 2);
-      return `${value.slice(0, prefixLength)}
-${value.slice(-(scanLimit - prefixLength))}`;
-    };
-    const scannedPrompt = sample(prompt);
-    const scannedSystemPrompt = systemPrompt ? sample(systemPrompt) : void 0;
+    const estimatedTokens = estimateTokens(fullText);
+    const limit = scanLimit(config.classifier.promptTruncationChars);
+    const scannedPrompt = sampleForScan(prompt, limit);
+    const scannedSystemPrompt = systemPrompt ? sampleForScan(systemPrompt, limit) : void 0;
     const ruleResult = classifyByRules(
       scannedPrompt,
       scannedSystemPrompt,
@@ -857,7 +926,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
       );
       return { ...decision2, tierConfigs, profile };
     }
-    const hasStructuredOutput = options.requiresStructuredOutput === true || (scannedSystemPrompt ? /json|structured|schema/i.test(scannedSystemPrompt) : false);
+    const hasStructuredOutput = options.requiresStructuredOutput === true;
     let tier;
     let confidence;
     const method = "rules";
@@ -913,6 +982,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-fable-5.1": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "anthropic/claude-haiku-4.5": {
     // override: The public catalog's `categories` omit "vision" for this Anthropic model even though the gateway accepts image input for it (the prior hand-maintained snapshot had it, and Anthropic's model card lists it). Without this the vision filter would silently drop it — reported against the catalog; remove once the categories carry vision.
     contextWindow: 2e5,
@@ -944,6 +1019,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-opus-5.5": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "anthropic/claude-sonnet-4.5": {
     contextWindow: 2e5,
     maxOutputTokens: 64e3,
@@ -963,6 +1044,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "anthropic/claude-sonnet-5.5": {
+    contextWindow: 1e6,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "cohere/north-mini-code": {
     // supportsTools: not probed — fails closed
     contextWindow: 256e3,
@@ -982,6 +1069,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "deepseek/deepseek-v4-flash-vision-exp": {
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: false,
+    supportsVision: true
+  },
   "deepseek/deepseek-v4-pro": {
     contextWindow: 1048576,
     maxOutputTokens: 65536,
@@ -1042,6 +1135,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "google/gemini-3.8-flash": {
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: true,
+    supportsVision: true
+  },
   "minimax/minimax-m2.7": {
     contextWindow: 204800,
     maxOutputTokens: 16384,
@@ -1067,7 +1166,7 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: false,
     supportsVision: true
   },
-  "nvidia/nemotron-3-nano-30b": {
+  "nvidia/muse-glimmer-30b": {
     // supportsTools: not probed — fails closed
     contextWindow: 131072,
     maxOutputTokens: 16384,
@@ -1075,10 +1174,11 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: false
   },
   "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning": {
+    // override: The catalog tags this model "vision", but a correctly sized probe does not hold up (ClawRouter, 2026-08-31): a 64x64 solid-red PNG was named correctly 1 of 4 times on Base, and on Solana the image was silently dropped and a text model answered "white". An HTTP 200 with a confident wrong answer gives the caller nothing to branch on, so image turns must not be routed here. It is ecoTiers.SIMPLE.fallback[0] since nemotron-3-nano-30b was delisted (2026-09-08); remove once a probe of this size comes back right on both chains.
     contextWindow: 256e3,
     maxOutputTokens: 16384,
     supportsTools: false,
-    supportsVision: true
+    supportsVision: false
   },
   "nvidia/nemotron-3-ultra-550b": {
     // supportsTools: not probed — fails closed
@@ -1136,6 +1236,12 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "openai/gpt-5.1": {
+    contextWindow: 4e5,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "openai/gpt-5.2": {
     contextWindow: 4e5,
     maxOutputTokens: 128e3,
@@ -1230,6 +1336,24 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "openai/gpt-6-astra": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "openai/gpt-6-luna": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "openai/gpt-6-sol": {
+    contextWindow: 105e4,
+    maxOutputTokens: 128e3,
+    supportsTools: true,
