I think one of my dev deployments got wedged in the HTTP actions layer.
Project: autotcg
Broken dev deployment: healthy-guineapig-155
Working replacement dev deployment: rightful-partridge-886
Symptoms on healthy-guineapig-155:
- Every HTTP action returned generic 500:
{"code":"InternalServerError","message":"Your request couldn't be completed. Try again later."} - This happened for all Better Auth endpoints under /api/auth/*
- I then replaced convex/http.ts with a single bare route:
GET /_debug-auth -> Response.json({ ok: true }) - function-spec confirmed only /_debug-auth was registered, but it still returned the same generic 500.
- Normal Convex query/mutation execution was not globally broken. A normal query reached app code and returned my
app’s expected auth error. - Dashboard logs showed only the generic message, no stack trace, no handler logs, and ~single-digit ms execution.
Sample old request ID:
fb221cbef4b3553b
I created a fresh dev deployment in the same project, rightful-partridge-886, copied env vars, pushed the same
code, and HTTP actions immediately worked again. Better Auth sign-in/sign-out works there.
But now I made some adjustments and proved the same circumstances bricked another deployment.
This is with KitCN + Better-auth.
Question: can someone inspect internal logs for healthy-guineapig-155 and see why HTTP actions return generic 500
before handler execution?