Repository navigation
Add fieldtwin to tools and libraries - #2537
nickgeorgeseo wants to merge 1 commit into
Conversation
|
@nickgeorgeseo is attempting to deploy a commit to the The GraphQL Foundation Team on Vercel. A member of the Team first needs to authorize it. |
|
|
|
Hi @nickgeorgeseo 👋🏻 thanks for submitting this PR! The team finds your app really interesting and at the same time, we're not quite sure how it fits into the GraphQL tooling picture, as the use case seems constrained to a single service. However, we're also fascinated at your case studies and want to start a conversation about whether there's a way to tell the story of how GraphQL is being used at these companies and in Fieldtwin in a way that is mutually beneficial. Are you open to a quick video call sometime? |
|
Thanks @bignimbus, glad the team finds it interesting, and that makes sense: today it’s built around one vendor’s API. I’d be happy to talk. The story I can tell is what testing an integration against a real company’s GraphQL API looks like (Jobber’s query-cost throttling, webhooks, and keeping it in sync with QuickBooks), and how a stateful twin of the schema makes that testable. I’m on US Eastern time; you can reach me at nick@nickgeorgeai.com to set up a time. |
Description
This PR adds fieldtwin to the tools on the Code page, as
src/code/tools/fieldtwin/general/fieldtwin.md, with the same frontmatter and structure as the other general tools (Microcks, Schemathesis).fieldtwin is an open-source (Apache-2.0), local, stateful twin of the Jobber GraphQL API, seeded with a fake trades business, for testing integrations and AI agents without a real account. It models the parts of Jobber's public schema a client depends on (Relay connections,
EncodedIdscalars,userErrors, and the documented query cost and throttling) and adds checkpoints, rewind, and faults on demand. The entry includes a few paragraphs on usage, as the contributing guide asks.pnpm lint --quietand Prettier pass on the new file.Disclosure: I'm the author of fieldtwin.