+    supportsVision: true
+  },
   "openai/o1": {
     contextWindow: 2e5,
     maxOutputTokens: 1e5,
@@ -1254,6 +1378,13 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
+  "poolside/laguna-s-2.1": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: false
+  },
   "poolside/laguna-xs-2.1": {
     // supportsTools: not probed — fails closed
     contextWindow: 131072,
@@ -1279,10 +1410,24 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: false
   },
-  "tencent/hy3": {
-    contextWindow: 262144,
-    maxOutputTokens: 128e3,
-    supportsTools: true,
+  "qwen/qwen3.8-flash": {
+    contextWindow: 1e6,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: true
+  },
+  "qwen/qwen3.8-max": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1e6,
+    maxOutputTokens: 131072,
+    supportsTools: false,
+    supportsVision: false
+  },
+  "tencent/hy4-preview": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 64e3,
+    supportsTools: false,
     supportsVision: false
   },
   "xai/grok-4.3": {
@@ -1297,12 +1442,30 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "xai/grok-4.6": {
+    contextWindow: 5e5,
+    maxOutputTokens: 16384,
+    supportsTools: true,
+    supportsVision: true
+  },
+  "xai/grok-4.7": {
+    contextWindow: 5e5,
+    maxOutputTokens: 16384,
+    supportsTools: true,
+    supportsVision: true
+  },
   "xai/grok-build-0.1": {
     contextWindow: 256e3,
     maxOutputTokens: 16384,
     supportsTools: true,
     supportsVision: false
   },
+  "xiaomi/mimo-v2.5": {
+    contextWindow: 1048576,
+    maxOutputTokens: 131072,
+    supportsTools: true,
+    supportsVision: true
+  },
   "xiaomi/mimo-v2.5-pro": {
     contextWindow: 1048576,
     maxOutputTokens: 131072,
@@ -2016,16 +2179,10 @@ function likelyNeedsParallelToolCalls(prompt, needsTools, toolCount, toolNames)
 }
 function classifyTask(prompt, systemPrompt, options) {
   const fullText = `${systemPrompt ?? ""} ${prompt}`;
-  const estimatedInputTokens = Math.ceil(fullText.length / 4);
-  const scanLimit = Math.max(1, Math.min(8e3, options.config.classifier.promptTruncationChars));
-  const sample = (value) => {
-    if (value.length <= scanLimit) return value;
-    const prefixLength = Math.ceil(scanLimit / 2);
-    return `${value.slice(0, prefixLength)}
-${value.slice(-(scanLimit - prefixLength))}`;
-  };
-  const scannedPrompt = sample(prompt);
-  const scannedSystemPrompt = sample(systemPrompt ?? "");
+  const estimatedInputTokens = estimateTokens(fullText);
+  const limit = scanLimit(options.config.classifier.promptTruncationChars);
+  const scannedPrompt = sampleForScan(prompt, limit);
+  const scannedSystemPrompt = sampleForScan(systemPrompt ?? "", limit);
   const scannedFullText = `${scannedSystemPrompt} ${scannedPrompt}`;
   const text = scannedPrompt.toLowerCase();
   const explicitCodeSignal = /```|\b(?:typescript|javascript|python|rust|java|sql|stack trace|traceback|exception)\b|\.(?:ts|tsx|js|py|go|rs)\b/i.test(
@@ -2037,7 +2194,9 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   const nativeCodeSignal = /\b(?:programmed|written|implemented?|code)\s+(?:in|using)\s+(?:c\+\+|c|rust|go)\b/i.test(
     scannedPrompt
   );
-  const hasCode = explicitCodeSignal || codeConstructSignal || nativeCodeSignal;
+  const errorClassSignal = /\b[A-Z][A-Za-z]*(?:Error|Exception)\b/.test(scannedPrompt);
+  const hasCode = explicitCodeSignal || codeConstructSignal || nativeCodeSignal || errorClassSignal;
+  const codeArtifact = /```|\b(?:stack trace|traceback)\b|\.(?:ts|tsx|js|py|go|rs)\b/i.test(scannedPrompt) || codeConstructSignal || nativeCodeSignal || errorClassSignal;
   const toolsAvailable = options.hasTools ?? false;
   const needsTools = options.requiresTools ?? (toolsAvailable && inferToolRequirement(scannedPrompt, scannedSystemPrompt));
   const likelyParallelToolCalls = likelyNeedsParallelToolCalls(
@@ -2109,10 +2268,10 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   const needsVision = options.hasVision ?? false;
   const needsStructuredOutput = options.requiresStructuredOutput ?? false;
   const latencySensitive = /\b(?:urgent|asap|fast|quick|low latency|real[- ]time)\b|尽快|马上|快速|低延迟/i.test(
-    scannedFullText
+    scannedPrompt
   );
   const highStakes = /\b(?:production|security|payment|legal|medical|financial|audit)\b|生产|安全|支付|法律|医疗|财务|审计/i.test(
-    scannedFullText
+    scannedPrompt
   );
   const terminalToolSignal = normalizedToolNames.some(
