[claude] Document and validate AirLLM Apple Silicon requirements (#1284) (#1298)
Some checks failed
Tests / lint (push) Has been cancelled
Tests / test (push) Has been cancelled

This commit was merged in pull request #1298.
This commit is contained in:
2026-03-24 01:52:16 +00:00
parent fc0a94202f
commit 00ef4fbd22
5 changed files with 178 additions and 3 deletions

View File

@@ -27,8 +27,12 @@
# ── AirLLM / big-brain backend ───────────────────────────────────────────────
# Inference backend: "ollama" (default) | "airllm" | "auto"
# "auto" → uses AirLLM on Apple Silicon if installed, otherwise Ollama.
# Requires: pip install ".[bigbrain]"
# "ollama" always use Ollama (safe everywhere, any OS)
# "airllm" → AirLLM layer-by-layer loading (Apple Silicon M1/M2/M3/M4 only)
# Requires 16 GB RAM minimum (32 GB recommended).
# Automatically falls back to Ollama on Intel Mac or Linux.
# Install extra: pip install "airllm[mlx]"
# "auto" → use AirLLM on Apple Silicon if installed, otherwise Ollama
# TIMMY_MODEL_BACKEND=ollama
# AirLLM model size (default: 70b).