Runtime
RustTyped orchestration core and shared domain models.- TokioAsync task runtime, cancellation and backpressure.
- ReqwestHTTP client layer over rustls for external services.
Back to registry01 02 03 04
main60835daWorking
ForgeFlow
Resilient async pipeline orchestrator for external services.
A pure-Rust orchestration core for high-throughput automation: Tokio tasks, Reqwest over rustls, Governor rate limits, Backoff policies, DashMap state and Prometheus metrics.
Private repository
Stack
Resilience
- GovernorRate-limit policies attached to pipeline edges.
- BackoffExplicit retry windows for transient failures.
- DashMapConcurrent state for job coordination.
Observability
PrometheusThroughput, latency and failure class metrics.- TracingStructured spans for pipeline debugging.
- WebSocketsLive console target for pipeline events.
Why I built this
External services fail in boring ways: timeouts, rate limits, temporary bans and partial responses.
ForgeFlow explores how to make those workflows explicit, observable and recoverable instead of hiding them in scripts.
How it works
Jobs move through bounded async stages with rate-limiters and retry policies attached to the edges.
State is kept in concurrent maps and operational visibility comes from structured tracing plus Prometheus exporters.
Key decisions
- 01
Tokio-first design
The orchestration model maps naturally to tasks, cancellation and backpressure.
- 02
Policy objects for retries
Retries and rate limits are explicit concerns, not scattered sleep calls.
- 03
Metrics as a product feature
Throughput, failure classes and latency need to be visible before automation can be trusted.
What I learned
- Resilience is mostly about making failure modes first-class.
- A good orchestrator needs boring controls before impressive throughput numbers.