@blockrun/clawrouter 0.8.6 → 0.8.8

package/dist/index.js

@@ -931,6 +931,18 @@ function getFallbackChain(tier, tierConfigs) {
   const config = tierConfigs[tier];
   return [config.primary, ...config.fallback];
 }
+function calculateModelCost(model, modelPricing, estimatedInputTokens, maxOutputTokens) {
+  const pricing = modelPricing.get(model);
+  const inputCost = pricing ? estimatedInputTokens / 1e6 * pricing.inputPrice : 0;
+  const outputCost = pricing ? maxOutputTokens / 1e6 * pricing.outputPrice : 0;
+  const costEstimate = inputCost + outputCost;
+  const opusPricing = modelPricing.get("anthropic/claude-opus-4");
+  const baselineInput = opusPricing ? estimatedInputTokens / 1e6 * opusPricing.inputPrice : 0;
+  const baselineOutput = opusPricing ? maxOutputTokens / 1e6 * opusPricing.outputPrice : 0;
+  const baselineCost = baselineInput + baselineOutput;
+  const savings = baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
+  return { costEstimate, baselineCost, savings };
+}
 function getFallbackChainFiltered(tier, tierConfigs, estimatedTotalTokens, getContextWindow) {
   const fullChain = getFallbackChain(tier, tierConfigs);
   const filtered = fullChain.filter((modelId) => {
@@ -1593,8 +1605,9 @@ var DEFAULT_ROUTING_CONFIG = {
   // Agentic tier configs - models that excel at multi-step autonomous tasks
   agenticTiers: {
     SIMPLE: {
-      primary: "anthropic/claude-haiku-4.5",
-      fallback: ["moonshot/kimi-k2.5", "xai/grok-4-fast-non-reasoning", "openai/gpt-4o-mini"]
+      primary: "moonshot/kimi-k2.5",
+      // Cheaper than Haiku ($0.5/$2.4 vs $1/$5), larger context
+      fallback: ["anthropic/claude-haiku-4.5", "xai/grok-4-fast-non-reasoning", "openai/gpt-4o-mini"]
     },
     MEDIUM: {
       primary: "xai/grok-code-fast-1",
@@ -3176,10 +3189,20 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       heartbeatInterval = void 0;
     }
     if (routingDecision && actualModelUsed !== routingDecision.model) {
+      const estimatedInputTokens = Math.ceil(body.length / 4);
+      const newCosts = calculateModelCost(
+        actualModelUsed,
+        routerOpts.modelPricing,
+        estimatedInputTokens,
+        maxTokens
+      );
       routingDecision = {
         ...routingDecision,
         model: actualModelUsed,
-        reasoning: `${routingDecision.reasoning} | fallback to ${actualModelUsed}`
+        reasoning: `${routingDecision.reasoning} | fallback to ${actualModelUsed}`,
+        costEstimate: newCosts.costEstimate,
+        baselineCost: newCosts.baselineCost,
+        savings: newCosts.savings
       };
       options.onRouted?.(routingDecision);
     }
@@ -3371,13 +3394,22 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
     throw err;
   }
   if (routingDecision) {
+    const estimatedInputTokens = Math.ceil(body.length / 4);
+    const accurateCosts = calculateModelCost(
+      routingDecision.model,
+      routerOpts.modelPricing,
+      estimatedInputTokens,
+      maxTokens
+    );
+    const costWithBuffer = accurateCosts.costEstimate * 1.2;
+    const baselineWithBuffer = accurateCosts.baselineCost * 1.2;
     const entry = {
       timestamp: (/* @__PURE__ */ new Date()).toISOString(),
       model: routingDecision.model,
       tier: routingDecision.tier,
-      cost: routingDecision.costEstimate,
-      baselineCost: routingDecision.baselineCost,
-      savings: routingDecision.savings,
+      cost: costWithBuffer,
+      baselineCost: baselineWithBuffer,
+      savings: accurateCosts.savings,
       latencyMs: Date.now() - startTime
     };
     logUsage(entry).catch(() => {