Hey HN,

Let me introduce "fault". A Rust-CLI to help engineers rapidly explore their reliability as close to the code as possible. Hopefully with a bit of fun while doing it.

Its features at glance:

- Native networking faults: latency, bandwidth, packet loss...

- Native support to change DNS or LLM responses

- OpenAPI scenario generator and runner with basic load test support builtin

- Easy command to inject itself into GCP, AWS or Kubernetes against a running application

- eBPF optional support for seamless interception

- LLM-based review of results with a system prompt leaning on SRE-analysis

- MCP server for AI-agents

- Fully Open Source and written in Rust

The premises are simple: a TCP proxying that you can configure to play on inbound and outbound streams.

For instance, injecting 300ms latency:

  $ fault run \
     --proxy "9090=127.0.0.1:7070" \
     --with-latency --latency-mean 300

Point your client to it and see how it behaves.

This is a small tool that is handy to have under your belt I believe. A bit of low-friction Chaos Engineering.

Would love to hear how you guys handle reliability so I know how to keep improving it.