CodeLit's Platform based migration - #194
rajat1saxena wants to merge 9 commits into
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
medialit/apps/api/src/index.ts
Lines 240 to 243 in 98de240
Configure the public bucket in the Compose deployment
The startup validation now requires CLOUD_PUBLIC_BUCKET_NAME, but the production docker-compose.yml only passes CLOUD_BUCKET_NAME to the API and has no way to forward the public bucket variable. Consequently, the documented Compose deployment exits during checkConfig() before serving requests, even when all of its currently required environment variables are supplied.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| services: | ||
| postgres: | ||
| image: postgres:16 |
There was a problem hiding this comment.
The new production Postgres service has no volumes entry, so all accounts, API keys, media metadata, sessions, and billing records live only in the container writable layer. Running docker compose down, recreating the database container, or moving the service to another host will start with an empty database and irreversibly orphan existing S3 objects.
Useful? React with 👍 / 👎.
| const publicApiUrl = ( | ||
| process.env.PUBLIC_API_URL || | ||
| process.env.API_SERVER || | ||
| `http://127.0.0.1:${port}` | ||
| ).replace(/\/$/, ""); | ||
| const webOrigin = ( | ||
| process.env.WEB_ORIGIN || | ||
| process.env.WEB_CLIENT || | ||
| "http://localhost:3000" | ||
| ).replace(/\/$/, ""); |
There was a problem hiding this comment.
In the production Compose deployment, SITE_ADDRESS is supplied only to Caddy; neither PUBLIC_API_URL/API_SERVER nor WEB_ORIGIN/WEB_CLIENT reaches the API. The new auth initialization therefore advertises http://127.0.0.1:8000 as its issuer and seeds redirects to http://localhost:3000, so OAuth and dashboard sign-in redirects on an externally served Compose instance point at the visitor's local machine instead of the deployed site.
Useful? React with 👍 / 👎.
Lemon had no subscriptions. OSS mode and an empty Dodo configuration now build an engine with no providers, offers, or catalog revision. Paid checkout still requires the full cloud catalog.
Root lint was parsing every JavaScript file with next/babel, which does not resolve outside the Next app. The legacy token and revoke routes are now rate-limited, refresh tokens are checked before new tokens are issued, and userinfo reads the bearer credential without a backtracking expression.
CodeQL types better-auth's handler property as undefined, and a typeof guard does not change that. An optional call is skipped by that query and still returns 500 when the handler is missing.
CodeQL treats better-auth's handler property as undefined, including through a typeof guard and optional chaining. Reflect.apply keeps the called value a known function.
CodeQL was reporting new Request as a call of undefined. The type-only Express import used the same name as the fetch constructor. The legacy forward path calls better-auth's handler directly again.
- Remove the scripts package’s duplicate database layer and migration files - Require the API to apply migrations before running the Mongo import
- Add Drizzle migration config and metadata - Run migrations via a dedicated command and Compose init service - Document the migration workflow
This branch has not been deployed
2 participants
No description provided.