Release of llm 0.36 introduces support for single-turn prompt models

The release of llm 0.36 introduces significant functionality by allowing model plugins to indicate if they do not support conversation mode. This means that models designed solely for single-turn prompts can now operate without confusion, leading to more streamlined interactions and fewer errors during usage.

When a model declares supports_conversation = False, the llm will raise an error if it receives assistant or tool history inputs. This proactive error handling prevents users from attempting to engage with models in a way they were not designed for, ultimately improving user experience and efficiency.

The update also enhances logging capabilities by wrapping reasoning traces in <details><summary> tags, which can provide clearer insights into model behavior and facilitate easier debugging. This is particularly beneficial for engineers looking to monitor performance or troubleshoot issues effectively.

With the introduction of new models, gpt-6-sol and gpt-6-luna, engineers now have access to updated capabilities that may offer improved performance or features compared to previous versions. The specific applications and advantages of these models depend on further documentation and practical testing.

Overall, the changes in llm 0.36 reflect a continuous effort to refine the usability and reliability of language models, ensuring they are equipped to handle a variety of user needs without running into compatibility issues.