@blockrun/clawrouter 0.9.4 → 0.9.6

package/dist/index.d.ts

@@ -581,7 +581,12 @@ declare const blockrunProvider: ProviderPlugin;
 declare const MODEL_ALIASES: Record<string, string>;
 /**
  * Resolve a model alias to its full model ID.
- * Returns the original model if not an alias.
+ * Also strips "blockrun/" prefix for direct model paths.
+ * Examples:
+ *   - "claude" -> "anthropic/claude-sonnet-4" (alias)
+ *   - "blockrun/claude" -> "anthropic/claude-sonnet-4" (alias with prefix)
+ *   - "blockrun/anthropic/claude-sonnet-4" -> "anthropic/claude-sonnet-4" (prefix stripped)
+ *   - "openai/gpt-4o" -> "openai/gpt-4o" (unchanged)
  */
 declare function resolveModelAlias(model: string): string;
 type BlockRunModel = {

package/dist/index.js

@@ -37,6 +37,7 @@ function resolveModelAlias(model) {
     const withoutPrefix = normalized.slice("blockrun/".length);
     const resolvedWithoutPrefix = MODEL_ALIASES[withoutPrefix];
     if (resolvedWithoutPrefix) return resolvedWithoutPrefix;
+    return withoutPrefix;
   }
   return model;
 }
@@ -2111,9 +2112,11 @@ var RequestDeduplicator = class {
   removeInflight(key) {
     const entry = this.inflight.get(key);
     if (entry) {
-      const errorBody = Buffer.from(JSON.stringify({
-        error: { message: "Original request failed, please retry", type: "dedup_origin_failed" }
-      }));
+      const errorBody = Buffer.from(
+        JSON.stringify({
+          error: { message: "Original request failed, please retry", type: "dedup_origin_failed" }
+        })
+      );
       for (const resolve of entry.resolvers) {
         resolve({
           status: 503,
@@ -3246,6 +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;
@@ -3554,6 +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;
@@ -3863,6 +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);
     }