Getting started with Temporal durable execution for Go and TypeScript

SDKs and development

Developers build workflows in Go or TypeScript. The Go SDK requires version 1.24 or higher. You can install the client, worker, and workflow packages using go get go.temporal.io/sdk.

Go offers more examples.

Developers choose between different SDKs based on their specific application requirements. The Go SDK provides the client, worker, and workflow modules. You must use Go version 1.24 or higher for the latest features. TypeScript 1.9.0 includes experimental workflow updates that are production ready. The Go SDK also has a longer lifetime and more examples than the TypeScript version.

TypeScript writing feels more straightforward.

If your application handles heavy CPU tasks, Go fits well. For Node.js workloads, you can run CPU-intensive activities in worker_threads or use abstractions like piscina. NodeJS processes use a single-threaded event loop model. This architecture means a worker process always binds to a single CPU core. You may encounter Out of Memory errors if you exceed the default memory limit, which reaches 2GB in some Node versions. The Go SDK provides built-in integration with the standard log package for version 1.21 or later. The workflowcheck tool can also detect non-determinism in Workflow Definitions. In Go, you can use internal/cmd/build for local development. This tool starts an embedded Temporal dev server for testing. You can use the -race flag to run tests.

The Go SDK has a smaller memory footprint.

Pricing and capacity

Temporal Cloud uses consumption-based pricing. You pay for Actions, Storage, and a monthly Plan. Actions include starting Workflows, recording a Heartbeat, or sending messages. Actions also cover Timers, Signals, Queries, and Updates.

Actions cost $50.

The Essentials plan costs the greater of $100 or 5% of your consumption. It includes 1 million Actions, 1 GB of Active Storage, and 40 GB of Retained Storage. The Business plan costs the greater of $500 or 10% of consumption. It includes 2.5 million Actions, 2.5 GB of Active Storage, and 100 GB of Retained Storage. Business plans include SAML SSO and improved support response times.

Active Storage measures usage for open Workflows at $0.042 per GBh. Retained Storage measures closed Workflow Event History at $0.00105 per GBh. Users set the retention period per Namespace between 1 and 90 days. For histories exceeding 90 days, Temporal recommends using the Export feature. The Essentials plan includes 1 million Actions.

Actions pricing starts at $50 per million and provides volume discounts. You can choose Pay-As-You-Go or Commitments. Commitments use Temporal Credits to pay for consumption and offer discounts based on volume and duration. Scheduled Workflows and Batch jobs also contribute to storage usage. You can purchase credits through AWS Marketplace or GCP Marketplace. Companies with less than $30 million in funding access the startup program for $6,000 in credits.

Orchestration and debugging

Temporal remains a code-centric engine. Camunda uses BPMN to help business and IT teams collaborate. It provides built-in task management for human-in-the-loop processes. This makes Camunda suitable for regulated industries like banking or insurance.

Orkes Conductor provides a hybrid approach. It uses Redis and Postgres and offers a visual workflow editor. It serves engineering-heavy organizations that manage microservices at Netflix-scale.

The VS Code extension for TypeScript improves debugging. It lets you load event histories. You can set breakpoints. The extension allows you to load event histories by Workflow ID or JSON file to enable the replayer to execute code deterministically while hitting sentinel breakpoints placed inside workflow or activity interceptors.

Temporal lacks a visual editor.

The plugin includes an IDE debugging UI for history upload and breakpoint management. It uses a serving layer to intercept debugging commands and coordinate with the IDE. The replayer processes the workflow’s event history deterministically.

Will the learning curve for durable execution semantics hinder your team?