Tiny Robot Lab
Learn AI by building it.
A hands-on, non-commercial learning lab designed to explore how autonomous agents, dynamic orchestration, grounded research, runtime observability, and Clean Architecture work in practice. No black boxes, no marketing hype—just working code and honest engineering tradeoffs.
Why This Exists
Hands-On Engineering Over Abstract Descriptions
Modern autonomous agents are often described with commercial hype or buried inside monolithic frameworks. Tiny Robot Lab was created to deconstruct these concepts by implementing them from scratch: asynchronous messaging queues, dynamic capability registration, heartbeat liveness policies, evidence-conditioned synthesis, and provider-neutral Clean Architecture boundaries.
Explore The Lab
What Would You Like to Learn?
Dive into dedicated learning sections covering our runtime mechanics, architectural boundaries, real engineering lessons, and roadmap.
How It Works
Trace the end-to-end task lifecycle: capability determination, dynamic candidate discovery, availability evaluation, and multi-stage research evidence processing.
Read Runtime FlowArchitecture
Inspect our full system architecture SVG and learn how Clean Architecture layer boundaries keep domain rules independent from Redis, Docker, and AI APIs.
Explore System DesignLearning Philosophy
Our 6-step loop (Build, Observe, Break, Understand, Improve, Share) and honest lessons from production defects—including why 400+ tests don't guarantee correctness.
View Engineering LessonsRoadmap & Limits
A factual breakdown of current v0.3.0 capabilities, honest evaluation caveats, and our planned v0.4.0 experiments in resilient multi-provider inference.
See Current StateLive Platform
Current Lab at a Glance
Everything documented here is running code backed by 401 passing tests in the repository.
Dynamic Orchestration
Runtime capability discovery and advertised queue targets replace hardcoded execution topology.
Specialist Microservices
General Worker and Research Agent running independently with dedicated queues and state streams.
Comprehensive Suite
Unit, contract, use case, concurrency, and UI tests running deterministically without heavy external services.
Prototype Console
Inspect agent readiness, live event logs, routing decisions, and attributed findings in real time.