TL;DR: Ollama v0.40.7 enhances OpenAI API compatibility, fixes streaming issues, and adds model introspection features, improving local LLM development workflows.
Summary: Ollama v0.40.7 was released, improving the stability of continuing OpenAI-compatible conversations, especially when switching models or handling compaction items. It also fixed misordered tool call events in streaming responses and introduced new renderer, parser, and template_source fields in ollama show and the API for better model introspection. Additionally, it updated llama.cpp and MLX, and improved speculative decoding for draft models.
Why it matters: These updates make Ollama a more robust platform for local LLM development, especially for those leveraging OpenAI API compatibility. Builders should explore the new model introspection features and test improved streaming and conversation continuity for their applications.
Source: github_releases