src/flask/brew_advice.py

@@ -25,6 +25,7 @@ from __future__ import annotations
 import json
 import logging
 import os
+import re
 import urllib.error
 import urllib.request
 
@@ -271,12 +272,38 @@ def _strip_reasoning(text: str) -> str:
     working did leak in, the brewer sees a slightly rambling reply
     rather than no advice at all.
     """
-    index = text.find(_ADVICE_MARKER)
-    if index == -1:
+    index = _find_marker(text)
+    if index is None:
         return text.strip()
 
-    advice = text[index + len(_ADVICE_MARKER):].strip()
-    return advice or ""
+    advice = text[index:].strip()
+
+    # The model may wrap the marker in markdown or a code block.  Strip
+    # the leftover formatting so the advice reads as prose.
+    advice = re.sub(r"^\*\*|\*\*$", "", advice).strip()
+    if advice.startswith("```"):
+        advice = advice[3:]
+    if advice.endswith("```"):
+        advice = advice[:-3]
+    if advice.startswith("##"):
+        advice = advice.lstrip("#").strip()
+
+    return advice.strip() or ""
+
+
+def _find_marker(text: str) -> int | None:
+    """Where the advice starts, or ``None`` if there is no marker.
+
+    The marker is matched case-insensitively and with the markdown
+    decoration the model might put around it - ``**ADVICE:**``,
+    ``## ADVICE:`` - because the instruction asks for "ADVICE:" but the
+    model formats it its own way often enough to matter.
+    """
+    match = re.search(r"^[\s#>\*]*advice\s*:[\s\*>]*", text,
+                      re.IGNORECASE | re.MULTILINE)
+    if match:
+        return match.end()
+    return None
 
 
 def generate_advice(run: dict, reading: dict, history: list[dict], *,

tests/test_brew_advice.py

@@ -274,6 +274,28 @@ def test_a_reply_with_the_marker_is_kept_whole():
     assert brew_advice._strip_reasoning(reply) == advice
 
 
+def test_the_marker_is_found_however_the_model_formats_it():
+    """The model formats the marker its own way often enough to matter.
+
+    The instruction asks for "ADVICE:", but the model writes
+    "**ADVICE:**", "## ADVICE:" or puts it in a code block.  Each of
+    those has to extract the advice cleanly, with no leftover markdown.
+    """
+    advice = "Warm it to 20C and wait."
+
+    for reply in (
+        f"ADVICE:\n{advice}",
+        f"**ADVICE:**\n{advice}",
+        f"## ADVICE:\n{advice}",
+        f"```\nADVICE:\n{advice}\n```",
+        f"**ADVICE:**\n```\n{advice}\n```",
+        f"advice:\n{advice}",
+    ):
+        assert brew_advice._strip_reasoning(reply) == advice, (
+            f"the marker was not extracted cleanly from {reply!r}"
+        )
+
+
 def test_a_reply_without_the_marker_is_used_as_it_stands():
     """The safe direction when the model does not use the marker.