Open Data Framework — CLI, UI & MCP (odf)

The CLI, UI server, MCP server, and chat surface for Open Data Framework. Depends on opendataframework for the core abstractions (Entity, Repository, Component, Service, Task, Pipeline, Layer, Context, Project) — see that package for those.

Install

pip install odf              # CLI + scaffolding only
pip install odf[ui]           # + UI dev server
pip install odf[mcp]          # + MCP server
pip install odf[chat]         # + chat window (requires [ui])

CLI

Scaffold a new project:

odf init --template default my-project

Available templates: default, data-analytics, data-engineering, data-science, research — each scaffolds a config.toml and folder layout tailored to that use case.

Run a project, optionally with the UI / MCP server / chat:

odf run
odf run --ui --mcp --chat

UI

A small FastAPI dev server (backgrounded, like any Service) that visualizes the resolved object graph — every Component/Service/ Repository, grouped by Layer, with lifecycle actions (start/stop/ execute) available from the UI. Grid positions persist across reloads. Repositories can declare a data_view() to pick how their data renders (table, map, timeseries, streaming video/audio, ...) instead of the default table.

server.start(ui=True)
print(server.ui_url)  # http://127.0.0.1:4747

Icons and colors can be extended beyond the built-in isometric set via [ui] icon-scripts / [ui.colors] in config, or by packages registering into odf.ui.extensions.

MCP Server

Exposes the same actions available in the UI — component start/stop, task/pipeline execution, log inspection — as MCP tools for any MCP-speaking client.

server.start(mcp=True)
print(server.mcp_url)  # http://127.0.0.1:4748/mcp

New components can be added to an existing project from within an AI chat: the MCP service knows the available component library and the project's current structure, generates the config delta and any needed code, and the project rebuilds.

Chat

An optional chat window added to the UI, backed by a local Ollama model. Requires ui=True; if mcp=True is also passed, the chat model gets tool-calling access to the same actions exposed as MCP tools.

server.start(ui=True, mcp=True, chat=True)

Connection parameters come from [project.chat] in config: model (default "gpt-oss") and ollama-host (default "http://localhost:11434").

Examples

See examples/ for the UI-server surface built on top of opendataframework. For the core abstractions themselves, see the opendataframework package and its own examples/.