rachid chabane.
Search
← All projects Project · 01 · agent-maintained

Sterna

open source 2025–2026 Python · Django · React FR / EN
1,264 automated tests
AGPL-3.0 open source

01What it is

A multi-model AI workspace, now open source under AGPL-3.0: side-by-side model chat, a sandboxed coding agent with a GitHub issue-to-PR flow, a RAG knowledge base, MCP connectors, voice rooms, and per-message cost accounting.

02Engineering

Sterna is a full-stack AI workspace: a Django/DRF backend pairing a React 19 + TypeScript frontend with FastAPI microservices, a Docker sandbox for code execution, and Kubernetes/Terraform infrastructure. Users run several models side by side in one conversation, hand GitHub issues to a sandboxed coding agent that writes reviewed implementation plans and pull requests through an in-browser IDE, query their own documents through pgvector RAG, join live multi-agent voice rooms, and see the measured tokens, cost, and latency on every single message.

The core was finished about eight months before release; what followed was polish. Rather than fund the hosting and on-call that a public SaaS demands, I chose to open-source the whole thing under AGPL-3.0 — the code is the product, self-hostable from one compose file, with recorded demos of every feature in the repository.

The codebase itself was driven by an autonomous Claude Code task-runner enforcing per-task quality gates, gate-repair loops, and structured failure records, making the development process a case study in agentic software engineering.

03 Architecture

Frontend
React 19TypeScript
Backend
Django/DRF
Domain apps
OpenRouter multi-LLMcoding agent (issue → plan → PR)sandboxed code executionMCP connectorsAI voice rooms
RAG and billing
pgvector RAGper-message cost accountingGDPR / rate-limiting hardening
Agentic build process
Claude Code task-runnerper-task quality gatesgate-repair loopsstructured failure records
Full-stack AI workspace driven by an autonomous Claude Code task-runner

04 Highlights

  • Open-sourced under AGPL-3.0: the full Django/DRF backend, React 19 + TypeScript frontend, FastAPI microservices, and the Kubernetes/Terraform infrastructure
  • Sandboxed coding agent that turns a GitHub issue into a reviewed implementation plan and a pull request, with an in-browser Monaco IDE over its workspace
  • Every message carries measured tokens, cost, and latency — pgvector RAG, MCP connectors, live multi-agent voice rooms, and BYOK provider keys round out the workspace
  • Entire codebase driven by an autonomous Claude Code task-runner with per-task quality gates and gate-repair loops

05Stack

PythonDjangoReactOpenRouterpgvectorKubernetesCloudflare

06Status

open source

07 Links