@blockrun/clawrouter 0.9.5 → 0.9.7

package/dist/index.js

@@ -1571,8 +1571,8 @@ var DEFAULT_ROUTING_CONFIG = {
   // Auto (balanced) tier configs - current default smart routing
   tiers: {
     SIMPLE: {
-      primary: "nvidia/kimi-k2.5",
-      // $0.55/$2.5 - best quality/price for simple tasks
+      primary: "moonshot/kimi-k2.5",
+      // $0.50/$2.40 - best quality/price for simple tasks
       fallback: [
         "google/gemini-2.5-flash",
         // 1M context, cost-effective
@@ -1624,14 +1624,14 @@ var DEFAULT_ROUTING_CONFIG = {
   // Eco tier configs - ultra cost-optimized (blockrun/eco)
   ecoTiers: {
     SIMPLE: {
-      primary: "nvidia/kimi-k2.5",
-      // $0.55/$2.5
+      primary: "moonshot/kimi-k2.5",
+      // $0.50/$2.40
       fallback: ["nvidia/gpt-oss-120b", "deepseek/deepseek-chat", "google/gemini-2.5-flash"]
     },
     MEDIUM: {
       primary: "deepseek/deepseek-chat",
       // $0.14/$0.28
-      fallback: ["xai/grok-code-fast-1", "google/gemini-2.5-flash", "nvidia/kimi-k2.5"]
+      fallback: ["xai/grok-code-fast-1", "google/gemini-2.5-flash", "moonshot/kimi-k2.5"]
     },
     COMPLEX: {
       primary: "xai/grok-4-0709",
@@ -3249,6 +3249,7 @@ var ROUTING_PROFILES = /* @__PURE__ */ new Set([
   "premium"
 ]);
 var FREE_MODEL = "nvidia/gpt-oss-120b";
+var MAX_MESSAGES = 200;
 var HEARTBEAT_INTERVAL_MS = 2e3;
 var DEFAULT_REQUEST_TIMEOUT_MS = 18e4;
 var MAX_FALLBACK_ATTEMPTS = 5;
@@ -3557,6 +3558,17 @@ function normalizeMessagesForThinking(messages) {
   });
   return hasChanges ? normalized : messages;
 }
+function truncateMessages(messages) {
+  if (!messages || messages.length <= MAX_MESSAGES) return messages;
+  const systemMsgs = messages.filter((m) => m.role === "system");
+  const conversationMsgs = messages.filter((m) => m.role !== "system");
+  const maxConversation = MAX_MESSAGES - systemMsgs.length;
+  const truncatedConversation = conversationMsgs.slice(-maxConversation);
+  console.log(
+    `[ClawRouter] Truncated messages: ${messages.length} \u2192 ${systemMsgs.length + truncatedConversation.length} (kept ${systemMsgs.length} system + ${truncatedConversation.length} recent)`
+  );
+  return [...systemMsgs, ...truncatedConversation];
+}
 var KIMI_BLOCK_RE = /<[｜|][^<>]*begin[^<>]*[｜|]>[\s\S]*?<[｜|][^<>]*end[^<>]*[｜|]>/gi;
 var KIMI_TOKEN_RE = /<[｜|][^<>]*[｜|]>/g;
 var THINKING_TAG_RE = /<\s*\/?\s*(?:think(?:ing)?|thought|antthinking)\b[^>]*>/gi;
@@ -3866,6 +3878,9 @@ async function tryModelRequest(upstreamUrl, method, headers, body, modelId, maxT
     if (Array.isArray(parsed.messages)) {
       parsed.messages = normalizeMessageRoles(parsed.messages);
     }
+    if (Array.isArray(parsed.messages)) {
+      parsed.messages = truncateMessages(parsed.messages);
+    }
     if (Array.isArray(parsed.messages)) {
       parsed.messages = sanitizeToolIds(parsed.messages);
     }