@blockrun/clawrouter 0.8.20 → 0.8.21

package/dist/index.d.ts

@@ -203,6 +203,10 @@ type RoutingConfig = {
     tiers: Record<Tier, TierConfig>;
     /** Tier configs for agentic mode - models that excel at multi-step tasks */
     agenticTiers?: Record<Tier, TierConfig>;
+    /** Tier configs for eco profile - ultra cost-optimized (blockrun/eco) */
+    ecoTiers?: Record<Tier, TierConfig>;
+    /** Tier configs for premium profile - best quality (blockrun/premium) */
+    premiumTiers?: Record<Tier, TierConfig>;
     overrides: OverridesConfig;
 };
 
@@ -225,7 +229,7 @@ declare function getFallbackChain(tier: Tier, tierConfigs: Record<Tier, TierConf
  * Calculate cost for a specific model (used when fallback model is used).
  * Returns updated cost fields for RoutingDecision.
  */
-declare function calculateModelCost(model: string, modelPricing: Map<string, ModelPricing>, estimatedInputTokens: number, maxOutputTokens: number): {
+declare function calculateModelCost(model: string, modelPricing: Map<string, ModelPricing>, estimatedInputTokens: number, maxOutputTokens: number, routingProfile?: "free" | "eco" | "auto" | "premium"): {
     costEstimate: number;
     baselineCost: number;
     savings: number;
@@ -264,6 +268,7 @@ declare const DEFAULT_ROUTING_CONFIG: RoutingConfig;
 type RouterOptions = {
     config: RoutingConfig;
     modelPricing: Map<string, ModelPricing>;
+    routingProfile?: "free" | "eco" | "auto" | "premium";
 };
 /**
  * Route a request to the cheapest capable model.

package/dist/index.js

@@ -41,16 +41,40 @@ function resolveModelAlias(model) {
   return model;
 }
 var BLOCKRUN_MODELS = [
-  // Smart routing meta-model — proxy replaces with actual model
+  // Smart routing meta-models — proxy replaces with actual model
   // NOTE: Model IDs are WITHOUT provider prefix (OpenClaw adds "blockrun/" automatically)
   {
     id: "auto",
-    name: "BlockRun Smart Router",
+    name: "Auto (Smart Router - Balanced)",
     inputPrice: 0,
     outputPrice: 0,
     contextWindow: 105e4,
     maxOutput: 128e3
   },
+  {
+    id: "free",
+    name: "Free (NVIDIA GPT-OSS-120B only)",
+    inputPrice: 0,
+    outputPrice: 0,
+    contextWindow: 128e3,
+    maxOutput: 4096
+  },
+  {
+    id: "eco",
+    name: "Eco (Smart Router - Cost Optimized)",
+    inputPrice: 0,
+    outputPrice: 0,
+    contextWindow: 105e4,
+    maxOutput: 128e3
+  },
+  {
+    id: "premium",
+    name: "Premium (Smart Router - Best Quality)",
+    inputPrice: 0,
+    outputPrice: 0,
+    contextWindow: 2e6,
+    maxOutput: 2e5
+  },
   // OpenAI GPT-5 Family
   {
     id: "openai/gpt-5.2",
@@ -352,8 +376,8 @@ var BLOCKRUN_MODELS = [
   {
     id: "xai/grok-4-0709",
     name: "Grok 4 (0709)",
-    inputPrice: 3,
-    outputPrice: 15,
+    inputPrice: 0.2,
+    outputPrice: 1.5,
     contextWindow: 131072,
     maxOutput: 16384,
     reasoning: true
@@ -909,7 +933,7 @@ function calibrateConfidence(distance, steepness) {
 }
 
 // src/router/selector.ts
-function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPricing, estimatedInputTokens, maxOutputTokens) {
+function selectModel(tier, confidence, method, reasoning, tierConfigs, modelPricing, estimatedInputTokens, maxOutputTokens, routingProfile) {
   const tierConfig = tierConfigs[tier];
   const model = tierConfig.primary;
   const pricing = modelPricing.get(model);
@@ -918,13 +942,13 @@ 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");
+  const opusPricing = modelPricing.get("anthropic/claude-opus-4.5");
   const opusInputPrice = opusPricing?.inputPrice ?? 0;
   const opusOutputPrice = opusPricing?.outputPrice ?? 0;
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
   const baselineOutput = maxOutputTokens / 1e6 * opusOutputPrice;
   const baselineCost = baselineInput + baselineOutput;
-  const savings = baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
+  const savings = routingProfile === "premium" ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
   return {
