Engineering notes from the team behind Graphwake. We write about what happens when one team changes an API that other teams depend on: OpenAPI breaking changes, checks in CI, telling the right consumer teams before a merge, and keeping track of who calls which endpoint. Practical guides with real config and SQL, plus a few opinions. Graphwake maps your API consumers and alerts only the teams a change will break. Try it at graphwake.com.
GitHub Actions and GitLab CI get a ready-made action and template from us. Azure DevOps, Bitbucket Pipelines and Jenkins get a snippet you paste into the pipeline, so here are those three, copy-paste
Usually not. Your AI assistant knows the code and files it can open, and the teams that call your endpoint keep their code somewhere else. So when you ask it whether a change is safe, it answers from
Cross-team integrations break at the seams. Each service passes its own tests, and then the two stop agreeing. Contract testing is the usual answer, and most rollouts stall before every consumer has w
Before a team changes an endpoint, someone has to answer a plain question: who calls this? In most companies the answer lives in people's heads, in an old wiki page, or in a Slack thread that starts a
Row-level security (RLS) is a PostgreSQL feature that filters the rows a query can see and write through a policy attached to the table. In a multi-tenant SaaS with one shared database, every tenant-o
An orphaned API endpoint is a live endpoint that no team has said it depends on. A second gap usually travels with it: an endpoint with no owner, so nobody answers for it. The first means a change the
A breaking API change is a change to an API's contract that can make an existing client fail, like a removed response field or a new required parameter. Talking about them in the abstract only gets yo
We ship a small open-source Go binary that sends a service's OpenAPI spec from a CI pipeline to Graphwake. The same step runs in GitHub Actions, GitLab CI, Azure DevOps, Bitbucket Pipelines and Jenkin
A risk number on a pull request is only useful if the reviewer can tell where it came from. When we added an impact score to Graphwake, we wanted one you could recompute on a napkin. So every change s
