fix: handle <think> reasoning tags in parseExtractorOutput (#2559)

Reasoning models (MiniMax-M3, DeepSeek-R1, etc.) return <think>...</think>
tags in the content field before the actual JSON output. This caused
parseExtractorOutput to fail in two ways:

1. The fence regex /^\`\`\`(json)?...$/ requires the fence at text start;
   <think> preceding it prevents matching, so the raw text (with trailing
   fences) hits JSON.parse and throws.

2. When think tags contain [ or { characters, indexOf finds them inside
   the reasoning block instead of the actual JSON array.

Changes:
- Strip <think>...</think> tags before any parsing (covers all reasoning models)
- Add JSON.parse fallback: truncate at last ] or } to handle trailing
  noise (leftover markdown fences after stripping)

Tests: 28/28 pass (3 new cases for think tags + trailing noise).

Co-authored-by: qaz8545355 <junjun@openclaw.local>

src/core/cycle/propose-takes.ts

@@ -299,6 +299,8 @@ export async function defaultExtractor(
 export function parseExtractorOutput(raw: string): ProposedTake[] {
   if (!raw || raw.trim().length === 0) return [];
   let text = raw.trim();
+  // Strip <think>...</think> reasoning tags (MiniMax-M3, DeepSeek-R1, etc.).
+  text = text.replace(/<think>[\s\S]*?<\/think>/g, '').trim();
   // Strip markdown code fence wrapper.
   const fenced = text.match(/^```(?:json)?\s*\n?([\s\S]*?)\n?```$/);
   if (fenced) text = (fenced[1] ?? '').trim();
@@ -311,7 +313,21 @@ export function parseExtractorOutput(raw: string): ProposedTake[] {
   try {
     parsed = JSON.parse(text.slice(start));
   } catch {
-    return [];
+    // Fallback: truncate at last ] or } to handle trailing noise (e.g. leftover
+    // markdown fences after <think> stripping). Try array-closing first.
+    const sliced = text.slice(start);
+    const lastArr = sliced.lastIndexOf(']');
+    const lastObj = sliced.lastIndexOf('}');
+    const end = Math.max(lastArr, lastObj);
+    if (end > 0) {
+      try {
+        parsed = JSON.parse(sliced.slice(0, end + 1));
+      } catch {
+        return [];
+      }
+    } else {
+      return [];
+    }
   }
   const arr = Array.isArray(parsed) ? parsed : [parsed];
   const out: ProposedTake[] = [];

test/propose-takes.test.ts

@@ -169,6 +169,26 @@ describe('parseExtractorOutput', () => {
     const out = parseExtractorOutput(raw);
     expect(out[0]!.domain).toBe('macro');
   });
+
+  test('strips <think> reasoning tags before parsing (MiniMax-M3, DeepSeek-R1)', () => {
+    const raw = '<think>Analyzing the prose... I see several claims.</think>\n\n```json\n[{"claim_text":"X","kind":"take","holder":"brain","weight":0.5}]\n```';
+    const out = parseExtractorOutput(raw);
+    expect(out).toHaveLength(1);
+    expect(out[0]!.claim_text).toBe('X');
+  });
+
+  test('strips multiple <think> blocks', () => {
+    const raw = '<think>First thought.</think>\n<tool_call>...</tool_call>\n<think>Second thought.</think>\n\n[{"claim_text":"Y","kind":"bet","holder":"brain","weight":0.7}]';
+    const out = parseExtractorOutput(raw);
+    expect(out).toHaveLength(1);
+  });
+
+  test('handles trailing noise after JSON (leftover fences)', () => {
+    const raw = '<think>done</think>\n```json\n[{"claim_text":"Z","kind":"take","holder":"brain","weight":0.6}]\n```\n';
+    const out = parseExtractorOutput(raw);
+    expect(out).toHaveLength(1);
+    expect(out[0]!.claim_text).toBe('Z');
+  });
 });
 
 // ─── contentHash ────────────────────────────────────────────────────