Dependency-aware execution
Cells execute in a deterministic order. Deleting a cell removes its variables and updates affected cells. For expensive computations, users can mark dependent outputs as stale instead of rerunning them automatically.
marimo
Free
An open-source reactive Python notebook for data scientists and AI engineers.
Qu'est-ce que marimo ?
marimo is an open-source Python notebook for exploring data, developing models, and building interactive applications. It stores notebooks as plain Python files, connecting notebook work with standard software development workflows.
Its reactive execution model tracks dependencies between cells and updates affected outputs when code or inputs change. The same notebook can serve as an analysis workspace, an executable script, or an application for people to explore through a browser.
Que peut faire marimo ?
Interactive data exploration
Dataframe tables offer sorting, filtering, search, column statistics, data types, and missing-value information. Sliders, dropdowns, and selectable plots let users control analyses through interactive inputs.
SQL alongside Python
Built-in SQL cells query dataframes and databases, with support for Pandas, Polars, PyArrow, DuckDB, SQLite, Postgres, and MySQL.
AI assistance and coding agents
Configure context-aware completion, suggested error fixes with visual diffs, and sidebar chat. marimo pair gives coding agents access to live notebook cells and state; the Agent sidebar supports ACP-compatible agents.
Python development workflow
Store notebooks as .py files for Git versioning, module reuse, and PyTest testing. The editor includes formatting, debugging panels, vim keybindings, and hotkeys, alongside a CLI and VS Code extension.
App serving and HTML export
Serve notebooks as web apps with the marimo CLI or export compatible notebooks to WebAssembly-powered HTML. Run the same files as scripts locally or in pipelines.
Cas d'utilisation
“Investigate datasets”
Data scientists can inspect data quality, adjust preprocessing parameters, and explore Altair or Plotly visualizations with hover details, zooming, and filtering.
“Share analysis with stakeholders”
Turn an existing notebook into an application where colleagues can adjust parameters and explore results independently, without rebuilding the analysis in a separate dashboard tool.
“Prototype AI applications”
AI engineers can combine custom chat widgets, file uploads, and dynamic visualizations to test model interactions and share working demos with colleagues.
“Build labeling and evaluation interfaces”
Create custom annotation tools, model comparison dashboards, and human-in-the-loop workflows for evaluating multimodal models or preparing fine-tuning examples.
Questions fréquentes
How do I install marimo?
Install the Python package with `uv pip install marimo`. Open a notebook for editing with `marimo edit notebook.py`.
Can a notebook record its package dependencies?
Yes. Notebooks can optionally include package requirements through PEP 723 metadata, with marimo managing virtual environments.
Which AI providers can I use?
marimo supports OpenAI, Anthropic, Google Gemini, Ollama, and OpenAI-compatible APIs. You can configure hosted providers or models running locally.
How do I launch a notebook as an app?
Run `marimo run notebook.py` to serve the notebook as an interactive application. Existing widgets provide controls for app users.
Is marimo free?
The notebook software is free and open source under Apache-2.0. AI assistance uses your configured model provider, which can be hosted or local. Model services and deployment infrastructure are separate from the notebook license.
Informations rapides
- Product type
- Reactive Python notebook
- License model
- Free, open source under Apache-2.0
- Primary users
- Data scientists, AI engineers, and Python developers
- Notebook format
- Plain Python (.py)
- Website
- https://marimo.io
Analyse du trafic de marimo
Les données de trafic sont fournies à titre indicatif uniquement.
- Visites mensuelles
- 171,4 k
- Classement mondial
- #205 482
- Taux de rebond
- 37.9%
- Durée moyenne de visite
- 01:11
- Pages par visite
- 3.76
- Évaluation du domaine
- 72
Tendances du trafic
| avr. | 171025 |
|---|---|
| mai | 156621 |
| juin | 171371 |
Sources de trafic
- Direct: 31.5%
- Recherche: 43.3%
- Référents: 18.7%
- Réseaux sociaux: 5.20%
- E-mail: 0.84%
- Affiliation: 0.00%
Principales régions
- États-Unis: 19.4%
- Allemagne: 7.53%
- Royaume-Uni: 7.32%
- Inde: 5.21%
- Brésil: 4.21%
- Autres: 56.3%
Alternatives à marimo
LiveKit
livekit.io
LiveKit est une plateforme open source pour développeurs permettant de créer, déployer et observer des agents IA vocaux, vidéo et physiques en Python ou TypeScript, avec LiveKit Cloud pour la production.
AiPy
aipy.app
AiPy est un assistant IA de bureau qui combine raisonnement IA et exécution Python pour les tâches locales, avec téléchargements pour Mac, Windows et Linux et déploiement local open source.
TensorFlow
tensorflow.org
Plateforme open source de machine learning pour créer, entraîner et déployer des modèles sur ordinateur, mobile, web, cloud et edge.
OpenCode
opencode.ai
OpenCode est un agent de programmation IA open source pour terminal, IDE et ordinateur, qui aide à comprendre les bases de code, planifier les changements et utiliser plusieurs fournisseurs de LLM.
Cline
cline.bot
Cline est un agent de programmation IA open source, disponible en extension VS Code, en CLI et en offre Entreprise. Il aide les développeurs à créer, modifier et vérifier du code grâce à des actions soumises à autorisation et à une facturation de l’inférence IA à l’usage.
Continue
continue.dev
Continue est un agent de programmation open source destiné aux développeurs.