# OpenAPI snapshot The **machine contract**. The server generates it with NSwag; this folder holds the published snapshot so the client can generate or verify types without booting the backend. > Last verified: 2026-07-29 against commit `c99e3f4` (server code last changed in `5885280`, 2026-07-28). | | | | --- | --- | | File | [`swagger.v1.json`](swagger.v1.json) | | Taken | 2026-07-29 | | Commit | `c99e3f4` | | **Paths** | **178** | | **Operations** | **186** | | Component schemas | 339 | | Generator | NSwag v14.7.1.0 · OpenAPI 3.0.0 | | Size | 612 K | ## How to regenerate The API binds **plain HTTP** on port 5002 (`launchSettings.json`), despite what most prose in this repo says — see contradiction C-3. Use whatever port you bind; the document is the same. ```bash cd server ASPNETCORE_ENVIRONMENT=Development dotnet run \ --project src/API/Baya.Web.Api/Baya.Web.Api.csproj # from another shell, once "Now listening" appears: curl -s --noproxy '*' http://127.0.0.1:5002/swagger/v1/swagger.json \ -o docs/integration/openapi/swagger.v1.json ``` Two things that cost time the first run: - Booting in `Development` **migrates and seeds** against the DB in `appsettings.Development.json` — currently a *remote* SQL Server. First boot takes ~40 s and logs `Demo world already seeded — no-op.` when the world is present. - `--noproxy '*'` matters. With a proxy configured in the environment, `curl` to `localhost` returns **502** rather than the document. Then update the table above — date, commit, and endpoint count — in the same change. A snapshot whose provenance is unrecorded is what this chain exists to stop. ## Scope `v1` only. [The old README](../../../dev/contracts/openapi/README.md) claimed the server publishes documents `v1` **and** `v1.1`, but only `v1` has ever been committed and `v1.1` was not fetched during the phase 0 survey. `UNVERIFIED:` check `/swagger/v1.1/swagger.json` when the server is next up (contradiction C-9). ## The human contract The prose half — one file per domain — lands in [`../domains/`](../index.md) in phase 2. **The two must agree**: this JSON is the wire truth, the markdown explains it. When they disagree, the JSON wins and the markdown is wrong.