@blockrun/clawrouter 0.9.26 → 0.9.28

package/dist/cli.js

@@ -511,6 +511,7 @@ function calibrateConfidence(distance, steepness) {
 }
 
 // src/router/selector.ts
+var BASELINE_MODEL_ID = "anthropic/claude-opus-4-5";
 function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPricing, estimatedInputTokens, maxOutputTokens, routingProfile) {
   const tierConfig = tierConfigs[tier];
   const model = tierConfig.primary;
@@ -520,7 +521,7 @@ function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPric
   const inputCost = estimatedInputTokens / 1e6 * inputPrice;
   const outputCost = maxOutputTokens / 1e6 * outputPrice;
   const costEstimate = inputCost + outputCost;
-  const opusPricing = modelPricing.get("anthropic/claude-opus-4.5");
+  const opusPricing = modelPricing.get(BASELINE_MODEL_ID);
   const opusInputPrice = opusPricing?.inputPrice ?? 0;
   const opusOutputPrice = opusPricing?.outputPrice ?? 0;
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
@@ -549,7 +550,7 @@ function calculateModelCost(model, modelPricing, estimatedInputTokens, maxOutput
   const inputCost = estimatedInputTokens / 1e6 * inputPrice;
   const outputCost = maxOutputTokens / 1e6 * outputPrice;
   const costEstimate = inputCost + outputCost;
-  const opusPricing = modelPricing.get("anthropic/claude-opus-4.5");
+  const opusPricing = modelPricing.get(BASELINE_MODEL_ID);
   const opusInputPrice = opusPricing?.inputPrice ?? 0;
   const opusOutputPrice = opusPricing?.outputPrice ?? 0;
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
@@ -1237,7 +1238,7 @@ var DEFAULT_ROUTING_CONFIG = {
         "openai/gpt-5.2",
         // Newer and cheaper input than gpt-4o
         "openai/gpt-4o",
-        "anthropic/claude-sonnet-4.6"
+        "anthropic/claude-sonnet-4-6"
       ]
     },
     REASONING: {
@@ -1300,7 +1301,7 @@ var DEFAULT_ROUTING_CONFIG = {
     SIMPLE: {
       primary: "moonshot/kimi-k2.5",
       // $0.50/$2.40 - good for simple coding
-      fallback: ["anthropic/claude-haiku-4.5", "google/gemini-2.5-flash", "xai/grok-code-fast-1"]
+      fallback: ["anthropic/claude-haiku-4-5", "google/gemini-2.5-flash", "xai/grok-code-fast-1"]
     },
     MEDIUM: {
       primary: "openai/gpt-5.2-codex",
@@ -1309,26 +1310,26 @@ var DEFAULT_ROUTING_CONFIG = {
         "moonshot/kimi-k2.5",
         "google/gemini-2.5-pro",
         "xai/grok-4-0709",
-        "anthropic/claude-sonnet-4.6"
+        "anthropic/claude-sonnet-4-6"
       ]
     },
     COMPLEX: {
-      primary: "anthropic/claude-opus-4.6",
+      primary: "anthropic/claude-opus-4-6",
       // Best quality for complex tasks
       fallback: [
         "openai/gpt-5.2-codex",
-        "anthropic/claude-opus-4.5",
-        "anthropic/claude-sonnet-4.6",
+        "anthropic/claude-opus-4-5",
+        "anthropic/claude-sonnet-4-6",
         "google/gemini-3-pro-preview",
         "moonshot/kimi-k2.5"
       ]
     },
     REASONING: {
-      primary: "anthropic/claude-sonnet-4.6",
+      primary: "anthropic/claude-sonnet-4-6",
       // $3/$15 - best for reasoning/instructions
       fallback: [
-        "anthropic/claude-opus-4.6",
-        "anthropic/claude-opus-4.5",
+        "anthropic/claude-opus-4-6",
+        "anthropic/claude-opus-4-5",
         "openai/o4-mini",
         // Newer and cheaper than o3 ($1.10 vs $2.00)
         "openai/o3",
@@ -1344,7 +1345,7 @@ var DEFAULT_ROUTING_CONFIG = {
       fallback: [
         "minimax/minimax-m2.5",
         // $0.30/$1.20 - agentic capable, cheaper than kimi
-        "anthropic/claude-haiku-4.5",
+        "anthropic/claude-haiku-4-5",
         "xai/grok-4-1-fast-non-reasoning",
         "openai/gpt-4o-mini"
       ]
@@ -1356,14 +1357,14 @@ var DEFAULT_ROUTING_CONFIG = {
         "minimax/minimax-m2.5",
         // $0.30/$1.20 - agentic capable