     model,
     tier,
@@ -940,20 +964,20 @@ function getFallbackChain(tier, tierConfigs) {
   const config = tierConfigs[tier];
   return [config.primary, ...config.fallback];
 }
-function calculateModelCost(model, modelPricing, estimatedInputTokens, maxOutputTokens) {
+function calculateModelCost(model, modelPricing, estimatedInputTokens, maxOutputTokens, routingProfile) {
   const pricing = modelPricing.get(model);
   const inputPrice = pricing?.inputPrice ?? 0;
   const outputPrice = pricing?.outputPrice ?? 0;
   const inputCost = estimatedInputTokens / 1e6 * inputPrice;
   const outputCost = maxOutputTokens / 1e6 * outputPrice;
   const costEstimate = inputCost + outputCost;
-  const opusPricing = modelPricing.get("anthropic/claude-opus-4");
+  const opusPricing = modelPricing.get("anthropic/claude-opus-4.5");
   const opusInputPrice = opusPricing?.inputPrice ?? 0;
   const opusOutputPrice = opusPricing?.outputPrice ?? 0;
   const baselineInput = estimatedInputTokens / 1e6 * opusInputPrice;
   const baselineOutput = maxOutputTokens / 1e6 * opusOutputPrice;
   const baselineCost = baselineInput + baselineOutput;
-  const savings = baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
+  const savings = routingProfile === "premium" ? 0 : baselineCost > 0 ? Math.max(0, (baselineCost - costEstimate) / baselineCost) : 0;
   return { costEstimate, baselineCost, savings };
 }
 function getFallbackChainFiltered(tier, tierConfigs, estimatedTotalTokens, getContextWindow) {
@@ -1582,15 +1606,17 @@ var DEFAULT_ROUTING_CONFIG = {
     // Tier boundaries on weighted score axis
     tierBoundaries: {
       simpleMedium: 0,
-      mediumComplex: 0.18,
-      complexReasoning: 0.4
-      // Raised from 0.25 - requires strong reasoning signals
+      mediumComplex: 0.3,
+      // Raised from 0.18 - prevent simple tasks from reaching expensive COMPLEX tier
+      complexReasoning: 0.5
+      // Raised from 0.4 - reserve for true reasoning tasks
     },
     // Sigmoid steepness for confidence calibration
     confidenceSteepness: 12,
     // Below this confidence → ambiguous (null tier)
     confidenceThreshold: 0.7
   },
+  // Auto (balanced) tier configs - current default smart routing
   tiers: {
     SIMPLE: {
       primary: "nvidia/kimi-k2.5",
@@ -1599,7 +1625,9 @@ var DEFAULT_ROUTING_CONFIG = {
         "google/gemini-2.5-flash",
         "nvidia/gpt-oss-120b",
         "nvidia/gpt-oss-20b",
-        "deepseek/deepseek-chat"
+        "deepseek/deepseek-chat",
+        "xai/grok-code-fast-1"
+        // Added for better quality fallback
       ]
     },
     MEDIUM: {
@@ -1614,7 +1642,8 @@ var DEFAULT_ROUTING_CONFIG = {
     },
     COMPLEX: {
       primary: "google/gemini-2.5-pro",
-      fallback: ["openai/gpt-5.2", "anthropic/claude-sonnet-4", "xai/grok-4-0709", "openai/gpt-4o"]
+      fallback: ["xai/grok-4-0709", "openai/gpt-4o", "openai/gpt-5.2", "anthropic/claude-sonnet-4"]
+      // Grok first for cost efficiency, Sonnet as last resort
     },
     REASONING: {
       primary: "xai/grok-4-1-fast-reasoning",
@@ -1628,6 +1657,52 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     }
   },
+  // Eco tier configs - ultra cost-optimized (blockrun/eco)
+  ecoTiers: {
+    SIMPLE: {
+      primary: "nvidia/kimi-k2.5",
+      // $0.001/$0.001
+      fallback: ["deepseek/deepseek-chat", "nvidia/gpt-oss-120b", "nvidia/gpt-oss-20b"]
+    },
+    MEDIUM: {
+      primary: "deepseek/deepseek-chat",
+      // $0.14/$0.28
+      fallback: ["xai/grok-code-fast-1", "google/gemini-2.5-flash", "nvidia/kimi-k2.5"]
+    },
+    COMPLEX: {
+      primary: "xai/grok-4-0709",
+      // $0.20/$1.50
+      fallback: ["deepseek/deepseek-chat", "google/gemini-2.5-flash", "openai/gpt-4o-mini"]
+    },
+    REASONING: {
+      primary: "deepseek/deepseek-reasoner",
+      // $0.55/$2.19
+      fallback: ["xai/grok-4-fast-reasoning", "moonshot/kimi-k2.5"]
+    }
+  },
+  // Premium tier configs - best quality (blockrun/premium)
+  premiumTiers: {
