OlunguSupport

LOCAL AI SETUP

Set up Ollama for Olungu

Choose your operating system and how you run Ollama. The page will show only the instructions for that setup, with the exact dashboard values and verification commands to use.

Choose your setup

These choices change the steps below. If you are unsure, keep the default option for your operating system.

Operating system

How Ollama runs

Model preference

macOS setup with the Ollama app

This is the usual Mac setup. The key detail is setting OLLAMA_ORIGINS with launchctl because menu bar apps inherit their environment from macOS.

  1. 1

    Install Ollama

    Ollama provides the local HTTP server that Olungu calls. Without the server running, the dashboard fields cannot connect to anything.

    How to do it

    • Go to the official Ollama download page.
    • Download the installer for your operating system.
    • Install it using the default options.
    • Open a new terminal after installation so the ollama command is available.

    Expected result: Running ollama --version should print a version number. If it does not, close and reopen the terminal.

  2. 2

    Download the model

    Ollama only serves models that have been pulled onto your machine. Olungu will return a model-not-found error if this step is skipped.

    How to do it

    • Use llama3.2:3b for this setup.
    • Run the pull command and wait for the download to finish.
    • If you later choose a different model in Olungu, pull that model first too.
    ollama pull llama3.2:3b
ollama list

    Expected result: The model list should include llama3.2:3b.

  3. 3

    Allow Olungu's extension origin

    Chrome extensions do not share the same origin as ordinary localhost pages. Ollama must explicitly allow extension origins or it can return HTTP 403.

    How to do it

    • Open Terminal.
    • Run the launchctl command below exactly as shown.
    • This stores the setting for GUI apps launched by macOS.
    launchctl setenv OLLAMA_ORIGINS "chrome-extension://*,http://localhost:*,http://127.0.0.1:*"

    Expected result: This command does not print much when it succeeds. The setting takes effect after Ollama restarts.

  4. 4

    Restart Ollama completely

    Changing OLLAMA_ORIGINS does not update an already-running Ollama process. A full restart is required.

    How to do it

    • Quit Ollama from the menu bar if it is visible.
    • Run the kill commands below to clear any leftover server process.
    • Start Ollama again from Applications or with open -a Ollama.
    pkill -9 Ollama
pkill -9 ollama
open -a Ollama

    Expected result: After restart, http://localhost:11434 should open a short Ollama running message in the browser.

  5. 5

    Fill in the Olungu dashboard

    Olungu needs to know that classification should use your local Ollama server instead of managed AI or browser built-in local AI.

    How to do it

    • Open the Olungu extension dashboard.
    • Find the AI configuration section.
    • Set AI Mode to Local Model / Ollama (best privacy).
    • Set Local Model Provider to Ollama local endpoint.
    • Set Ollama Base URL to http://localhost:11434.
    • Set Ollama Model to llama3.2:3b.

    Expected result: Do not include /api/generate or /api/chat in the base URL. Olungu adds the Ollama API path itself.

  6. 6

    Verify from the browser-origin path

    A normal local curl can work while browser extension requests still fail. This test includes an Origin header so it catches the real CORS issue.

    How to do it

    • Run the command below from Terminal, PowerShell, or Command Prompt.
    • A successful setup returns HTTP 200 and a JSON response from Ollama.
    • If you see HTTP 403, go back to the origin step and restart Ollama again.
    curl -i http://localhost:11434/api/generate \
  -H 'Origin: chrome-extension://test' \
  -H 'Content-Type: application/json' \
  -d '{"model":"llama3.2:3b","prompt":"Return only {\"ok\":true}","stream":false,"format":"json"}'

    Expected result: After this passes, open a normal browsing page and let Olungu classify it once. That confirms the extension path, not only the dashboard page.

What Olungu should contain

Use these values after completing the setup steps above.

AI Mode

Local Model / Ollama (best privacy)

Local Model Provider

Ollama local endpoint

Ollama Base URL

http://localhost:11434

Ollama Model

llama3.2:3b

Troubleshooting

Start with the error shown in Olungu or in the verification command.

Olungu shows HTTP 403

Ollama is running, but the extension origin is not allowed. Recheck OLLAMA_ORIGINS for your selected setup and fully restart Ollama.

Olungu shows HTTP 404

Use http://localhost:11434 as the base URL and make sure the selected model appears in ollama list.

The first classification is slow

The model is loading into memory. This is normal. Smaller models start faster; larger models usually classify better.

Classifications are weak

Pull qwen2.5:7b or llama3.1:8b, then change only the Ollama Model field in Olungu.

Security note

chrome-extension://* is convenient because it works before you look up the exact Olungu extension ID. It also allows any Chrome extension on your machine to call Ollama. For a stricter setup, replace it with Olungu's exact extension origin.

Download OllamaOllama environment docs