LLM 0.36 Enhances Logs and Manages Single-Turn Models

Version 0.36 of the command-line tool for language models refines log display and is enriched with fixes signed by new contributors. It also introduces explicit control for single-turn limited models and references two identifiers for GPT-6 Sol and GPT-6 Luna.

Collapsible Logs and Fixes from Five New Contributors

The reasoning traces produced in the logs in Markdown format are now grouped within <details><summary> tags, allowing them to be collapsed for clearer reading. This version also includes bug fixes contributed by five new contributors.

Explicit Framing of Non-Conversation Models

Model plugins can now indicate supports_conversation = False for models that only handle single-turn prompts. If an assistant or tool history is passed to these models, LLM triggers the exception llm.ConversationNotSupported, and the llm chat command dismisses them before even opening a session. The first plugin to utilize this feature is llm-typesafe. This mechanism applies to models that do not support multi-turn conversations.

Two Identifiers for OpenAI Models Accessible via CLI

LLM allows access to language models from the command line and now references the identifiers gpt-6-sol for GPT-6 Sol and gpt-6-luna for GPT-6 Luna. These additions are part of version 0.36 and expand the list of models available through the tool.