Skip to main content

Frequently Asked Questions (FAQ)

Q: Why does M31A only support NVIDIA NIM models?​

A: High-assurance autonomous engineering requires verified model alignments, deterministic JSON tool calling, and high-throughput streaming. NVIDIA NIM provides standardized enterprise inference endpoints. Other providers are intentionally retired in production to eliminate multi-provider behavioral discrepancies.

Q: Where does M31A store its database?​

A: In platform user data directories (~/.local/share/m31a/m31a.db on Linux). It never commits databases into your repository.

Q: How do I cancel an autonomous task in progress?​

A: In the TUI, press Ctrl+C. In CLI mode, send SIGINT (Ctrl+C). M31A initiates instant cancellation, terminates child process trees, and seals output spools.

Q: Can M31A run without an internet connection?​

A: LLM reasoning requires internet access to reach the NVIDIA NIM endpoint. However, all deterministic verification checks, compilers, linters, and SQLite transactions execute 100% locally on your machine.