v0.5.1 — Production Stable: Windows-Safe + Auto-Warmup

raix v0.5.1 — Production Stable

Critical Fixes

Windows-safe auto-detection — TCP port check (socketConnection) replaces curl for backend detection. No more segfaults on Windows when checking if Ollama is running.

Model pre-warming — On setup, raix sends a warmup request to load the model into RAM. Your first real call never times out.

Smart retry — Detects cold-start vs. connection failure. First call gets 180s timeout, subsequent calls 120s. Automatic retry if model is still loading.

Helpful error messages — Distinguishes between:

• Model loading into memory → suggests waiting and retrying
• Ollama not running → shows start command
• Model not found → shows pull command
• Connection refused → suggests switching provider

All Features (v0.5.x)

• 33 functions: Chat, Code Gen, Developer Agent, Compute Engine
• 13+ providers with custom endpoint support
• Small LLM optimization (auto-detects 7B-9B models)
• Modern Shiny GUI with chat bubbles and typing indicators
• Developer agent: solve, script, notebook, pipeline, compile, benchmark
• Cross-language: terminal, Python, C++/Rcpp, shell scripts

Install

remotes::install_github("twomathematicians-code/raix@v0.5.1")