Examples

Small, focused odf projects demonstrating the CLI/MCP/UI-server surface that sits on top of opendataframework. These depend on opendataframework as a real package dependency, so this directory doubles as an integration check that the split works for external consumers.

#	Example	Concept(s)	Docs
1	01-table-view	A repository implementing no data_view() at all, so the UI falls back to the implicit default plain table.	View
2	02-data-analytics	The data-analytics template layout: a SummarizeReadings @Analytics @Task writing a JSON report to reports/.	Task, Layer
3	03-data-science	The data-science template layout: data/raw/ ingested via Project, then explored from notebooks/ through a Repository built standalone, outside the Context.	Repository
4	04-data-engineering	The data-engineering template layout: SeedReadings + ExportReadingsSummary @Tasks composed by a Pipeline into a raw-to-processed ETL flow. Its [ui.topology] config hides the Config and SQLite nodes.	Task, Pipeline, Topology view config
5	05-research	The research template layout: a single RunExperiment @Task loading data/ into a repository and writing results/summary.json, alongside a doc/notes.md lab-notebook stub.	Task
6	06-location-view	A repository implementing data_view() -> LocationView so the UI renders a map instead of a table.	View
7	07-streaming-video-view	A stream-only Webcam repository (data_view() -> StreamingVideoView) rendered as a live video player with a Start/Stop toggle.	View
8	08-video-view	A SecurityClips repository (data_view() -> VideoView) rendered as a grid of video cards, each previewing a frame of its clip and opening it in a player.	View
9	09-streaming-audio-view	A stream-only DispatchRadio repository (data_view() -> StreamingAudioView) rendered as a live audio player.	View
10	10-audio-view	A VoiceMemos repository (data_view() -> AudioView) rendered as a grid of audio cards, each showing its clip's waveform and opening it in a player.	View
11	11-image-view	A DeliveryPhotos repository (data_view() -> ImageView) rendered as a grid of thumbnails that open full size, where labelled boxes can be drawn on them.	View
12	12-document-view	An OrderReceipts repository (data_view() -> DocumentView) rendered as cards that open each record's document: a JSON tree, text, or a PDF. File receipts are named and typed by content_metadata().	View
13	13-timeseries-view	An Orders repository (data_view() -> TimeseriesView) rendered as a line chart, one point per record, where time ranges and points can be labelled.	View
14	14-chart	A SalesByStore component implementing chart() -> str, a matplotlib-backed self-contained HTML chart the UI renders behind a View Chart button.	Component
15	15-chart-plotly	The same ChartProtocol contract as 14-chart, backed by Plotly instead of matplotlib for real hover tooltips.	Component
16	16-mcp-chat	odf.server.Server.start(ui=True, mcp=True, chat=True) — the same resolved Context exposed as MCP tools and driven by a local Ollama chat window.	Chat
17	17-custom-icon	A config.toml-registered .js file ([ui] icon-scripts) and [ui.colors] entries adding a "Lighthouse" icon and named color swatches to the topology UI's pickers — pre-applied to the Beacons node via a committed layout.json — plus a custom favicon, header logo, and brand name ([ui] favicon / [ui] logo / [ui] brand).	UI
18	18-custom-mcp-tool	A StockRoom component whose mcp_tools() adds a typed stock-room.low_stock tool next to the fixed MCP tools.	MCP Server
19	19-agent	A Librarian @Agent whose invoke(question, limit=3) takes arguments: called through context.call(), invoked from a form in the UI, and exposed as its own typed agent.librarian.invoke MCP tool.	MCP Server, Chat
20	20-function	Adopting plain functions with @Function: to_eur and store_total keep their signatures and still work when called directly, while context.call() injects their Rates/Orders dependencies and another function through Annotated. Each one gets a Call action in the UI and its own typed <key>.call MCP tool.	MCP Server, Chat

Examples 2–5 mirror odf init --template <name>'s scaffolded layouts — each is runnable both as a plain script (python main.py) and via odf run (see each example's own README.md for specifics). Most "Docs" entries above refer to concept pages on the opendataframework docs site — Entity, Repository, Task, Pipeline, Layer, View, and Component are all core abstractions documented there, not here. The UI, MCP server and chat each have their own page here: UI, MCP Server, Chat.

Running an example

Clone the repository, install odf with the UI (pip install "odf[ui]"; some examples need more, such as opencv-python or matplotlib, listed in their own README.md), then run from inside the example's own directory:

cd examples/01-table-viewodf runUI running at http://127.0.0.1:4747

Open the UI and look at Books — it renders as a plain table with no extra step needed. Examples 2–5 can also run as a plain script instead of the UI (python main.py) — see each example's own README.md for specifics.