Merge cd25cbad83 into 4fade39c90

agent/trajectory.py

@@ -7,24 +7,39 @@ the file-write logic live here.
 
 import json
 import logging
+import re
 from datetime import datetime
 from typing import Any, Dict, List
 
 logger = logging.getLogger(__name__)
 
+_SCRATCHPAD_OPEN_TAG = "<REASONING_SCRATCHPAD>"
+_SCRATCHPAD_CLOSE_TAG = "</REASONING_SCRATCHPAD>"
+_FENCED_CODE_BLOCK_RE = re.compile(r"```.*?```", re.DOTALL)
+_INLINE_CODE_RE = re.compile(r"`[^`\n]*`")
+_BLOCKQUOTE_RE = re.compile(r"(?m)^[ \t]*>.*(?:\n|$)")
+
 
 def convert_scratchpad_to_think(content: str) -> str:
     """Convert <REASONING_SCRATCHPAD> tags to <think> tags."""
-    if not content or "<REASONING_SCRATCHPAD>" not in content:
+    if not content or _SCRATCHPAD_OPEN_TAG not in content:
         return content
-    return content.replace("<REASONING_SCRATCHPAD>", "<think>").replace("</REASONING_SCRATCHPAD>", "</think>")
+    return content.replace(_SCRATCHPAD_OPEN_TAG, "<think>").replace(_SCRATCHPAD_CLOSE_TAG, "</think>")
+
+
+def _strip_markdown_context_for_scratchpad_check(content: str) -> str:
+    """Remove markdown contexts where scratchpad tags should be treated as literal text."""
+    stripped = _FENCED_CODE_BLOCK_RE.sub("", content)
+    stripped = _INLINE_CODE_RE.sub("", stripped)
+    return _BLOCKQUOTE_RE.sub("", stripped)
 
 
 def has_incomplete_scratchpad(content: str) -> bool:
     """Check if content has an opening <REASONING_SCRATCHPAD> without a closing tag."""
     if not content:
         return False
-    return "<REASONING_SCRATCHPAD>" in content and "</REASONING_SCRATCHPAD>" not in content
+    visible_content = _strip_markdown_context_for_scratchpad_check(content)
+    return visible_content.count(_SCRATCHPAD_OPEN_TAG) > visible_content.count(_SCRATCHPAD_CLOSE_TAG)
 
 
 def save_trajectory(trajectory: List[Dict[str, Any]], model: str,

tests/agent/test_trajectory.py

@@ -0,0 +1,36 @@
+from agent.trajectory import convert_scratchpad_to_think, has_incomplete_scratchpad
+
+
+def test_convert_scratchpad_to_think_rewrites_tags():
+    content = "<REASONING_SCRATCHPAD>think</REASONING_SCRATCHPAD> done"
+    assert convert_scratchpad_to_think(content) == "<think>think</think> done"
+
+
+def test_has_incomplete_scratchpad_detects_real_unclosed_tag():
+    content = "Answering...\n<REASONING_SCRATCHPAD>still thinking"
+    assert has_incomplete_scratchpad(content) is True
+
+
+def test_has_incomplete_scratchpad_ignores_fenced_code_block_mentions():
+    content = """Here is the grep output:
+
+```text
+<REASONING_SCRATCHPAD>
+```
+"""
+    assert has_incomplete_scratchpad(content) is False
+
+
+def test_has_incomplete_scratchpad_ignores_blockquote_mentions():
+    content = "> literal token <REASONING_SCRATCHPAD>\n\nFinal answer."
+    assert has_incomplete_scratchpad(content) is False
+
+
+def test_has_incomplete_scratchpad_ignores_inline_code_mentions():
+    content = "The user literally typed `<REASONING_SCRATCHPAD>` in the prompt."
+    assert has_incomplete_scratchpad(content) is False
+
+
+def test_has_incomplete_scratchpad_still_flags_real_tag_after_quote():
+    content = "> quoted literal <REASONING_SCRATCHPAD>\n\n<REASONING_SCRATCHPAD>real"
+    assert has_incomplete_scratchpad(content) is True