+    SIMPLE: {
+      primary: "google/gemini-2.5-flash",
+      // $0.075/$0.30
+      fallback: ["openai/gpt-4o-mini", "anthropic/claude-haiku-4.5", "moonshot/kimi-k2.5"]
+    },
+    MEDIUM: {
+      primary: "openai/gpt-4o",
+      // $2.50/$10
+      fallback: ["google/gemini-2.5-pro", "anthropic/claude-sonnet-4", "xai/grok-4-0709"]
+    },
+    COMPLEX: {
+      primary: "anthropic/claude-opus-4.5",
+      // $15/$75
+      fallback: ["openai/gpt-5.2", "anthropic/claude-sonnet-4", "google/gemini-2.5-pro"]
+    },
+    REASONING: {
+      primary: "openai/o3",
+      // $10/$40
+      fallback: ["anthropic/claude-opus-4.5", "openai/o1", "google/gemini-2.5-pro"]
+    }
+  },
   // Agentic tier configs - models that excel at multi-step autonomous tasks
   agenticTiers: {
     SIMPLE: {
@@ -1669,21 +1744,34 @@ function route(prompt, systemPrompt, maxOutputTokens, options) {
   const fullText = `${systemPrompt ?? ""} ${prompt}`;
   const estimatedTokens = Math.ceil(fullText.length / 4);
   const ruleResult = classifyByRules(prompt, systemPrompt, estimatedTokens, config.scoring);
-  const agenticScore = ruleResult.agenticScore ?? 0;
-  const isAutoAgentic = agenticScore >= 0.69;
-  const isExplicitAgentic = config.overrides.agenticMode ?? false;
-  const useAgenticTiers = (isAutoAgentic || isExplicitAgentic) && config.agenticTiers != null;
-  const tierConfigs = useAgenticTiers ? config.agenticTiers : config.tiers;
+  const { routingProfile } = options;
+  let tierConfigs;
+  let profileSuffix = "";
+  if (routingProfile === "eco" && config.ecoTiers) {
+    tierConfigs = config.ecoTiers;
+    profileSuffix = " | eco";
+  } else if (routingProfile === "premium" && config.premiumTiers) {
+    tierConfigs = config.premiumTiers;
+    profileSuffix = " | premium";
+  } else {
+    const agenticScore = ruleResult.agenticScore ?? 0;
+    const isAutoAgentic = agenticScore >= 0.5;
+    const isExplicitAgentic = config.overrides.agenticMode ?? false;
+    const useAgenticTiers = (isAutoAgentic || isExplicitAgentic) && config.agenticTiers != null;
+    tierConfigs = useAgenticTiers ? config.agenticTiers : config.tiers;
+    profileSuffix = useAgenticTiers ? " | agentic" : "";
+  }
   if (estimatedTokens > config.overrides.maxTokensForceComplex) {
     return selectModel(
       "COMPLEX",
       0.95,
       "rules",
-      `Input exceeds ${config.overrides.maxTokensForceComplex} tokens${useAgenticTiers ? " | agentic" : ""}`,
+      `Input exceeds ${config.overrides.maxTokensForceComplex} tokens${profileSuffix}`,
       tierConfigs,
       modelPricing,
       estimatedTokens,
-      maxOutputTokens
+      maxOutputTokens,
+      routingProfile
     );
   }
   const hasStructuredOutput = systemPrompt ? /json|structured|schema/i.test(systemPrompt) : false;
@@ -1707,11 +1795,7 @@ function route(prompt, systemPrompt, maxOutputTokens, options) {
       tier = minTier;
     }
   }
-  if (isAutoAgentic) {
-    reasoning += " | auto-agentic";
-  } else if (isExplicitAgentic) {
-    reasoning += " | agentic";
-  }
+  reasoning += profileSuffix;
   return selectModel(
     tier,
     confidence,
@@ -1720,7 +1804,8 @@ function route(prompt, systemPrompt, maxOutputTokens, options) {
     tierConfigs,
     modelPricing,
     estimatedTokens,
-    maxOutputTokens
+    maxOutputTokens,
+    routingProfile
   );
 }
 
@@ -1878,12 +1963,13 @@ async function getStats(days = 7) {
 function formatStatsAscii(stats) {
   const lines = [];
   lines.push("\u2554\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2557");
-  lines.push("\u2551              ClawRouter Usage Statistics                   \u2551");
+  lines.push("\u2551          ClawRouter by BlockRun v0.8.20                    \u2551");
+  lines.push("\u2551                Usage Statistics                            \u2551");
   lines.push("\u2560\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2563");
   lines.push(`\u2551  Period: ${stats.period.padEnd(49)}\u2551`);
