llm 0.36

The latest version of LLM, the popular command-line tool and Python library for interacting with large language models, introduces significant updates that expand model support and improve the handling of conversation limitations. Version 0.36 delivers new OpenAI model integrations alongside architectural improvements designed to better accommodate models with varying technical capabilities and conversation requirements.

LLM 0.36 introduces two new OpenAI models to its supported ecosystem: gpt-6-sol for GPT-6 Sol and gpt-6-luna for GPT-6 Luna. These additions extend the tool's compatibility with OpenAI's latest model offerings, allowing users to access cutting-edge language model capabilities through the LLM interface. The release also implements a critical architectural improvement through model plugins, which can now declare supports_conversation = False for models that only accept single-turn prompts. When these single-turn-only models receive assistant or tool history, LLM now raises an llm.ConversationNotSupported exception, providing clear error handling and improved user experience.

• Enhanced model diversity enables developers and researchers to experiment with newer OpenAI models through a unified interface
• The conversation support flag addresses a significant usability gap by preventing silent failures when incompatible conversation data is passed to limited models
• Improved error messaging through dedicated exceptions helps users understand model constraints and adjust workflows accordingly
• Plugin architecture flexibility supports a broader ecosystem of language models with varying technical specifications
• Better conversation handling reduces debugging time and supports more reliable production deployments

As the landscape of large language models continues to expand with models featuring different architectural limitations, LLM 0.36 demonstrates the importance of building flexible infrastructure that can accommodate diverse model capabilities. The explicit handling of conversation limitations represents a maturation of the LLM tooling ecosystem, moving beyond one-size-fits-all approaches toward more nuanced support for varied model types. For developers relying on LLM for production applications, these improvements provide both greater flexibility in model selection and more predictable error handling that supports robust application development.