         "moonshot/kimi-k2.5",
-        "anthropic/claude-haiku-4.5",
-        "anthropic/claude-sonnet-4.6"
+        "anthropic/claude-haiku-4-5",
+        "anthropic/claude-sonnet-4-6"
       ]
     },
     COMPLEX: {
-      primary: "anthropic/claude-sonnet-4.6",
+      primary: "anthropic/claude-sonnet-4-6",
       fallback: [
-        "anthropic/claude-opus-4.6",
+        "anthropic/claude-opus-4-6",
         // Latest Opus - best agentic
         "minimax/minimax-m2.5",
         // $0.30/$1.20 - cheap agentic fallback
@@ -1373,10 +1374,10 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     },
     REASONING: {
-      primary: "anthropic/claude-sonnet-4.6",
+      primary: "anthropic/claude-sonnet-4-6",
       // Strong tool use + reasoning for agentic tasks
       fallback: [
-        "anthropic/claude-opus-4.6",
+        "anthropic/claude-opus-4-6",
         "minimax/minimax-m2.5",
         // $0.30/$1.20 - reasoning + agentic
         "xai/grok-4-1-fast-reasoning",
@@ -1465,19 +1466,23 @@ function route(prompt, systemPrompt, maxOutputTokens, options) {
 
 // src/models.ts
 var MODEL_ALIASES = {
-  // Claude - short names
-  claude: "anthropic/claude-sonnet-4.6",
-  sonnet: "anthropic/claude-sonnet-4.6",
-  opus: "anthropic/claude-opus-4.6",
-  // Updated to latest Opus 4.6
-  "opus-46": "anthropic/claude-opus-4.6",
-  "opus-45": "anthropic/claude-opus-4.5",
-  haiku: "anthropic/claude-haiku-4.5",
+  // Claude - short names (use dashes in version, not dots - Anthropic API format)
+  claude: "anthropic/claude-sonnet-4-6",
+  sonnet: "anthropic/claude-sonnet-4-6",
+  opus: "anthropic/claude-opus-4-6",
+  "opus-46": "anthropic/claude-opus-4-6",
+  "opus-45": "anthropic/claude-opus-4-5",
+  haiku: "anthropic/claude-haiku-4-5",
   // Claude - provider/shortname patterns (common in agent frameworks)
-  "anthropic/sonnet": "anthropic/claude-sonnet-4.6",
-  "anthropic/opus": "anthropic/claude-opus-4.6",
-  "anthropic/haiku": "anthropic/claude-haiku-4.5",
-  "anthropic/claude": "anthropic/claude-sonnet-4.6",
+  "anthropic/sonnet": "anthropic/claude-sonnet-4-6",
+  "anthropic/opus": "anthropic/claude-opus-4-6",
+  "anthropic/haiku": "anthropic/claude-haiku-4-5",
+  "anthropic/claude": "anthropic/claude-sonnet-4-6",
+  // Backward compatibility - old dot notation still works
+  "anthropic/claude-sonnet-4.6": "anthropic/claude-sonnet-4-6",
+  "anthropic/claude-opus-4.6": "anthropic/claude-opus-4-6",
+  "anthropic/claude-opus-4.5": "anthropic/claude-opus-4-5",
+  "anthropic/claude-haiku-4.5": "anthropic/claude-haiku-4-5",
   // OpenAI
   gpt: "openai/gpt-4o",
   gpt4: "openai/gpt-4o",
@@ -1666,7 +1671,7 @@ var BLOCKRUN_MODELS = [
   },
   // Anthropic - all Claude models excel at agentic workflows
   {
-    id: "anthropic/claude-haiku-4.5",
+    id: "anthropic/claude-haiku-4-5",
     name: "Claude Haiku 4.5",
     inputPrice: 1,
     outputPrice: 5,
@@ -1675,7 +1680,7 @@ var BLOCKRUN_MODELS = [
     agentic: true
   },
   {
-    id: "anthropic/claude-sonnet-4.6",
+    id: "anthropic/claude-sonnet-4-6",
     name: "Claude Sonnet 4.6",
     inputPrice: 3,
     outputPrice: 15,
@@ -1695,7 +1700,7 @@ var BLOCKRUN_MODELS = [
     agentic: true
   },
   {
-    id: "anthropic/claude-opus-4.5",
+    id: "anthropic/claude-opus-4-5",
     name: "Claude Opus 4.5",
     inputPrice: 5,
     outputPrice: 25,
@@ -1705,7 +1710,7 @@ var BLOCKRUN_MODELS = [
     agentic: true
   },
   {
-    id: "anthropic/claude-opus-4.6",
+    id: "anthropic/claude-opus-4-6",
     name: "Claude Opus 4.6",
     inputPrice: 5,
     outputPrice: 25,
@@ -3785,6 +3790,83 @@ var PROVIDER_ERROR_PATTERNS = [
   /request.*size.*exceeds/i,
   /payload too large/i
 ];
+var DEGRADED_RESPONSE_PATTERNS = [
+  /the ai service is temporarily overloaded/i,
