marimo-flow 0.3.0

Reactive marimo notebooks for ML with MLflow tracking, PINA physics-informed neural networks, and a multi-agent team on pydantic-graph + Ollama Cloud.

Marimo Flow 🌊

---

Like marimo algae drifting in crystal waters, your data flows and evolves – each cell a living sphere of computation, gently touching others, creating ripples of reactive change. In this digital ocean, data streams like currents, models grow like organic formations, and insights emerge naturally from the depths. Let your ML experiments flow freely, tracked and nurtured, as nature intended.

https://github.com/user-attachments/assets/3bc24463-ff42-44a7-ae61-5d500d29688c

Why Marimo Flow is Powerful 🚀

Marimo Flow combines reactive notebook development with AI-powered assistance and robust ML experiment tracking:

• 🤖 AI-First Development with MCP: Model Context Protocol (MCP) integration brings live documentation, code examples, and AI assistance directly into your notebooks - access up-to-date library docs for Marimo, Polars, Plotly, and more without leaving your workflow
• 🔄 Reactive Execution: Marimo's dataflow graph ensures your notebooks are always consistent - change a parameter and watch your entire pipeline update automatically
• 📊 Seamless ML Pipeline: MLflow integration tracks every experiment, model, and metric without breaking your flow
• 🎯 Interactive Development: Real-time parameter tuning with instant feedback and beautiful visualizations

This combination eliminates the reproducibility issues of traditional notebooks while providing AI-enhanced, enterprise-grade experiment tracking.

Features ✨

🤖 AI-Powered Development (MCP)

• Model Context Protocol Integration: Live documentation and AI assistance in your notebooks
• Context7 Server: Access up-to-date docs for any Python library without leaving marimo
• Marimo MCP Server: Specialized assistance for marimo patterns and best practices
• Local LLM Support: Ollama integration for privacy-focused AI code completion

📊 ML Development Workflow

• 📓 Reactive Notebooks: Git-friendly .py notebooks with automatic dependency tracking
• 🔬 MLflow Tracking: Complete ML lifecycle management with model registry
• 🎯 Interactive Development: Real-time parameter tuning with instant visual feedback
• 💾 SQLite Backend: Lightweight, file-based storage for experiments

🐳 Deployment

• Docker: docker-compose setup with CPU, CUDA, and XPU image variants
• 🧠 PINA Integration: Physics-informed neural networks with Walrus foundation model
• 📚 MCP-Powered Docs: Live documentation via Context7 and Marimo MCP servers

Quick Start 🏃‍♂️

With Docker (Recommended)

# Clone repository
git clone https://github.com/synapticore-io/marimo-flow.git
cd marimo-flow

# Build and start services
docker compose -f docker/docker-compose.yaml up --build -d

# Access services
# Marimo: http://localhost:2718
# MLflow: http://localhost:5000

# View logs
docker compose -f docker/docker-compose.yaml logs -f

# Stop services
docker compose -f docker/docker-compose.yaml down

Docker Image Variants

Variant	Image Tag	Use Case
CPU	ghcr.io/synapticore-io/marimo-flow:latest	No GPU (lightweight)
CUDA	ghcr.io/synapticore-io/marimo-flow:cuda	NVIDIA GPUs
XPU	ghcr.io/synapticore-io/marimo-flow:xpu	Intel Arc/Data Center GPUs

# NVIDIA GPU (requires nvidia-docker)
docker compose -f docker/docker-compose.cuda.yaml up -d

# Intel GPU (requires Intel GPU drivers)
docker compose -f docker/docker-compose.xpu.yaml up -d

Local Development

# Install dependencies
uv sync

# Start MLflow server (in background or separate terminal)
uv run mlflow server \
  --host 0.0.0.0 \
  --port 5000 \
  --backend-store-uri sqlite:///data/experiments/db/mlflow.db \
  --default-artifact-root ./data/experiments/artifacts \
  --serve-artifacts

# Start Marimo (in another terminal)
uv run marimo edit examples/

Example Notebooks 📚

All notebooks live in examples/ and can be opened with uv run marimo edit examples/<file>.py.

• 01_pina_poisson_solver.py – Solve the Poisson equation with baseline PINNs or the Walrus foundation model. Training is tracked in MLflow with integrated Optuna sweep analytics and experiment history.

