[Bug]: auxiliary title generation config is hardcoded to 30s and ignores config

[veryheavypickle]

Bug Description

Problem

What happened

After a session completes, I get the message ⚠ Auxiliary title generation failed: Request timed out.. After investigating this issue, I updated the config auxiliary.title_generation.timeout to something high like 1800, and still get this issue.

Why this happens

I run local LLMs and setting context to high on qwen3.6:27b uses almost double the VRAM that I have, so the model runs very slowly.

What I expected

For the config parameter to work.

Solution

I copied what I had here into the analysis part

Steps to Reproduce

1. Use a very slow model (would have to be local)
2. Chat with hermes about whatever
3. Wait about 3 minutes
4. ⚠ Auxiliary title generation failed: Request timed out.

Expected Behavior

The title_generation uses the auxiliary.title_generation.timeout config

Actual Behavior

The title_generation does not use the auxiliary.title_generation.timeout config

Affected Component

Configuration (config.yaml, .env, hermes setup)

Messaging Platform (if gateway-related)

No response

Debug Report

Logs contain personal information (domains, usernames etc) so I won't share it. However nothing of importance is in these logs.

Operating System

Alpine Linux

Python Version

3.13.5

Hermes Version

0.14.0 (2026.5.16)

Additional Logs / Traceback (optional)

Root Cause Analysis (optional)

I have identified in the code where this disconnect from the config file happens. So I can later create a pull request to fix this.
In agent/title_generator.py we can trace what happens, -> maybe_auto_title -> auto_title_session ->
Once we get to generate_title, we can see that auto_title_session passes no timeout to generate_title, so it defaults to 30s.

Proposed Fix (optional)

The solution is to use the auxiliary.title_generation.timeout config. I haven't checked the rest of the codebase for how using the config is implemented but it should be a simple fix.

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR

[veryheavypickle]

Most relevant issue is Auxiliary LLM calls (session_search, skills_hub, etc.) use short 30s default timeout with no local-endpoint detection — causes retry storms on slow local inference

[alt-glitch]

Duplicate of #18129 (open issue) — same bug: auto_title_session calls generate_title without passing the configured timeout, so it defaults to 30s. Open fix PR: #18839.

[veryheavypickle]

@alt-glitch Thank you, there are so many issues open that I must have skimmed past it. I will check out the PR

[veryheavypickle]

This issue will be resolved with merge of #18129

[bsbofmusic]

I independently hit and verified the same configuration issue after updating Hermes.

Current behavior: generate_title() still defaults to a hardcoded timeout value, so the configured auxiliary timeout cannot fully control title generation:

def generate_title(..., timeout: float = 30.0, ...):
    ...
    call_llm(..., timeout=timeout, ...)

The local minimal fix that makes auxiliary.title_generation.timeout effective is to let call_llm() inherit the auxiliary config by default:

from typing import Optional

def generate_title(..., timeout: Optional[float] = None, ...):
    ...
    call_llm(..., timeout=timeout, ...)

I added a regression test locally to assert the default call passes timeout=None rather than 30.0:

def test_default_timeout_allows_auxiliary_config_inheritance():
    ...
    generate_title("question", "answer")
    assert captured_kwargs["timeout"] is None

Verification:

python -m pytest -q tests/agent/test_title_generator.py

This is a config correctness bug rather than a preference change: if a user has set auxiliary.title_generation.timeout, the title generation path should not silently override it with a hardcoded 30 seconds.