Snag Capture Architecture
Explore this system
Step through curated paths without changing the source diagram.
Beat
Diagram guide
Explore this system
Inspecting compiled semantics
E ExportT ThemeS Style0 Reset+ Zoom in- Zoom outEsc Close
Find a node
/
No matching nodes
Route probe
Choose a start node
Pick two semantic nodes on the diagram
Semantic lens
Compare system roles
Choose up to two semantic kinds. One reveals its real traffic; two compare only direct authored relationships.
Choose a kind to inspect its nodes and touching relationships.
Semantic radar
Building overview
Semantic radar needs more MAP space.
How Snag captures and keeps content
Snag is a local-first content capture system that grabs social content, classifies it, and makes it reusable.
The architecture:
- Capture sources pull from the share sheet and the browser.
- The Snag API ingests content and applies PATCH updates locally.
- An in-process queue holds background jobs.
- A classifier stages each capture by impact and effort (proposed, not final).
- A SQLite store keeps captures and decisions; Litestream replicates the WAL for durability.
- Faster-whisper transcribes audio locally.
- Durable copies go to S3/B2.
- A viewer lets the user review and correct, so the human sees why each item matters.
It is a working example of Chris's local-first AI tooling, in development now: 9 captures across 5 sources so far.
Capture
- • Anything saved arrives through the API
- • Audio is transcribed locally, no cloud send
Decision
- • Classifier proposes stage, impact, and effort
- • No calibrated confidence score, user accepts or corrects
Action
- • The capture app that turns anything you save into your next action
- • Durability via Litestream to S3 or B2