Project Structure 📁

marimo-flow/
├── examples/                    # Marimo notebooks
│   └── 01_pina_poisson_solver.py
├── src/marimo_flow/             # Installable package
│   └── core/                    # PINA solvers, training, visualization
├── docs/                        # Project documentation
├── docker/                      # Dockerfiles + compose (CPU, CUDA, XPU)
├── data/mlflow/                 # MLflow storage (artifacts, db)
└── pyproject.toml               # Dependencies

The marimo_flow Package

from marimo_flow.core import (
    ModelFactory,        # Create PINA neural network models
    ProblemManager,      # Define PDE problems and domains
    SolverManager,       # Configure PINN / SAPINN solvers
    WalrusAdapter,       # Walrus foundation model adapter
    build_optuna_history_figure,
    build_optuna_param_importance_figure,
    build_optuna_parallel_figure,
)

MCP (Model Context Protocol) Integration 🔌

Marimo Flow is AI-first with built-in Model Context Protocol (MCP) support for intelligent, context-aware development assistance.

Why MCP Matters

Traditional notebooks require constant context-switching to documentation sites. With MCP:

• 📚 Live Documentation: Access up-to-date library docs directly in marimo
• 🤖 AI Code Completion: Context-aware suggestions from local LLMs (Ollama)
• 💡 Smart Assistance: Ask questions about libraries and get instant, accurate answers
• 🔄 Always Current: Documentation updates automatically, no more outdated tutorials

Pre-Configured MCP Servers

Context7 - Universal Library Documentation

Access real-time documentation for any Python library:

# Ask: "How do I use polars window functions?"
# Get: Current polars docs, code examples, best practices

# Ask: "Show me plotly 3D scatter plot examples"
# Get: Latest plotly API with working code samples

Supported Libraries:

• Polars, Pandas, NumPy - Data manipulation
• Plotly, Altair, Matplotlib - Visualization
• Scikit-learn, PyTorch - Machine Learning
• And 1000+ more Python packages

Marimo - Specialized Notebook Assistance

Get expert help with marimo-specific patterns:

# Ask: "How do I create a reactive form in marimo?"
# Get: marimo form patterns, state management examples

# Ask: "Show me marimo UI element examples"
# Get: Complete UI component reference with code

Real-World Examples

Example 1: Learning New Libraries

# You're exploring polars window functions
# Type: "polars rolling mean example"
# MCP returns: Latest polars docs + working code
df.with_columns(
    pl.col("sales").rolling_mean(window_size=7).alias("7d_avg")
)

Example 2: Debugging

# Stuck on a plotly error?
# Ask: "Why is my plotly 3D scatter not showing?"
# Get: Common issues, solutions, and corrected code

Example 3: Best Practices

# Want to optimize code?
# Ask: "Best way to aggregate in polars?"
# Get: Performance tips, lazy evaluation patterns

AI Features Powered by MCP

• Code Completion: Context-aware suggestions as you type (Ollama local LLM)
• Inline Documentation: Hover over functions for instant docs
• Smart Refactoring: AI suggests improvements based on current libraries
• Interactive Q&A: Chat with AI about your code using latest docs

Configuration

MCP servers are pre-configured in .marimo.toml:

[mcp]
presets = ["context7", "marimo"]

[ai.ollama]
model = "gpt-oss:20b-cloud"
base_url = "http://localhost:11434/v1"

If you're running inside Docker, the same mcp block lives in docker/.marimo.toml, so both local and containerized sessions pick up identical presets.

Adding Custom MCP Servers

You can extend functionality by adding custom MCP servers in .marimo.toml:

[mcp.mcpServers.your-custom-server]
command = "npx"
args = ["-y", "@your-org/your-mcp-server"]

MLflow Trace Server (Optional)

Expose MLflow trace operations to MCP-aware IDEs/assistants (e.g., Claude Desktop, Cursor) by running:

mlflow mcp run

Run the command from an environment where MLFLOW_TRACKING_URI (or MLFLOW_BACKEND_STORE_URI/MLFLOW_DEFAULT_ARTIFACT_ROOT) points at your experiments. The server stays up until interrupted and can be proxied alongside Marimo/MLflow so every tool shares the same MCP context.

Learn More:

• Marimo MCP Guide - Official MCP documentation
• Model Context Protocol - MCP specification and resources

Claude Code Integration 🤖

Marimo Flow includes full Claude Code support with domain-specific skills, MCP servers, and automated hooks.

