Docs

Studio

Pipeline inspector

Register Anvia pipelines in Studio, inspect their graphs, run inputs, and compare deterministic workflow output.

Studio can run pipelines beside agents. A pipeline is ordinary TypeScript with input validation and graph metadata, so the Studio page is useful for checking workflow shape before the workflow is called from a worker, route, or job.

Run the pipeline inspector example:

pnpm cookbook:studio:07

Then open:

http://localhost:4021/ui/pipelines

Studio pipeline inspector showing a ticket triage pipeline graph and input panel.

Graph Shape

07-pipeline-inspector.ts creates a ticket triage pipeline with deterministic transforms, parallel branch analysis, and an agent prompt step. Studio renders the graph from pipeline metadata so you can inspect step names, descriptions, and connections.

Use clear step names because they become the operational language for debugging runs.

Multiple Pipelines

Run the multiple pipeline example:

pnpm cookbook:studio:08

It registers an order status pipeline and a ticket routing pipeline. Studio presents them as separate workflow targets with different metadata and sample inputs.

This is the right shape for internal operations where the same runtime exposes several deterministic workflows, not just chat agents.

Persistence

Pipeline logs and run history are in memory by default. Use the persistent store example when you need run history across restarts:

pnpm cookbook:studio:10

That example passes the same SQLite-backed store for sessions, traces, pipeline logs, and pipeline runs.

  • examples/cookbook/09_studio/07-pipeline-inspector.ts
  • examples/cookbook/09_studio/08-multiple-pipelines.ts
  • examples/cookbook/09_studio/10-persistent-store.ts