AI Chat Platform — Infrastructure & Kubernetes
Full Stack
Backend
Tech Stack
Terraform
Ansible
Kubernetes
Docker
GitHub Actions
CI/CD
Linux
Git
Description
Infrastructure repository for deploying the multi-repo AI chat application on Grid'5000 (Kubernetes on bare metal). Covers the full path from cluster provisioning to production-style workloads with ingress, secrets, and observability hooks.
Built to demonstrate system design, async processing, caching, message queues, and infrastructure as code — the same platform described across chat-app-frontend, chat-app-gateway, chat-app-worker, and this repo.
- Terraform + Grid'5000 provider: OAR reservation (oarsub -t deploy), Kadeploy, and RKE cluster creation.
- Ansible: NVIDIA container toolkit, GPU device plugin, kubectl apply for manifests.
- Kubernetes (Kustomize): gateway, worker, frontend, PostgreSQL, Redis, RabbitMQ, Ollama with ingress /api rewrite.
- Docker Hub images (soufian1/chat-app-*): built and pushed via GitHub Actions before deploy.
- CPU and GPU profiles (chiclet vs chifflot) for integration tests vs Ollama on NVIDIA P100/V100.
Page Info
Kubernetes workloads
Kustomize manifests for gateway, worker, frontend, PostgreSQL, Redis, RabbitMQ, and GPU-backed Ollama in the chat-app namespace.
Grid'5000 provisioning
Terraform reserves nodes via OAR, deploys images with Kadeploy, and bootstraps an RKE cluster before Ansible configures the runtime.