   lines.push(`\u2551  Total Requests: ${stats.totalRequests.toString().padEnd(41)}\u2551`);
   lines.push(`\u2551  Total Cost: $${stats.totalCost.toFixed(4).padEnd(43)}\u2551`);
-  lines.push(`\u2551  Baseline Cost (Opus): $${stats.totalBaselineCost.toFixed(4).padEnd(33)}\u2551`);
+  lines.push(`\u2551  Baseline Cost (Opus 4.5): $${stats.totalBaselineCost.toFixed(4).padEnd(30)}\u2551`);
   const savingsLine = `\u2551  \u{1F4B0} Total Saved: $${stats.totalSavings.toFixed(4)} (${stats.savingsPercentage.toFixed(1)}%)`;
   if (stats.entriesWithBaseline < stats.totalRequests && stats.entriesWithBaseline > 0) {
     lines.push(savingsLine.padEnd(61) + "\u2551");
@@ -2411,7 +2497,16 @@ async function checkForUpdates() {
 // src/proxy.ts
 var BLOCKRUN_API = "https://blockrun.ai/api";
 var AUTO_MODEL = "blockrun/auto";
-var AUTO_MODEL_SHORT = "auto";
+var ROUTING_PROFILES = /* @__PURE__ */ new Set([
+  "blockrun/free",
+  "free",
+  "blockrun/eco",
+  "eco",
+  "blockrun/auto",
+  "auto",
+  "blockrun/premium",
+  "premium"
+]);
 var FREE_MODEL = "nvidia/gpt-oss-120b";
 var HEARTBEAT_INTERVAL_MS = 2e3;
 var DEFAULT_REQUEST_TIMEOUT_MS = 18e4;
@@ -3093,6 +3188,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   let isStreaming = false;
   let modelId = "";
   let maxTokens = 4096;
+  let routingProfile = null;
   const isChatCompletion = req.url?.includes("/chat/completions");
   if (isChatCompletion && body.length > 0) {
     try {
@@ -3108,58 +3204,83 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       const normalizedModel = typeof parsed.model === "string" ? parsed.model.trim().toLowerCase() : "";
       const resolvedModel = resolveModelAlias(normalizedModel);
       const wasAlias = resolvedModel !== normalizedModel;
-      const isAutoModel = normalizedModel === AUTO_MODEL.toLowerCase() || normalizedModel === AUTO_MODEL_SHORT.toLowerCase();
+      const isRoutingProfile = ROUTING_PROFILES.has(normalizedModel);
+      if (isRoutingProfile) {
+        const profileName = normalizedModel.replace("blockrun/", "");
+        routingProfile = profileName;
+      }
       console.log(
-        `[ClawRouter] Received model: "${parsed.model}" -> normalized: "${normalizedModel}"${wasAlias ? ` -> alias: "${resolvedModel}"` : ""}, isAuto: ${isAutoModel}`
+        `[ClawRouter] Received model: "${parsed.model}" -> normalized: "${normalizedModel}"${wasAlias ? ` -> alias: "${resolvedModel}"` : ""}${routingProfile ? `, profile: ${routingProfile}` : ""}`
       );
-      if (wasAlias && !isAutoModel) {
+      if (wasAlias && !isRoutingProfile) {
         parsed.model = resolvedModel;
         modelId = resolvedModel;
         bodyModified = true;
       }
-      if (isAutoModel) {
-        const sessionId = getSessionId(
-          req.headers
-        );
-        const existingSession = sessionId ? sessionStore.getSession(sessionId) : void 0;
-        if (existingSession) {
-          console.log(
-            `[ClawRouter] Session ${sessionId?.slice(0, 8)}... using pinned model: ${existingSession.model}`
-          );
-          parsed.model = existingSession.model;
-          modelId = existingSession.model;
+      if (isRoutingProfile) {
+        if (routingProfile === "free") {
+          const freeModel = "nvidia/gpt-oss-120b";
+          console.log(`[ClawRouter] Free profile - using ${freeModel} directly`);
+          parsed.model = freeModel;
+          modelId = freeModel;
           bodyModified = true;
-          sessionStore.touchSession(sessionId);
+          await logUsage({
+            timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+            model: freeModel,
+            tier: "SIMPLE",
+            cost: 0,
+            baselineCost: 0,
+            savings: 1,
+            // 100% savings
+            latencyMs: 0
+          });
         } else {
-          const messages = parsed.messages;
-          let lastUserMsg;
-          if (messages) {
-            for (let i = messages.length - 1; i >= 0; i--) {
