# Receipt — `~/.ollama/models` moved to the T7 (tic 751, ruled /review 751 second pass Q4)

- **What moved:** 15 GB / 51 files (biomed, gpt-oss, llama2, llama3.2, qwen3, qwen3-embedding — served by no lane; `ollama ps` empty at every reading this tic) from the internal disk to `/Volumes/T7 Shield/models/ollama-library/models/` (the directory already held 68 files; merged, nothing overwritten by checksum).
- **How:** `rsync -a` (16,435,579,933 bytes sent) → checksum dry-run verify: **0 files differing**, src 51 / dst 119 → only then `rm -rf ~/.ollama/models` → `ln -s "/Volumes/T7 Shield/models/ollama-library/models" ~/.ollama/models`. The first attempt refused itself (macOS rsync rejected a flag; the verify read dst=0; the guard kept the internal copy) and was re-issued — nothing was deleted unverified.
- **Result:** `~/.ollama/models` is a symlink; `ollama` works whenever the T7 is mounted (set `OLLAMA_MODELS` to the T7 path if you ever want the app independent of the link). Internal volume: **11 → 28 GiB free** (`df`), container free 30.2 GB.
- **Log:** `audit-logs/services/local-inference/ollama-models-move-tic751.log`.
- **Not touched:** the T7 fleet (Qwen3.6-27B-MTP, organization-engine-lora, the rest), the llama.cpp path the laps use, anything else on the internal disk.