Playbooks

Screen-by-screen walkthroughs of Annie’s main flows, so the expected flow is clear without running the program.

Screens live in docs/playbooks/_screens/. Most are real screenshots, captured against the bundled [Example] CMU-MOSEI Mini config — so you can reproduce them by picking that config on the Dataset tab. The Segment-review screens are still SVG mockups, because no bundled example carries a segmentation source yet.

Regenerating the screenshots

The captures are scripted, so they can be refreshed whenever a screen changes instead of drifting out of date:

$ make install-docs   # once — brings in Playwright and Pillow
$ make screenshots

That boots Annie against the example config, drives it with Playwright, and rewrites the PNGs in place; commit whatever changes. The script serves the app from a copy of the repo at a neutral path and uses a throwaway ANNIE_HOME, so no personal directory layout or review state can end up in a published image. See scripts/screenshot_docs.py.

• Playbook — Dataset, Browse & Curation
• Playbook — Segment review (accept / drop clips)

The flows

• Dataset, Browse & Curation — define a dataset from data sources, reshape label columns into better filter facets, select videos in the read-only Browse viewer, then curate them (like/dislike/note) in the Annotator.

• Segment review — declare a segmentation CSV, then accept or drop each per-clip segment of a long video, comparing competing start/end bands, and export the kept and discarded sets as two files.