Setting Up Ollama for AI Summaries

What is Ollama?

Ollama is a free, open-source application that runs AI models on your computer. Myna uses it to generate meeting summaries and action items — everything stays on your machine, nothing is sent to the cloud.

Option A: Install Using the Ollama App (Easiest)

If you're not comfortable with the command line, this is the simplest approach.

1. Go to ollama.com and click Download
2. Run the installer — it takes about a minute
3. Once installed, open the Ollama app from your Start menu
4. Ollama opens a simple chat window. Type any message (e.g. “hello”) and press Enter
5. Ollama will automatically download a default model and respond — this may take a few minutes on the first run
6. Once you see a response, Ollama is ready. You can close the chat window — Ollama keeps running in your system tray

That’s it! Go back to Myna, open any transcribed recording, and click Summarise.

Option B: Install a Specific Model (Recommended for Best Results)

For better summaries, we recommend installing llama3.1:8b. This requires using the terminal briefly.

1. Go to ollama.com and click Download
2. Run the installer
3. Open PowerShell or Command Prompt (search “PowerShell” in your Start menu)
4. Type the following and press Enter:

   ollama pull llama3.1:8b

5. Wait for the download to complete (~4.7 GB). You’ll see a progress bar
6. Once done, Ollama runs in the background automatically

Checking It Works

In Myna, go to any transcribed recording and click Summarise. If Ollama is running and has a model installed, you’ll see a summary appear within 30–60 seconds.

If you see “Ollama not detected”, check that Ollama is running (look for the llama icon in your system tray near the clock).

Recommended Models

Not sure which to pick? Start with phi3:mini if your PC has less than 8 GB of RAM. Otherwise, use llama3.1:8b.

Troubleshooting

“Ollama not detected”

• Check your system tray (near the clock) for the Ollama icon
• If it’s not there, open Ollama from the Start menu
• Open PowerShell and run ollama list — if you see a model listed, Ollama is working

“Memory layout cannot be allocated” error

This means the model is too large for your available RAM or GPU memory. Try a smaller model:

ollama pull phi3:mini

Then in Myna, the smaller model will be detected automatically.

“Summary is taking a very long time”

Long meetings (30+ minutes) produce large transcripts. The first summary may take 2–5 minutes depending on your hardware. Subsequent summaries are faster as the model stays loaded in memory.

Advanced: Installing Ollama on a Different Drive

By default, Ollama installs to your C: drive and stores models in C:\Users\YourName\.ollama. If your C: drive is full, you can move models to another drive.

Move Ollama models to another drive

1. Open System Properties → Environment Variables (search “environment variables” in Start menu)
2. Under User variables, click New
3. Variable name: OLLAMA_MODELS
4. Variable value: the path on your other drive, e.g. D:\Ollama\models
5. Click OK, then restart Ollama
6. Any new models you pull will download to the new location

Note: If you already have models downloaded, you’ll need to re-download them after changing this setting, or manually move the files from C:\Users\YourName\.ollama\models to your new location.

Move the entire Ollama installation

If you want Ollama itself on another drive:

1. Set the OLLAMA_MODELS environment variable as above
2. Uninstall Ollama from C: drive
3. Re-download from ollama.com
4. During installation, choose a custom install path on your other drive

Privacy

Ollama runs entirely on your computer. When Myna sends your transcript to Ollama for summarisation, the data travels from Myna → Ollama on localhost (127.0.0.1). Nothing leaves your machine. You can verify this in Myna’s Network Log (Settings → Network Log).