From d7bac9f01e7326d3a4fd7e7dbe2b6ccd15fcf984 Mon Sep 17 00:00:00 2001
From: liuhao1024 <sunsky.lau@gmail.com>
Date: Mon, 8 Jun 2026 13:40:16 +0800
Subject: [PATCH] fix(agent): respect auxiliary.title_generation.timeout config
 for title generation

generate_title() defaulted to timeout=30.0, which caused call_llm() to use
that hardcoded value instead of reading auxiliary.title_generation.timeout
from config.yaml. Changed the default to None so call_llm's config-driven
resolution kicks in (reads auxiliary.title_generation.timeout from config).

Fixes #41812
---
 agent/title_generator.py            |  7 +++++-
 tests/agent/test_title_generator.py | 38 +++++++++++++++++++++++++++++
 2 files changed, 44 insertions(+), 1 deletion(-)

diff --git a/agent/title_generator.py b/agent/title_generator.py
index a7f1e158e1a6..357ea1e06459 100644
--- a/agent/title_generator.py
+++ b/agent/title_generator.py
@@ -29,7 +29,7 @@
 def generate_title(
     user_message: str,
     assistant_response: str,
-    timeout: float = 30.0,
+    timeout: Optional[float] = None,
     failure_callback: Optional[FailureCallback] = None,
     main_runtime: dict = None,
 ) -> Optional[str]:
@@ -39,6 +39,11 @@ def generate_title(
     auxiliary LLM client (cheapest/fastest available model).
     Returns the title string or None on failure.
 
+    ``timeout`` defaults to ``None``, which delegates to
+    ``call_llm``'s config-driven resolution (reads
+    ``auxiliary.title_generation.timeout`` from config.yaml).
+    An explicit value overrides the config.
+
     ``failure_callback`` is invoked with ``(task, exception)`` when the
     auxiliary call raises — the caller typically wires this to
     ``AIAgent._emit_auxiliary_failure`` so the user sees a warning instead
diff --git a/tests/agent/test_title_generator.py b/tests/agent/test_title_generator.py
index 56286f6ecc99..e2b4a4909539 100644
--- a/tests/agent/test_title_generator.py
+++ b/tests/agent/test_title_generator.py
@@ -93,6 +93,44 @@ def test_no_callback_matches_legacy_behavior(self):
         with patch("agent.title_generator.call_llm", side_effect=RuntimeError("nope")):
             assert generate_title("q", "a") is None
 
+    def test_default_timeout_delegates_to_config(self):
+        """Regression: timeout=None delegates to call_llm config resolution (#41812).
+
+        Previously generate_title defaulted to timeout=30.0, which caused
+        call_llm to use that hardcoded value instead of reading
+        auxiliary.title_generation.timeout from config.yaml.
+        """
+        captured_kwargs = {}
+
+        def mock_call_llm(**kwargs):
+            captured_kwargs.update(kwargs)
+            resp = MagicMock()
+            resp.choices = [MagicMock()]
+            resp.choices[0].message.content = "Config-Aware Title"
+            return resp
+
+        with patch("agent.title_generator.call_llm", side_effect=mock_call_llm):
+            generate_title("question", "answer")
+
+        # Default must be None so call_llm reads from config
+        assert captured_kwargs.get("timeout") is None
+
+    def test_explicit_timeout_is_forwarded(self):
+        """When an explicit timeout is passed, it must reach call_llm."""
+        captured_kwargs = {}
+
+        def mock_call_llm(**kwargs):
+            captured_kwargs.update(kwargs)
+            resp = MagicMock()
+            resp.choices = [MagicMock()]
+            resp.choices[0].message.content = "Explicit Timeout"
+            return resp
+
+        with patch("agent.title_generator.call_llm", side_effect=mock_call_llm):
+            generate_title("question", "answer", timeout=120.0)
+
+        assert captured_kwargs.get("timeout") == 120.0
+
     def test_truncates_long_messages(self):
         """Long user/assistant messages should be truncated in the LLM request."""
         captured_kwargs = {}