Title generation skips if message generation is too fast #123
Description
When using the library with automatic title generation, if the main chat request finishes before the title generation request completes (for example, due to a very short assistant reply or high first-token latency on the title model), the generated title is never saved.
Steps to reproduce:
- Send a message with title generation enabled (for example, ask the assistant to respond with just a single word).
- If necessary, use a slower model for title generation.
- Observe that no title is generated.
The same issue occurs when using a thinking model, such as the GPT‑5 series. Thinking models should of course normally not be used for title generation, but this helps illustrate the problem.