Pre-Configured MCP Servers

Server	Purpose	Config
marimo	Notebook inspection, debugging, linting	HTTP on port 2718
mlflow	Trace search, feedback, evaluation	stdio via mlflow mcp run
context7	Live library documentation	stdio via npx
serena	Semantic code search	stdio via uvx

Start marimo MCP server:

# Install once (recommended)
uv tool install "marimo[lsp,recommended,sql,mcp]>=0.18.0"

# Start server
marimo edit --mcp --no-token --port 2718 --headless

Domain Skills

Three specialized skills in .claude/Skills/ provide expert guidance:

Skill	Triggers	MCP Tools
marimo	marimo, reactive notebook, mo.ui	Notebook inspection, linting, context7 docs
mlflow	mlflow, experiment tracking, genai tracing	Trace search, feedback, evaluation, context7 docs
pina	pina, pinns, pde solver, neural operator	MLflow tracking, context7 docs

Pre-resolved context7 library IDs (no lookup needed):

• /marimo-team/marimo - marimo docs (2,413 snippets)
• /mlflow/mlflow - mlflow docs (9,559 snippets)
• /mathlab/pina - PINA docs (2,345 snippets)

Automated Hooks

Cross-platform hooks in .claude/settings.json:

Hook	Trigger	Action
SessionStart	Session begins	Start marimo MCP server
PostToolUse	Edit/Write .py files	Auto-format with ruff
PreToolUse	Edit uv.lock	Block (protection)

VS Code Copilot

MCP config for VS Code Copilot in .vscode/mcp.json:

{
  "servers": {
    "marimo": { "type": "http", "url": "http://127.0.0.1:2718/mcp/server" },
    "mlflow": { "type": "stdio", "command": "mlflow", "args": ["mcp", "run"] }
  }
}

Configuration ⚙️

Environment Variables

Docker setup (configured in docker/docker-compose.yaml):

• MLFLOW_BACKEND_STORE_URI: sqlite:////app/data/experiments/db/mlflow.db
• MLFLOW_DEFAULT_ARTIFACT_ROOT: /app/data/experiments/artifacts
• MLFLOW_HOST: 0.0.0.0 (allows external access)
• MLFLOW_PORT: 5000
• OLLAMA_BASE_URL: http://host.docker.internal:11434 (requires Ollama on host)

Local development:

• MLFLOW_TRACKING_URI: http://localhost:5000 (default)

Docker Services

The Docker container runs both services via docker/start.sh:

• Marimo: Port 2718 - Interactive notebook environment
• MLflow: Port 5000 - Experiment tracking UI

GPU Support: Use docker-compose.cuda.yaml for NVIDIA GPUs or docker-compose.xpu.yaml for Intel GPUs. The default docker-compose.yaml is CPU-only.

API Endpoints 🔌

MLflow REST API

• Experiments: GET /api/2.0/mlflow/experiments/list
• Runs: GET /api/2.0/mlflow/runs/search
• Models: GET /api/2.0/mlflow/registered-models/list

Marimo Server

• Notebooks: GET / - File browser and editor
• Apps: GET /run/<notebook> - Run notebook as web app

PINA Multi-Agent Team (marimo_flow.agents) 🧑‍🚀🧑‍🚀🧑‍🚀

Reactive multi-agent team that orchestrates PINA workflows via pydantic-graph, backed by MLflow for tracing + persistence, exposed via marimo's chat UI and optionally as A2A and AG-UI ASGI servers.

from marimo_flow.agents import lead_chat, FlowDeps
import marimo as mo

deps = FlowDeps(mlflow_tracking_uri="sqlite:///mlruns.db")
chat = mo.ui.chat(lead_chat(deps=deps))
chat

Roles (each loads its .claude/Skills/<name>/SKILL.md as instructions=):

• notebook — marimo MCP cell ops (skills: marimo, marimo-pair)
• problem — defines a PINA Problem from an open spec (skill: pina)
• model — designs a neural architecture for the problem (skill: pina)
• solver — wires Solver + Trainer config (skill: pina)
• mlflow — MLflow MCP tracking + registry (skill: mlflow)

A RouteNode classifier dispatches between sub-nodes; the lead agent wraps the graph as a single tool so the same backend powers marimo chat, A2A, and AG-UI.

Models: Ollama Cloud at http://localhost:11434/v1 (:cloud-suffixed tags). Defaults in marimo_flow.agents.deps.DEFAULT_MODELS.

Standalone servers:

uv run python -m marimo_flow.agents.server.a2a    # A2A on :8000
uv run python -m marimo_flow.agents.server.ag_ui  # AG-UI on :8001

See examples/lab.py for the full demo notebook.

Contributing 🤝

We welcome contributions! Please see our Contributing Guidelines for details on:

• Development setup and workflow
• Code standards and style guide
• Testing requirements
• Pull request process

Quick Start for Contributors:

1. Fork the repository
2. Create a feature branch: git checkout -b feature-name
3. Make your changes following the coding standards
4. Test your changes: uv run pytest
5. Submit a pull request

See CONTRIBUTING.md for comprehensive guidelines.

Changelog 📋

See CHANGELOG.md for a detailed version history and release notes.

Current Version: 0.2.0

License 📄

This project is licensed under the MIT License - see the LICENSE file for details.

---

Built with ❤️ using Marimo and MLflow