+  /service is temporarily overloaded/i,
+  /please try again in a moment/i
+];
+var DEGRADED_LOOP_PATTERNS = [
+  /the boxed is the response\./i,
+  /the response is the text\./i,
+  /the final answer is the boxed\./i
+];
+function extractAssistantContent(payload) {
+  if (!payload || typeof payload !== "object") return void 0;
+  const record = payload;
+  const choices = record.choices;
+  if (!Array.isArray(choices) || choices.length === 0) return void 0;
+  const firstChoice = choices[0];
+  if (!firstChoice || typeof firstChoice !== "object") return void 0;
+  const choice = firstChoice;
+  const message = choice.message;
+  if (!message || typeof message !== "object") return void 0;
+  const content = message.content;
+  return typeof content === "string" ? content : void 0;
+}
+function hasKnownLoopSignature(text) {
+  const matchCount = DEGRADED_LOOP_PATTERNS.reduce(
+    (count, pattern) => pattern.test(text) ? count + 1 : count,
+    0
+  );
+  if (matchCount >= 2) return true;
+  const lines = text.split(/\r?\n/).map((line) => line.trim()).filter(Boolean);
+  if (lines.length < 8) return false;
+  const counts = /* @__PURE__ */ new Map();
+  for (const line of lines) {
+    counts.set(line, (counts.get(line) ?? 0) + 1);
+  }
+  const maxRepeat = Math.max(...counts.values());
+  const uniqueRatio = counts.size / lines.length;
+  return maxRepeat >= 3 && uniqueRatio <= 0.45;
+}
+function detectDegradedSuccessResponse(body) {
+  const trimmed = body.trim();
+  if (!trimmed) return void 0;
+  if (DEGRADED_RESPONSE_PATTERNS.some((pattern) => pattern.test(trimmed))) {
+    return "degraded response: overloaded placeholder";
+  }
+  if (hasKnownLoopSignature(trimmed)) {
+    return "degraded response: repetitive loop output";
+  }
+  try {
+    const parsed = JSON.parse(trimmed);
+    const errorField = parsed.error;
+    let errorText = "";
+    if (typeof errorField === "string") {
+      errorText = errorField;
+    } else if (errorField && typeof errorField === "object") {
+      const errObj = errorField;
+      errorText = [
+        typeof errObj.message === "string" ? errObj.message : "",
+        typeof errObj.type === "string" ? errObj.type : "",
+        typeof errObj.code === "string" ? errObj.code : ""
+      ].filter(Boolean).join(" ");
+    }
+    if (errorText && PROVIDER_ERROR_PATTERNS.some((pattern) => pattern.test(errorText))) {
+      return `degraded response: ${errorText.slice(0, 120)}`;
+    }
+    const assistantContent = extractAssistantContent(parsed);
+    if (!assistantContent) return void 0;
+    if (DEGRADED_RESPONSE_PATTERNS.some((pattern) => pattern.test(assistantContent))) {
+      return "degraded response: overloaded assistant content";
+    }
+    if (hasKnownLoopSignature(assistantContent)) {
+      return "degraded response: repetitive assistant loop";
+    }
+  } catch {
+  }
+  return void 0;
+}
 var FALLBACK_STATUS_CODES = [
   400,
   // Bad request - sometimes used for billing errors
@@ -4324,6 +4406,22 @@ async function tryModelRequest(upstreamUrl, method, headers, body, modelId, maxT
         isProviderError: isProviderErr
       };
     }
+    const contentType = response.headers.get("content-type") || "";
+    if (contentType.includes("json") || contentType.includes("text")) {
+      try {
+        const responseBody = await response.clone().text();
+        const degradedReason = detectDegradedSuccessResponse(responseBody);
+        if (degradedReason) {
+          return {
+            success: false,
+            errorBody: degradedReason,
+            errorStatus: 503,
+            isProviderError: true
+          };
+        }
+      } catch {
+      }
+    }
     return { success: true, response };
   } catch (err) {
     const errorMsg = err instanceof Error ? err.message : String(err);