v2.1.0

2.1.0 (2026-07-26)

✨ Features

• add oMLX (ol.mlx) to openai-compatible preset apps (#533) (6a191ba)

🐛 Bug Fixes

• enforce direct-exec for single-step tool plans and reject stop-only planner plans (#541) (da5a430)
• make OpenAI-compatible warmup send actual inference to pre-load model (#539) (537c989)
• mcp: flatten real SDK content blocks instead of their repr (#548) (856b2ce)
• mcp: honour per-server timeout_sec and stop swallowing empty error messages (#550) (de7133f)
• re-encode tool_calls arguments as JSON strings for OpenAI compat (#560) (841f893)
• reply: strengthen SMALL tool incentives and add weather-specific tool-use rule (cf8e87d)
• use /api/chat instead of /api/generate for Ollama warmup to exercise full inference pipeline (#542) (1325941), closes #537
• wake-worded statements no longer misclassified as 'not directed' (#543) (4d2c8ec)

⚡ Performance Improvements

• reduce voice response latency for simple queries (#534) (a3fe9b2)

📝 Documentation

• add triage patterns from full-log review, wake word no-query, hallucination scripts (#538) (17920c9)
• restructure triage skill diagnosis into ordered flow with common patterns (#536) (42b9e48)
• split review-pr into standard and ultra tiers (#552) (38591b8)

---

⚡ Prerequisites

• Ollama (default), or any OpenAI-compatible server (LM Studio, Jan, llama.cpp, vLLM, oMLX, LocalAI, …)

📦 Downloads

Platform	File	Notes
Windows	Jarvis-Windows-x64.zip	Extract → Run Jarvis.exe
macOS (Apple Silicon)	Jarvis-macOS-arm64.zip	Extract → Move to Applications → Right-click → Open
macOS (Intel)	Jarvis-macOS-x64.zip	Extract → Move to Applications → Right-click → Open
Linux	Jarvis-Linux-x64.tar.gz	tar -xzf → Run ./Jarvis/Jarvis