Ollama local setup and model configuration

Running AI models locally on your Mac ensures absolute data privacy. Meeting notes and corporate strategy never leave your hardware. Oyma connects seamlessly to Ollama, an open-source runner for local large language models.

1. Installing Ollama on macOS

1. Download Ollama from the official project website (ollama.com) or install via Homebrew:

   brew install ollama

2. Launch the Ollama application. An Ollama icon will appear in your macOS menu bar.
3. Verify that the local daemon is responding in Terminal:

   curl http://127.0.0.1:11434
   # Returns: "Ollama is running"

2. Pulling recommended models

Download your chosen model weights using the Terminal. For meeting note synthesis, we recommend:

Llama 3.2 3B (Fastest — Ideal for MacBook Air / 8 GB Macs)

ollama run llama3.2:3b

• Download size: ~2.0 GB
• RAM footprint: ~3.5 GB
• Speed: ~35–50 tokens/second on Apple Silicon

Llama 3.1 8B (Highest Quality — Recommended for 16 GB+ Macs)

ollama run llama3.1:8b

• Download size: ~4.7 GB
• RAM footprint: ~8.0 GB
• Speed: ~18–25 tokens/second on Apple Silicon

3. Configuring Oyma

1. Open Settings › AI › Provider.
2. Select Ollama (Local).
3. Confirm the Host URL: http://127.0.0.1:11434.
4. Click Refresh Models. Oyma queries your local Ollama installation and displays your installed models in a dropdown.
5. Select your model (e.g. llama3.2:3b) and click Test Connection.

Once verified, all summary generation and Ask conversational queries execute 100% on your Mac’s hardware.

Related guides

• Cloud BYOK keys in Bring your own AI API keys.
• Summary templates in AI meeting summary archetypes.
• Offline privacy in Local-only mode.