AppendixΒΆ
Operational material that complements the architecture and reference sections: how to install PROTEA, how to configure it, recipes for common tasks, the script to reproduce the thesis results, and the on-call runbook.
- PROTEA stack
The eight repositories that make up the PROTEA stack: the platform, the contracts package, the inference layer, source / runner / backend plugins, the LightGBM lab, and the cafaeval fork. Read this first when you need to jump between repositories.
- Installation and Quickstart
Bring up the full stack from a fresh checkout: dependencies,
manage.sh, the eleven process roles, and a ten-minute end-to-end smoke test.- Configuration Reference
protea/config/system.yamlreference, environment-variable overrides, and the per-environment knobs that change behaviour without code edits.- How-to Guides
Task-oriented recipes (load an ontology, ingest GOA, upload a FASTA query set, compute embeddings, predict GO terms, scale workers) with the exact
curlandmanage.shcommands. Read this when you have one specific thing to accomplish.- Reproduction guide
The full, ordered procedure that regenerates every figure and table in the Results chapter from a clean database. Read this when you want to reproduce the thesis evaluation end-to-end.
- Operational Runbook
On-call procedures: diagnosing stuck jobs, draining queues, recovering after a worker crash, restoring from backup.
- Monitoring
Visitor analytics middleware, the optional Grafana stack with auto- provisioned dashboard, and the privacy design behind the rotating daily salt.
- Secrets management (sops + age)
sops + age workflow for encrypted secret material: one-time install and key generation, day-to-day editing and decryption, CI integration, and recipient rotation.