Inline datasets: CSV carried in the dashboard YAML (0.11.0)

[dankor]

A chart's dataset: name now resolves two ways — a managed dataset in the store, or an inline one defined in the dashboard's own optional datasets: block:

datasets:
  order_data: |
    id,status,amount
    1,paid,42

The point is prototyping without leaving the editor: invent a table, chart it, iterate, with no upload step and nothing to clean up. A dashboard that carries its own data is also a single file you can hand to someone, and the assistant's prompt now teaches the block, so "mock me up some sales data and chart it" is one turn.

The block holds CSV and nothing else. A path or a header-only line is rejected with a message saying so, because either would otherwise parse as a valid one-column, zero-row table and fail much later complaining about a missing column.

An inline name overrides a managed one: the block is part of the file being rendered, so quietly reading someone else's stored data would be the surprising choice. Anything it doesn't define falls through to the store as before.

Each block converts to Parquet once and is cached by content checksum in the temp dir, so an unchanged block costs a stat, an edit lands on a fresh path, and two dashboards with the same sample data share a file. The write is a rename, which is atomic, so concurrent renders can't read a half-written file.

Also:

• Dashboard.dataset_names() reports only managed names, so a chart on inline data doesn't pin a stored dataset against the portal's delete-guard.
• The edit modal's column dropdowns resolve inline datasets, which would otherwise be empty for such a chart.
• The starter dashboard is a file, not a 440-line string in app.py. It lives in demo/dashboards/, is read at import, and docker-compose maps that folder in as the demo path — so what you open in the browser is the file in git, nothing is copied or seeded, and _seed_demo_path is gone. The path is hardcoded and the file required: missing it is a broken checkout, so the import fails and says so. compose mounts the folder at both /app/demo (what the app reads) and /paths/demo (the path you browse), so an edit shows up in both without a rebuild; a test asserts any service mounting the source mounts the demo folder too, which is the drift that broke the container.
• The starter dashboard carries its own data, and files/ is gone entirely. The sample now lives in the dashboard that uses it, so the Parquet (a derived binary in git), the CSV, the generator that made them, the Dockerfile COPY and the compose bind-mounts were all machinery around a file nothing read any more. The orders dataset is no longer seeded at startup in either mode.
• The docs show YAML rather than Python. The examples in architecture.md and the README passed a .csv path, which has not worked since datasets moved to Parquet — scan() calls scan_parquet, so every one of those snippets failed on its first line. Rather than correct paths in Python that duplicated the YAML anyway, the product-facing examples are now the YAML a user actually writes. The README's "Use it as a library" section is dropped for now; the contributor guides (SKILL.md, PARAM_SKILL.md) keep their Python, since they show how to write chart and widget code.
• Two dashboard examples in the docs did not parse: each placed one chart twice, which is only legal as a contiguous vertical span. Found by rendering every complete example in the docs rather than reading them; all three now parse and render. Stale claims went with them — the README and compose no longer promise a seeded orders dataset, which nothing creates any more.
• The route tests carry their own data too. They had leaned on that seeded dataset, which meant they were passing off a leftover file on the developer's disk and would have failed on a clean checkout; the suite now runs with no dataset store present at all.

Minor bump: datasets: is optional, so existing dashboards are untouched.

523 tests pass.