-              if (messages[i].role === "user") {
-                lastUserMsg = messages[i];
-                break;
-              }
-            }
-          }
-          const systemMsg = messages?.find((m) => m.role === "system");
-          const prompt = typeof lastUserMsg?.content === "string" ? lastUserMsg.content : "";
-          const systemPrompt = typeof systemMsg?.content === "string" ? systemMsg.content : void 0;
-          const tools = parsed.tools;
-          const hasTools = Array.isArray(tools) && tools.length > 0;
-          if (hasTools) {
-            console.log(`[ClawRouter] Tools detected (${tools.length}), agentic mode via keywords`);
-          }
-          routingDecision = route(prompt, systemPrompt, maxTokens, routerOpts);
-          parsed.model = routingDecision.model;
-          modelId = routingDecision.model;
-          bodyModified = true;
-          if (sessionId) {
-            sessionStore.setSession(sessionId, routingDecision.model, routingDecision.tier);
+          const sessionId = getSessionId(
+            req.headers
+          );
+          const existingSession = sessionId ? sessionStore.getSession(sessionId) : void 0;
+          if (existingSession) {
             console.log(
-              `[ClawRouter] Session ${sessionId.slice(0, 8)}... pinned to model: ${routingDecision.model}`
+              `[ClawRouter] Session ${sessionId?.slice(0, 8)}... using pinned model: ${existingSession.model}`
             );
+            parsed.model = existingSession.model;
+            modelId = existingSession.model;
+            bodyModified = true;
+            sessionStore.touchSession(sessionId);
+          } else {
+            const messages = parsed.messages;
+            let lastUserMsg;
+            if (messages) {
+              for (let i = messages.length - 1; i >= 0; i--) {
+                if (messages[i].role === "user") {
+                  lastUserMsg = messages[i];
+                  break;
+                }
+              }
+            }
+            const systemMsg = messages?.find((m) => m.role === "system");
+            const prompt = typeof lastUserMsg?.content === "string" ? lastUserMsg.content : "";
+            const systemPrompt = typeof systemMsg?.content === "string" ? systemMsg.content : void 0;
+            const tools = parsed.tools;
+            const hasTools = Array.isArray(tools) && tools.length > 0;
+            if (hasTools) {
+              console.log(`[ClawRouter] Tools detected (${tools.length}), agentic mode via keywords`);
+            }
+            routingDecision = route(prompt, systemPrompt, maxTokens, {
+              ...routerOpts,
+              routingProfile: routingProfile ?? void 0
+            });
+            parsed.model = routingDecision.model;
+            modelId = routingDecision.model;
+            bodyModified = true;
+            if (sessionId) {
+              sessionStore.setSession(sessionId, routingDecision.model, routingDecision.tier);
+              console.log(
+                `[ClawRouter] Session ${sessionId.slice(0, 8)}... pinned to model: ${routingDecision.model}`
+              );
+            }
+            options.onRouted?.(routingDecision);
           }
-          options.onRouted?.(routingDecision);
         }
       }
       if (bodyModified) {
@@ -3343,7 +3464,8 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
         actualModelUsed,
         routerOpts.modelPricing,
         estimatedInputTokens,
-        maxTokens
+        maxTokens,
+        routingProfile ?? void 0
       );
       routingDecision = {
         ...routingDecision,
@@ -3552,7 +3674,8 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       routingDecision.model,
       routerOpts.modelPricing,
       estimatedInputTokens,
-      maxTokens
+      maxTokens,
+      routingProfile ?? void 0
     );
     const costWithBuffer = accurateCosts.costEstimate * 1.2;
     const baselineWithBuffer = accurateCosts.baselineCost * 1.2;