AI completion¶
marimo comes with GitHub Copilot, a tool that helps you write code faster by suggesting in-line code suggestions based on the context of your current code.
marimo also comes with the ability to use AI for refactoring a cell, finishing writing a cell, or writing a full cell from scratch. This feature is currently experimental and is not enabled by default.
GitHub Copilot¶
The marimo editor natively supports GitHub Copilot, an AI pair programmer, similar to VS Code.
Get started with Copilot:
Install Node.js.
Enable Copilot via the settings menu in the marimo editor.
Note: Copilot is not yet available in our conda distribution; please install
marimo using pip
if you need Copilot.
Codeium Copilot¶
Go to the Codeium website and sign up for an account: https://codeium.com/
Install the browser extension: https://codeium.com/chrome_tutorial
Open the settings for the Chrome extension and click on “Get Token”
Right-click on the extension window and select “Inspect” to open the developer tools for the extension. Then click on “Network”
Copy the token and paste it into the input area, and then press “Enter Token”
This action will log a new API request in the Network tab. Click on “Preview” to get the API key.
Paste the API key in the marimo settings in the UI, or add it to your
~/.marimo.toml
file as follows:
[completion]
copilot = "codeium"
codeium_api_key = ""
Using AI to modify cells¶
This feature is currently experimental and is not enabled by default. To enable it:
Install openai:
pip install openai
Add the following to your
~/.marimo.toml
:
[ai.open_ai]
# Get your API key from https://platform.openai.com/account/api-keys
api_key = "sk-..."
# Choose a model, we recommend "gpt-4-turbo"
model = "gpt-4-turbo"
# Change the base_url if you are using a different OpenAI-compatible API
base_url = "https://api.openai.com/v1"
Once enabled, you can use AI completion by pressing Ctrl/Cmd-Shift-e
in a
cell. This will open an input to modify the cell using AI.
Using Anthropic¶
To use Anthropic with marimo:
Sign up for an account at Anthropic and grab your Anthropic Key.
Add the following to your
~/.marimo.toml
:
[ai.open_ai]
model = "claude-3-5-sonnet-20240620"
# or any model from https://docs.anthropic.com/en/docs/about-claude/models
[ai.anthropic]
api_key = "sk-..."
Using other AI providers¶
marimo supports OpenAI’s GPT-3.5 API by default. If your provider is compatible with OpenAI’s API, you can use it by changing the base_url
in the configuration.
For other providers not compatible with OpenAI’s API, please submit a feature request or “thumbs up” an existing one.
Using Ollama¶
Ollama allows you to run open-source LLMs (e.g. Llama 3.1, Phi 3, Mistral, Gemma 2) on your local machine. To integrate Ollama with marimo:
Download and install Ollama.
Download the model you want to use:
ollama pull llama2
We also recommend
codellama
(code specific) orllama3.1
(more advanced, but larger).
Start the Ollama server:
ollama run llama2
Visit http://localhost:11434 to confirm that the server is running.
Add the following to your
~/.marimo.toml
:
[ai.open_ai]
api_key = "ollama" # This is not used, but required
model = "llama2" # or the model you downloaded from above
base_url = "http://localhost:11434/v1"