AFIE — Autonomous FinOps Intelligence Engine for Kubernetes. A closed-loop system that scrapes cluster telemetry, learns a resource-adjustment policy with reinforcement learning, validates each recommendation against a policy engine, and applies changes through GitOps pull requests.
This directory contains the engineering documentation. The IEEE paper drafts
live under ../paper/; the phase-by-phase implementation workflow lives in
AFIE_Workflow_updated.docx at the repo owner's desktop.
| Doc | What it covers |
|---|---|
| architecture.md | System overview, components, data flow, tech stack |
| telemetry-service.md | Phase 3 service — Prometheus scraping, MetricEvent JSONL sink |
| feature-engineering-service.md | Phase 4 service — sliding window, 47-dim state vector, /state endpoint, Postgres persistence |
| infrastructure.md | KIND cluster, Prometheus/Grafana, ArgoCD, Postgres |
| development.md | Build, test, and run locally |
| roadmap.md | Planned scope for Phases 5–9 |
| doc.md | Local dev credentials cheatsheet |
- Phase 1 (env + monorepo): complete
- Phase 2 (KIND + Prometheus/Grafana + ArgoCD): complete
- Phase 3 (telemetry pipeline): complete — see telemetry-service.md
- Phase 4 (feature engineering + Postgres): complete — see feature-engineering-service.md
- Phases 5–9: not started — see roadmap.md
src/
api/
telemetry/ Phase 3 — ASP.NET Core scraper (built)
feature-engineering/ Phase 4 — 47-dim state vector + Postgres (built)
bff/ Phase 7 — scaffold only
operator/ Phase 6 — scaffold only
ml/ Phase 5 — scaffold only
dashboard/ Phase 7 — scaffold only
shared/
AFIE.Contracts/ Shared DTOs (MetricEvent)
infra/
gitops/ ArgoCD apps, CRDs, plain manifests
terraform/ Phase 8 — Azure IaC (scaffold)
benchmarks/ Phase 9 — experiment workloads
tests/
AFIE.Telemetry.Tests/ xUnit tests for Phase 3
AFIE.FeatureEngineering.Tests/ xUnit tests for Phase 4
AFIE.Operator.Tests/ Phase 6 — scaffold only
experiments/ Local telemetry sinks + Phase 9 analysis
scripts/ Phase 2 bootstrap
paper/ IEEE draft