How qiskit/qiskit-serverless works
Project type and purpose
• This repository implements “Qiskit Serverless,” a full-stack, distributed quantum-classical execution platform.
• It consists of three modules:
• A Python “client” library for users to define and submit Qiskit Functions.
• A “gateway” service (Django/ASGI) exposing REST APIs, handling authentication, job metadata, and scheduling.
• Kubernetes/Helm “charts” to deploy gateway, Ray compute cluster, database, storage, and supporting services.
File structure
• Top-level directories:
• client/ – Python SDK and CLI
• gateway/ – Django-based API server and job scheduler scripts
• charts/ – Helm chart definitions for Kubernetes deployment
• Other orchestrations:
• Dockerfile-ray-node, docker-compose*.yaml, Makefile — for local Docker-based setups
• .github/workflows – CI/CD pipelines for building client, gateway, Helm chart and running tests
• gateway/api: MVC-style organization with models, serializers, views, services, use-cases, and repositories
• charts/qiskit-serverless: Helm templates for gateway, Ray head/worker, Postgres, PVCs, ingress, hpa, secrets
Key insights from README
• Purpose: let users submit long-running Qiskit Functions that automatically scale across CPUs, GPUs, QPUs
• Quickstart:
• Docker Compose for local dev (components: gateway, ray-head, scheduler, postgres)
• Kubernetes via Helm/Kind for production-like deployment