     (name) => /^(?:terminalexec|terminalinspect|terminalsendkeys)$/.test(name)
@@ -2148,7 +2307,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   );
   const terminalSafetySensitive = terminalToolSignal && (highStakes || terminalCredentialSignal);
   const implicitTerminalCode = needsTools && terminalToolSignal && agentRisk === "standard" && !highStakes && !complexTerminalOperation && numberedSteps < 3 && prompt.length <= 1e3 && simpleTerminalArtifact;
-  const language = /[\u3400-\u9fff]/.test(scannedFullText) ? "zh" : "other";
+  const language = /[\u3400-\u9fff]/.test(scannedPrompt) ? "zh" : "other";
   const multipleChoiceSignals = (scannedPrompt.match(/(?:^|\n)\s*[A-D][.)]\s+/gim) ?? []).length;
   const numericSignals = (scannedPrompt.match(/-?\d+(?:[.,]\d+)?/g) ?? []).length;
   const compactMathProblem = !hasCode && prompt.length < 2500 && numericSignals >= 2 && (/[+×÷=%$€£¥]|\b(?:total|each|per|times|half|twice|percent|how many|how much|calculate)\b/i.test(
@@ -2163,7 +2322,7 @@ ${value.slice(-(scanLimit - prefixLength))}`;
   else if (needsTools) taskType = "tool_agent";
   else if (multipleChoiceSignals >= 3) taskType = "reasoning_mcq";
   else if (compactMathProblem) taskType = "reasoning_math";
-  else if (/\b(?:bug|debug|error|failure|failing|regression|crash|修复|报错|错误|调试)\b/i.test(text))
+  else if (errorClassSignal || /\b(?:bug|debug|error|failure|failing|regression|crash|修复|报错|错误|调试)\b/i.test(text))
     taskType = "debug";
   else if (hasCode || /\b(?:refactor|implement|patch|edit|rewrite|重构|实现|修改)\b/i.test(text))
     taskType = "code_edit";
@@ -2189,7 +2348,8 @@ ${value.slice(-(scanLimit - prefixLength))}`;
     agentRisk,
     terminalToolSignal,
     terminalSafetySensitive,
-    implicitTerminalCode
+    implicitTerminalCode,
+    codeArtifact
   };
 }
 function affinity(modelId, task, language = "other", agentDomain = "other", deepWebResearch = false, agentRisk = "standard", terminalToolSignal = false, terminalSafetySensitive = false) {
@@ -2545,6 +2705,19 @@ function evidenceCandidates(task) {
   }
   return [];
 }
+var VERIFIED_TOOL_FALLBACKS = [
+  "anthropic/claude-fable-5.1",
+  "anthropic/claude-opus-5.5",
+  "anthropic/claude-sonnet-5.5",
+  "google/gemini-3.8-flash",
+  "openai/gpt-5.1",
+  "openai/gpt-6-astra",
+  "openai/gpt-6-luna",
+  "openai/gpt-6-sol",
+  "xai/grok-4.6",
+  "xai/grok-4.7",
+  "xiaomi/mimo-v2.5"
+];
 function isEligible(modelId, features, maxOutputTokens, options) {
   const model = options.modelCapabilities?.[modelId] ?? DEFAULT_MODEL_CAPABILITIES[modelId];
   if (!model) return true;
@@ -2579,27 +2752,38 @@ function profileScore(modelId, options, now) {
 }
 var PortfolioStrategy = class {
   name = "portfolio";
+  // Rules set the capability tier; the strategy is stateless, so one instance serves every request.
+  rules = new RulesStrategy();
   route(prompt, systemPrompt, maxOutputTokens, options) {
     const features = classifyTask(prompt, systemPrompt, options);
-    const base = new RulesStrategy().route(prompt, systemPrompt, maxOutputTokens, {
+    const base = this.rules.route(prompt, systemPrompt, maxOutputTokens, {
       ...options,
       requiresTools: features.needsTools
     });
     const tierConfigs = base.tierConfigs;
     if (!tierConfigs) return base;
-    const targetTier = (features.taskType === "reasoning_mcq" || features.taskType === "reasoning_math") && (base.tier === "SIMPLE" || base.tier === "MEDIUM") ? "REASONING" : base.tier;
+    const targetTier = (features.taskType === "reasoning_mcq" || features.taskType === "reasoning_math") && (base.tier === "SIMPLE" || base.tier === "MEDIUM") ? "REASONING" : (features.taskType === "debug" || features.taskType === "code_edit") && features.codeArtifact && base.tier === "SIMPLE" ? "MEDIUM" : base.tier;
     const tierConfig = tierConfigs[targetTier];
     const configuredCandidates = tierConfig ? getFallbackChain(targetTier, tierConfigs) : [];
     const unavailable = new Set(options.unavailableModels ?? []);
+    const verifiedToolFallbacks = features.needsTools ? VERIFIED_TOOL_FALLBACKS.filter((model2) => options.modelPricing.has(model2)) : [];
+    if (options.routingProfile !== "premium") {
+      const fallbackCost = (model2) => estimatedCost(model2, options, features.estimatedInputTokens, maxOutputTokens);
+      verifiedToolFallbacks.sort((a, b) => fallbackCost(a) - fallbackCost(b));
+    }
     const chain = [
