Atomizer¶
Queue-based processor for VeriLib repositories: clone/upload to S3, run language probes, parse Schema 3.0 JSON, and persist atoms for the web app.
| Repo | Beneficial-AI-Foundation/verilib-atomizer (private) |
| Runtime | Python 3.10+, RabbitMQ consumers, Docker-out-of-Docker probes on ECS-on-EC2 |
| Status | active |
What this repo owns¶
| Owns | Does not own |
|---|---|
| Upload + atomize workers | Web UX / certify UI (frontend) |
| Language plugins + probe orchestration | Certificate probe worker (certificates) |
| S3 handoff of repo trees / probe output | CLI local atomize (verilib-cli) |
| Persist atoms / deps / statuses to MySQL | Science-team probe source repos (linked only) |
Legacy DB-polling (main.py) may still exist in older deployments; production is RabbitMQ consumers (scripts/run_upload_processor.py, scripts/run_atomize_processor.py).
Hub pages¶
- Languages and plugins
- Probe ecosystem — the four probes, the Schema 3.0 contract,
merge/enrich/summary - Data flow
- Queue workers
- JSON mapping — Schema 3.0 envelope
- Atom statuses and colours — verification-status → colour scheme
- Probe upgrading
- ECR on ECS
- Disk cleanup
Handoff from frontend¶
Pending uploads and reatomize actions publish queue messages. Private GitHub tokens arrive on the message (encrypted), not via a MySQL lookup — see queue workers and upstream docs/private-github-upload-processor.md.