VibeSim
Living telemetry for apps under construction. Describe a world in one sentence, and VibeSim runs it as a stateful simulation that streams realistic, story-driven data over SSE, WebSocket, MQTT, Prometheus or OpenTelemetry. Once compiled, it runs fully offline.
pip install vibesim
vibesim run "5 delivery vans in rainy Bristol"
simulation 0867f0da-… (logistics, 5 entities)
console http://127.0.0.1:8000/?sim=0867f0da-…
stream http://127.0.0.1:8000/simulations/0867f0da-…/stream
websocket ws://127.0.0.1:8000/simulations/0867f0da-…/ws
Mock data is usually static JSON or random numbers. VibeSim gives a dashboard something worth showing:
- Five built-in domains: delivery fleets on real road routes, microservices that leak memory and restart, smart buildings with thermostats, online stores that sell out, and markets with order books.
- Any other world in plain YAML: drifting, mean-reverting, cyclic and derived fields, plus status state machines. Or ask Claude to design one (
--compiler claude).
- Chaos on demand:
"sudden network drop on half the vans for 30 seconds", ramps, latency and jitter, scripted incident timelines, seeded random incidents.
- Time control: compressed days, fast-forward, snapshots that survive restarts, and recordings that replay event-for-event.
- Fits your stack: a browser client, MQTT, Prometheus
/metrics, OpenTelemetry, a Python library and a pytest fixture.
from vibesim import Engine, compile_prompt, Anomaly
engine = Engine(compile_prompt("3 services with a memory leak", seed=1))
engine.advance(600)
engine.inject(Anomaly(name="outage", drop_rate=1.0, duration_ticks=30))
Docs: https://jfrusher.github.io/VibeSim/: the quickstart, describing a world, chaos, integrations and the config reference.
Examples: examples/ has a scenario per domain, a React dashboard, pytest usage, and Prometheus and MQTT recipes.
Development
pip install -e ".[dev]"
ruff check . && ruff format --check . && mypy && pytest && mkdocs build --strict
MIT licensed. See CONTRIBUTING.md, CHANGELOG.md and ROADMAP.md.
