Visual screenshot testing is an excellent control surface for directing coding agents. Using snapshots inside the coding loop can increase snapshot counts by 100x to 1000x, which is wildly unaffordable with existing centralized snapshot services such as Chromatic.
Selfie rethinks this model:
- coding models can (and should!) generate snapshots locally to validate and document their work
- this local work can (optionally) be trusted to dramatically reduce cloud snapshot usage (~100x faster and cheaper)
- agents and humans should be able to directly annotate the snapshots and their deltas, during local development and cloud review
Selfie is piloting with StorybookJS, but PRs to support other testing models are welcome.
A hosted instance runs at hackathon.selfie.dev.
We previously launched Selfie, a snapshot testing library focused on data rather than screenshots with no cloud or agent integrations. We are backgrounding it to focus on this completely new and unrelated service, which shares the same general theme and name but none of its code. Everything in this repo was built from November 1, 2026 onward for the Meta Global AI Developer Hackathon.
Pre-existing work used here:
- pgstencil (MIT, DiffPlug LLC), consumed as published npm packages to bootstrap the Postgres backend. None of its source is copied into this repo.
Every model Selfie calls at runtime is a Meta model.
- Muse Spark (Meta Model API), in the cloud service: reads the before/after snapshots together with the code diff and the stated intent of the change (commit message, PR description, or agent prompt), groups the changed snapshots by the code change that caused them, and flags changes that weren't asked for so they can be approved or rejected in batches.
- Muse Code, in the local loop: the coding agent that makes a change, runs
review, and fixes the regressions Selfie flags. Selfie itself works with any coding agent.
These parts are ordinary code, not model calls: rendering snapshots (Storybook), hashing and pixel-diffing images, finding baselines with git, and storing data (SQLite locally; in the cloud, Neon Postgres for accounts, a D1 database per project for snapshot metadata, and R2 for image bytes).
Selfie stores data in two places:
- a cloud service holding golden snapshot data, as well as runs which are pending approval before being accepted as golden
- local data stored at
~/.config/selfie/<git-origin-slugified>/, Selfie stores locally generated snapshots here, and also mirrors data from the cloud service as necessary to improve performance
Selfie has the following commands:
- find-baseline - returns the Git sha of the nearest merge-base with a complete set of goldens
- review - records snapshots (only if necessary) and shows the diff against the baseline
- image snapshots are fully opaque 8-bit RGB, and transparency is not supported
- image snapshots are always referred to by the SHA256 hash of the snapshot pixel data in the PAM format
- the db primary key of images is based on this SHA256, truncated to the safe range for positive javascript integers
- images whose truncated hash collides are stored in the same table under monotonic negative integers
- local app is built with SQLite, Typescript and Vite
- cloud server is implemented with Hono (typically Cloudflare Workers), Typescript and Vite, with user accounts in Neon Postgres (via pgstencil, connecting through Hyperdrive), snapshot metadata in a per-project Cloudflare D1 database, and image bytes in Cloudflare R2
- Storybook (MIT)
- Hono (MIT)
- Vite (MIT)
- SQLite (public domain)
- PostgreSQL (PostgreSQL License)
- pgstencil (MIT)
FSL-1.1-MIT — Copyright 2026 DiffPlug LLC