I shipped ragleap-terraform v0.1.0 8 hours ago. It's our 9th library. And it changes what RagLeap is.
RagLeap started as RAG. Now it's DevOps for ANY software. Here's the honest breakdown — what's live-verified, what's next, and what we won't claim.
ragleap-terraform 0.1.0 — kind cluster + full observability stack via Terraform, verified on a real Docker host
What it does:
A Terraform module that creates a local kind cluster and installs ragleap-observability with helm_release.
hcl
module "kind-cluster" {
source = "your-registry/ragleap-terraform"
cluster_name = "ragleap-dev"
grafana_admin_password = var.grafana_password
}
**Variables:** cluster_name, namespace, install_observability, observability_chart_path, grafana_admin_password, install_postgres_exporter (default false), helm_timeout, kubeconfig_path
**What's inside:** examples/kind ships inside the wheel. pip install ragleap-terraform and you have it.
**Verified — Not Mocked:**
Applied and destroyed on a real Docker Desktop host with Terraform 1.9.8. All chart pods Ready. Your kubectl context untouched. This is why I say "live-verified".
**Known Limitations (honest):**
Local kind only. No AWS/GCP/Azure modules yet — those need a real cloud account to verify, not just code.
postgres-exporter off by default — needs ragleap-ops and a Secret no chart creates yet.
State and kubeconfig hold credentials in plain text. Do not interrupt apply/destroy.
This is v0.1.0 for a reason. It works, it's verified, and we're transparent about gaps.
**
## ✅ What's Done & Live-Verified So Far
**
**1. ragleap-ops 0.4.1**
Helm chart, SecurityContext, read-only root filesystem (not neo4j), zero-permission ServiceAccounts, NetworkPolicies, Ingress + TLS, multi-env values, Postgres and neo4j backup/restore with RTO, runbooks drilled against broken clusters.
This wasn't helm install and hope. We broke clusters on purpose and drilled restore.
**2. ragleap-app-chart 0.2.1**
Generic Helm chart for any developer's services. If you have a service, this deploys it with the same security standards as ragleap-ops. This is the "ANY software" part.
**3. ragleap-observability 0.6.0**
Prometheus v3, Grafana 12, Loki 3.6, Promtail, AlertManager, Slack and email receivers (delivery proven against a stand-in), pods-only RBAC, configurable Grafana password, Trivy scans of all images.
We started at 383 findings. Down to 14 on four images. postgres-exporter still at 48, Promtail at 86 — next bump cycle will fix.
**4. Release Process**
PyPI keywords, TestPyPI → PyPI, tags, GitHub Releases, docs surfaces. Boring but critical.
**
## 🔜 What's Next — Ordered by Level
**
I organize everything by solo → team → corporate. You only need what you need.
#**## SOLO DEVELOPER (Next 2 weeks):**
* Finish Loki retention → ship 0.6.1
* Real Slack/email send — right now only stand-in is proven
* Automate postgres-exporter Secret + DB role — no chart does it today, it's manual
* ragleap-ansible for VPS bootstrap (never started)
* Image bumps
**
## TEAM (Next month — needs your sign-off):
**
* HPA and PodDisruptionBudgets in ragleap-ops + app-chart
* Helm chart testing in CI (ct) + scheduled Trivy job — both edit CI files
* GitOps (ArgoCD or Flux) — this is the blocker for canary and blue-green
* SLO and SLI dashboards
* ServiceAccounts, RBAC and PDBs for app-chart
**
## CORPORATE (Later — needs cloud accounts & audits):
**
* Cloud Terraform modules (AWS, GCP, Azure) — they'd be unverified without an account, so I won't fake it
* Multi-AZ, multi-node, cluster autoscaling (blocked — kind is one node)
* Ingress rate limiting + WAF, audit logging, image SBOM + provenance
* Tracing, SSO for Grafana, chaos testing
* Compliance (SOC 2) — needs external audit, not just engineering
## Suggested Order:
Loki retention → 0.6.1 → HPA/PDB → Ansible → Trivy/ct in CI → GitOps
**
## 🎯 The Honest Claim
**
This is important. Even with every item done, the accurate claim is:
**
## "enterprise-grade foundations, live-verified"
**
It is NOT "proven at global scale". Everything has run on one node, with single-fault tests.
I'm a solo dev building in public from Chennai. I won't claim what I haven't tested. Many startups claim "enterprise-ready" after one Helm chart. I drilled backup/restore against broken clusters before claiming RTO.
**
## 📦 Where We Are
**
46 AI Employees — modular RAG & self-thinking intelligent agents
9 Libraries — 8 Python + 1 Java
45k+ Downloads
MIT Licensed, Self-Hosted, BYOK — 17 LLMs + Ollama local
45+ Vector DBs
Links:
Core: github.com/antonyrag/ragleap-core (17 stars, help us reach 50?)
Docs: docs.ragleap.com
Packages: packages.ragleap.com
PyPI: pypi.org/project/ragleap-terraform
If you want DevOps that doesn't hide behind marketing, try:
pip install ragleap-terraform
Feedback welcome. Issues welcome. Stars appreciated.
Building in public. Shipping truth.
— Antony
source & further reading
dev.to — original article
Stop managing six AI vendor accounts: one key for 109 models
AI-assisted coding interviews test a different skill. Here's how to practice it
Run Claude Code on a GitHub Actions cron without installing the GitHub App