Skip to main content

Ollama

Run everything locally. No data leaves your Mac, no API costs.

Setup​

  1. Install Ollama.
  2. Pull a model: 
    ollama pull llama3.2
  3. Make sure the Ollama server is running (the menu-bar app handles this).
  4. In ClipSlop, open Settings → Providers → Ollama.
  5. Confirm the host (http://localhost:11434 by default) and pick a model.

ClipSlop will list every model Ollama has pulled.

ClipSlop transformations are short, single-shot, and benefit from instruction-following more than raw size. Good starting points:

  • llama3.2 (3B / 8B) — solid baseline for grammar, formatting, summarising.
  • qwen2.5 (7B / 14B) — strong on translation, especially Asian languages.
  • phi3 — small, fast, good for grammar fixes when you want minimum latency.
  • mistral-nemo — wide language coverage, good for Rewrite tasks.

For coding-focused prompts (Dev → Beautify Code, Explain Stack Trace), try a code-specialised model like deepseek-coder or qwen2.5-coder.

Trade-offs​

  • ✅ Private. Nothing leaves your Mac.
  • ✅ Free.
  • ⚠️ Slower than cloud APIs unless you have a beefy GPU.
  • ⚠️ Quality is below frontier models. Expect more retries on nuanced rewrites.

Mixing providers​

You can mix Ollama with cloud providers per prompt. E.g. Fix Grammar on Ollama, Translate on Claude.