High-Performance AI & IoT Boilerplate
A production-grade backend for AI agent systems and IoT pipelines.
What you get
Complete dual-path data architecture
SQLModel (structural) + asyncpg raw SQL (high-frequency telemetry), enforced by AST-level tests — not just documentation.
IoT pipeline end-to-end, <10ms latency
MQTT → Redis Streams → asyncpg INSERT → pg_notify → SSE → UI.
LangGraph 1.x agent runtime
Stateful graphs with Redis checkpointing + Postgres fallback. pgvector RAG, Ollama local LLM, streaming SSE responses.
4 Docker Compose profiles
core / ai / iot / observability — run only what you need.
Prometheus + Grafana pre-configured
API latency, MQTT throughput, circuit breaker state, Redis stream lag — all wired on first boot.
18 Architecture Fitness Tests
AST-level enforcement of layer boundaries. Runs in 0.4s, zero external dependencies. AI-assisted development without architectural drift.
ARCHITECTURE.md — 20 sections
Failure semantics, schema evolution, resource isolation, event taxonomy, ADLs with rationale and trade-offs.
CLAUDE.md
Architecture contract for Cursor and Claude Code. Full refactoring permission with structural guarantees.
4–6 weeks of architecture decisions, setup, and debugging that most teams redo from scratch on every project.
Solo license: use for your own products and internal systems.
For client project delivery, see Agency license (€2,500).
Contact: pedro.murinelo@protonmail.com