-      .../* @__PURE__ */ new Set([...configuredCandidates, ...evidenceCandidates(features.taskType)])
+      .../* @__PURE__ */ new Set([
+        ...configuredCandidates,
+        ...evidenceCandidates(features.taskType),
+        ...verifiedToolFallbacks
+      ])
     ].filter(
       (model2) => typeof model2 === "string" && model2.length > 0 && !unavailable.has(model2)
     );
-    const eligible = chain.filter(
-      (model2) => isEligible(model2, features, maxOutputTokens, options)
+    const eligible = new Set(
+      chain.filter((model2) => isEligible(model2, features, maxOutputTokens, options))
     );
-    const eligibleCandidates = eligible.length > 0 ? eligible : chain;
+    const eligibleCandidates = eligible.size > 0 ? [...eligible] : chain;
     if (eligibleCandidates.length === 0) return base;
     const profileName = options.routingProfile === "eco" ? "eco" : options.routingProfile === "premium" ? "premium" : "auto";
     const portfolio = options.config.portfolio ?? DEFAULT_PORTFOLIO_WEIGHTS;
@@ -2634,11 +2818,11 @@ var PortfolioStrategy = class {
     const rankedEntries = candidates.map((model2, index) => {
       const cost = estimatedCost(model2, options, features.estimatedInputTokens, maxOutputTokens);
       const costScore = Number.isFinite(cost) && maxCost > minCost ? 1 - (cost - minCost) / (maxCost - minCost) : 0.5;
-      const capabilityScore = isEligible(model2, features, maxOutputTokens, options) ? 1 : 0;
+      const capabilityScore = eligible.has(model2) ? 1 : 0;
       const profile = profileScore(model2, options, now);
       const observedQuality = profile?.quality === void 0 ? getAffinity(model2) : getAffinity(model2) * (1 - profile.freshness) + profile.quality * profile.freshness;
-      const observedSpeed = profile ? profile.speed * profile.freshness : 0.5;
-      const observedTailSpeed = profile ? profile.tailSpeed * profile.freshness : 0.5;
+      const observedSpeed = profile ? profile.speed * profile.freshness + 0.5 * (1 - profile.freshness) : 0.5;
+      const observedTailSpeed = profile ? profile.tailSpeed * profile.freshness + 0.5 * (1 - profile.freshness) : 0.5;
       const observedReliability = profile ? profile.reliability * profile.freshness + (1 - profile.freshness) : 1;
       const legacyScore = 1 - index / Math.max(1, candidates.length - 1);
       const qualityWeight = profileWeights.quality + (features.highStakes ? portfolio.highStakesBoost.quality : 0);
@@ -3895,8 +4079,13 @@ var DEFAULT_ROUTING_CONFIG = {
       primary: "nvidia/nemotron-3.5-lightning",
       // FREE — NVIDIA free tier flagship, 1M ctx
       fallback: [
-        "nvidia/nemotron-3-nano-30b",
-        // FREE — fastest free model (~121 tok/s)
+        // Was nvidia/nemotron-3-nano-30b until blockrun delisted it on 2026-09-08
+        // (NVIDIA deprovisioned it for the account — a structured per-account
+        // 404). nano-omni is blockrun's own redirect target for that id, so this
+        // follows the same rule as the head: the router and the gateway name the
+        // same model.
+        "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+        // FREE — nano-30b's successor, 256K ctx
         // The free head keeps rotting with NVIDIA's hosting (deepseek-v4-flash
         // 410 2026-08-12, seed-oss-36b 410 2026-08-03, gpt-oss-120b/20b 400
         // 2026-08-21, and on 2026-08-30 FOUR of the five visible free models at
@@ -4725,7 +4914,7 @@ function getCostSummary() {
 }
 
 // src/version.ts
-var SDK_VERSION = "3.24.0";
+var SDK_VERSION = "3.24.2";
 var USER_AGENT = `blockrun-ts/${SDK_VERSION}`;
 
 // src/client.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blockrun/llm",
-  "version": "3.24.0",
+  "version": "3.24.2",
   "type": "module",
   "description": "TypeScript SDK for BlockRun - every frontier model behind one OpenAI-compatible client. Pay with a BlockRun API key or per-call in USDC. Smart routing picks the cheapest capable model. No rate limits.",
   "main": "dist/index.cjs",
@@ -58,7 +58,7 @@
   },
   "devDependencies": {
     "@blockrun/core": "^0.1.0",
-    "@blockrun/router-core": "https://codeload.github.com/BlockRunAI/router-core/tar.gz/814fd3b6bc7880a84bbe030bd08c98e8670031e7",
+    "@blockrun/router-core": "https://codeload.github.com/BlockRunAI/router-core/tar.gz/65a92c85ea1f187a64e7611302ef6add236b0cba",
     "@eslint/js": "^10.0.1",
     "@solana/kit": "^5.5.1",
     "@types/node": "^26.4.1",