cliskill 1.0.1 → 1.0.3

package/dist/bootstrap/cli.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   runCli
-} from "../chunk-CBGJMMA3.js";
+} from "../chunk-PBLJ6557.js";
 import "../chunk-AJENHWD3.js";
 export {
   runCli

package/dist/{chunk-CBGJMMA3.js → chunk-PBLJ6557.js}

@@ -388,6 +388,19 @@ var BaseAdapter = class {
     }
     return timeoutSignal;
   }
+  /**
+   * Build an AbortSignal for streaming requests with extended timeout.
+   * Streaming responses can take much longer than regular requests,
+   * so we use 3x the configured timeout to avoid premature aborts.
+   */
+  buildStreamingSignal(external) {
+    const streamTimeout = Math.max(this.config.timeout * 3, 6e5);
+    const timeoutSignal = AbortSignal.timeout(streamTimeout);
+    if (external) {
+      return AbortSignal.any([timeoutSignal, external]);
+    }
+    return timeoutSignal;
+  }
   /**
    * Fetch with automatic retry for rate-limit (429), server errors (5xx),
    * and network failures (TypeError: fetch failed).
@@ -461,7 +474,7 @@ var GenericCompatAdapter = class extends BaseAdapter {
       method: "POST",
       headers: this.buildHeaders(),
       body: JSON.stringify(body),
-      signal: this.buildSignal(request.signal)
+      signal: this.buildStreamingSignal(request.signal)
     }));
     if (!response.ok) {
       const text = await response.text().catch(() => "unknown error");
@@ -670,7 +683,7 @@ var GLMAdapter = class extends BaseAdapter {
       method: "POST",
       headers: this.buildHeaders(),
       body: JSON.stringify(body),
-      signal: this.buildSignal(request.signal)
+      signal: this.buildStreamingSignal(request.signal)
     }));
     if (!response.ok) {
       const text = await response.text().catch(() => "unknown error");
@@ -5594,40 +5607,36 @@ ${result.summary}` }]
         break;
       }
       if (parser.stopReason === "max_tokens") {
-        const truncatedCalls = extractToolCalls(parser.assistantContent);
-        const hasEmptyInput = truncatedCalls.some((tc) => Object.keys(tc.input).length === 0);
-        if (hasEmptyInput || truncatedCalls.length === 0) {
-          if (maxOutputTokensRecoveryCount === 0 && maxOutputTokensOverride === void 0) {
-            maxOutputTokensOverride = ESCALATED_MAX_TOKENS;
-            maxOutputTokensRecoveryCount++;
-            state.messages.pop();
-            yield { type: "assistant_text", text: " [escalating token limit...]" };
-            continue;
-          }
-          if (maxOutputTokensRecoveryCount < MAX_OUTPUT_TOKENS_RECOVERY_LIMIT) {
-            const textOnly = parser.assistantContent.filter((b) => b.type === "text");
-            state.messages[state.messages.length - 1] = {
-              role: "assistant",
-              content: textOnly.length > 0 ? textOnly : [{ type: "text", text: "[Response truncated \u2014 output token limit]" }]
-            };
-            state.messages.push({
-              role: "user",
-              content: [{
-                type: "text",
-                text: "Output token limit hit. Resume directly \u2014 no apology, no recap. Pick up mid-thought if that is where the cut happened. Break remaining work into smaller pieces."
-              }]
-            });
-            maxOutputTokensRecoveryCount++;
-            maxOutputTokensOverride = void 0;
-            yield { type: "assistant_text", text: " [truncated \u2014 requesting continuation...]" };
-            continue;
-          }
-          yield {
-            type: "error",
-            error: new Error("Output token limit reached after multiple recovery attempts. Consider splitting your request into smaller tasks.")
+        if (maxOutputTokensRecoveryCount === 0 && maxOutputTokensOverride === void 0) {
+          maxOutputTokensOverride = ESCALATED_MAX_TOKENS;
+          maxOutputTokensRecoveryCount++;
+          state.messages.pop();
+          yield { type: "assistant_text", text: " [escalating token limit...]" };
+          continue;
+        }
+        if (maxOutputTokensRecoveryCount < MAX_OUTPUT_TOKENS_RECOVERY_LIMIT) {
+          const textOnly = parser.assistantContent.filter((b) => b.type === "text");
+          state.messages[state.messages.length - 1] = {
+            role: "assistant",
+            content: textOnly.length > 0 ? textOnly : [{ type: "text", text: "[Response truncated \u2014 output token limit]" }]
           };
-          break;
+          state.messages.push({
+            role: "user",
+            content: [{
+              type: "text",
+              text: "Output token limit hit. Resume directly \u2014 no apology, no recap. Pick up mid-thought if that is where the cut happened. IMPORTANT: Break remaining work into smaller pieces \u2014 write files in multiple smaller writes instead of one huge write."
+            }]
+          });
+          maxOutputTokensRecoveryCount++;
+          maxOutputTokensOverride = void 0;
+          yield { type: "assistant_text", text: " [truncated \u2014 requesting continuation...]" };
+          continue;
         }
+        yield {
+          type: "error",
+          error: new Error("Output token limit reached after multiple recovery attempts. Consider splitting your request into smaller tasks.")
+        };
+        break;
       } else {
         maxOutputTokensRecoveryCount = 0;
         maxOutputTokensOverride = void 0;
@@ -10021,4 +10030,4 @@ export {
   MessageList,
   runCli
 };
-//# sourceMappingURL=chunk-CBGJMMA3.js.map
+//# sourceMappingURL=chunk-PBLJ6557.js.map