Apple's Open-Weight Drop and RAG-Native Reasoning Hit Local AI

Ollama v0.9.1: Critical Stability Patches for v0.9 Teething Issues

Ollama 0.9.1 is a maintenance drop that cleans up the jitter and failure modes introduced by the massive multi-GPU release last week. If you’re running the v0.9 stack for production inference, this is a mandatory stability patch.

vLLM Adds DSPy Auto-Tuning Support for Prompt Engineering

vLLM just merged support for the DSpy integration. It’s a major convenience feature for people doing automated prompt tuning (DSPy) who want to skip the export step and tune directly against a live, local vLLM server.

AutoRAG v0.2: Auto-Optimization for Local RAG Pipelines

If you're building private RAG pipelines, AutoRAG just got a major upgrade for automated benchmarking. It lets you run ablation studies across different chunking strategies and vector stores locally without writing custom boilerplate.