Note: dashboard

Context

The dashboard feature provides a living-docs view of a registered corpus through a three-stage pipeline:

  1. Refresh — build_snapshot() runs the benchmark against the corpus and returns a snapshot dict. If queries.yaml is missing, it returns a partial snapshot that includes an error field rather than failing outright.
  2. Render — render() writes a self-contained HTML report to a file path you specify, with the snapshot embedded as JSON between the __ATTUNE_SNAPSHOT__ and __ATTUNE_TITLE__ sentinels in the template.
  3. Show — display() pretty-prints the snapshot to the terminal using Rich.

Each stage has a dedicated source file and its own CLI entry point under attune-rag dashboard.

Content

The dashboard module exposes top-level functions rather than classes. You call them directly without instantiating anything first.

Function File Purpose
build_snapshot(corpus_package, queries_path) dashboard/refresh.py Returns a snapshot dict; returns a partial dict with an error key if queries.yaml is absent
render(out, snapshot, title) dashboard/render.py Renders the dashboard HTML template to out with snapshot embedded as JSON; returns the resolved output path
display(snapshot, console) dashboard/show.py Prints the snapshot to the terminal via Rich; accepts an optional Console instance for testing or redirection

The refresh and show stages each also expose a main() function (return code 0 on success) that wires the stage into the CLI.

Two template sentinels — _SENTINEL_SNAPSHOT (__ATTUNE_SNAPSHOT__) and _SENTINEL_TITLE (__ATTUNE_TITLE__) — mark the injection points in the HTML template. They are internal constants and are not part of the public API.

Source files

Tags: dashboard, living-docs, html, terminal, snapshot, freshness