Conversation
fixed signup name optional error
…for protected routes, enhancing user redirection to the sign-in page.
…cument Feature/update employee document
Feature/setting up docker
fixed-UI-problems and added reame for testing
…-improvement updated OCR Processing step, embedding storage model, and prompts
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![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
Here are some automated review suggestions for this pull request.
Reviewed commit: 1d67a00320
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".
| if ( | ||
| (server.INNGEST_EVENT_KEY == null || server.INNGEST_EVENT_KEY.length === 0) | ||
| ) { | ||
| throw new Error("INNGEST_EVENT_KEY is required in production"); |
There was a problem hiding this comment.
This throws whenever INNGEST_EVENT_KEY is missing, even though JOB_RUNNER supports non-Inngest backends and the comment says the key is only required in production. Because env.server is parsed at startup, development or JOB_RUNNER=trigger-dev deployments will fail to boot before any dispatcher selection happens. Gate this check on NODE_ENV === "production" and/or JOB_RUNNER === "inngest" instead of enforcing it unconditionally.
Useful? React with 👍 / 👎.
| (!mimeType || | ||
| mimeType === "application/pdf" || |
There was a problem hiding this comment.
Treating !mimeType as PDF here misroutes any upload that omits MIME metadata (e.g., .txt, .html, image files) into the PDF OCR pipeline instead of the ingestion adapters. Since mimeType is optional in the request/event shape, this causes valid non-PDF documents to hit routeDocument/normalizeDocument with the wrong processor and fail extraction.
Useful? React with 👍 / 👎.
| const data = (await resp.json()) as { embeddings: number[][] }; | ||
| return data.embeddings; |
There was a problem hiding this comment.
The sidecar response is parsed as embeddings-only and the returned vector size is never checked, but downstream storage/retrieval assumes vector(1536). If the sidecar model dimension differs (including the default sentence-transformer setup in this commit), inserts/casts will fail at runtime and document jobs will error after vectorization. Validate dimensions (or normalize/truncate/pad) before constructing vectorizedChunks.
Useful? React with 👍 / 👎.
1 participant
No description provided.