ollama-tune — quickstart
========================

1. Put the binary somewhere on your PATH (or run it in place).
   Windows: ollama-tune.exe        Linux: chmod +x ollama-tune

2. Free, forever — the full tuning report:
     ollama-tune list
     ollama-tune profile <model>          # e.g. ollama-tune profile qwen3.5:9b
   The report measures ON YOUR BOX: GPU placement per context size, generation
   and prefill speed, the largest context that stays 100% on GPU, silent-truncation
   and reasoning-model risks, and the exact recommended config.

3. Paid features (one $49 seat = one computer, 12 months of updates):
     ollama-tune apply <model>            # bake <model>-tuned + persist env
     ollama-tune profile <model> --deep   # A/B the KV cache type empirically
     ollama-tune serve                    # local dashboard on :8770
   Activate your seat:
     ollama-tune license activate <purchase-key>
   (opens the activation page in your browser; on an offline box, run
    `ollama-tune license request` and activate from any device instead)

4. Notes
   - `--deep` restarts your local Ollama several times and ALWAYS restores stock
     settings afterward, even on error. Don't run it mid-inference.
   - Linux build is glibc-based (built on Ubuntu). Linux/macOS hardware probes are
     newer than the Windows one — report oddities to 413mrm@gmail.com and they'll
     get fixed fast.

Docs & downloads: see the product page. License terms: LICENSE.txt.
