Rasputin

Runtime recorder

Optional. Adds arguments, returns, and call history to errors. Error reporting works without it.

How it works

During a request, Rasputin can record what your functions received and returned. That record is attached when the request fails.

The @rasputin-ai/sveltekit/vite plugin instruments client and SSR modules at build time.

The recorder needs a long-running Node process with adapter-node. It is not built for adapter-vercel, adapter-netlify, Cloudflare, or other serverless and edge adapters.

What you need to set up

If you followed Installation, both pieces are already in place:

1. Instrument functions

The @rasputin-ai/sveltekit/vite plugin instruments client and SSR modules.

2. Scope an execution

rasputinServerHooks creates one execution per HTTP request. Recorded state is only kept inside that request and attached if it throws.

Executions tab

The project Executions tab lists observed HTTP routes and named jobs from SDKs that report a catalog. You can capture a bounded set of complete recordings, including successes, without changing how errors are reported.

Where to go next

  1. Check auto-instrumentation support for adapter-node and your Node runtime.
  2. Follow Instrumentation & executions for non-HTTP work and recorder tuning.