vihren

README

Vihren

Vihren is a Go toolkit for building typed Temporal workflows with less boilerplate. You write ordinary Go activities and workflows, mark them with //vihren:activity and //vihren:workflow, and vihren-gen generates the Temporal registration, workflow-safe activity proxies, and typed clients you'd otherwise hand-write.

If you're building Temporal workflows in Go, Vihren gives you:

• Code Generation — typed activity/workflow plumbing without the hand-written boilerplate.
• Large Payload Storage — a claim-check pattern that keeps large data out of Temporal's workflow history.
• Embedded Temporal — an in-process server so examples and local development need no external infrastructure.

Install

Requires Go 1.26 or newer.

go get vihren.dev/vihren
go install vihren.dev/vihren/cmd/vihren-gen@latest

Code Generation

Write the activity and workflow:

type GreetingInput struct{ Name string }
type GreetingOutput struct{ Message string }

//vihren:activity
func (a *GreetingActivities) ComposeGreeting(ctx context.Context, in GreetingInput) (GreetingOutput, error) {
	return GreetingOutput{Message: fmt.Sprintf("%s, %s", a.Prefix, in.Name)}, nil
}

//vihren:workflow
func HelloWorkflow(ctx workflow.Context, in GreetingInput) (GreetingOutput, error) {
	ctx = workflow.WithActivityOptions(ctx, workflow.ActivityOptions{StartToCloseTimeout: time.Second})
	return Activity.ComposeGreeting(ctx, in)
}

Generate the plumbing:

go generate ./...

vihren-gen writes Register (worker registration), the type-checked Activity proxy used above, and a typed NewClient for starting workflows and awaiting typed results — none of it hand-maintained.

Run it with zero infrastructure:

go run ./examples/codegenhello/cmd/codegenhello-embedded
# Hello, Ada

See examples/codegenhello for the full walkthrough.

Large Payload Storage

Large workflow data — agentic LLM context, documents, ML models — can blow past Temporal's workflow history size limits. blob solves this with a claim-check pattern: activities offload data to a backend and pass a typed handle through the workflow instead of the bytes.

//vihren:activity
func GenerateLargePrompt(ctx context.Context, input string) (blob.Value[string], error) {
	veryLargeString := "..." // e.g., fetched from a database
	return blob.Offload(ctx, veryLargeString)
}

//vihren:activity
func CallLLM(ctx context.Context, prompt blob.Value[string]) (string, error) {
	actualPrompt, err := prompt.Resolve(ctx)
	if err != nil {
		return "", err
	}
	return runInference(actualPrompt), nil
}

The workflow only ever passes prompt around; it never touches the underlying bytes. See blob and examples/conversationstorage for backend setup and a full walkthrough.

Embedded Temporal

embeddedtemporal runs a Temporal server inside your own process — no Docker, no daemon, no temporal server start-dev. It's built for examples, demos, and single-binary local development, not production deployments.

server, err := embeddedtemporal.Start()
if err != nil {
	return err
}
defer server.Close()

See docs/embedded-temporal.md.

License

Vihren is licensed under the Apache License, Version 2.0. See the NOTICE file for attribution.