feat(vector): support 2560-dimensional embeddings

[JustinMissmahl]

Summary

Add 2560 to the vector dimensions supported by the Agent component.

This enables embedding models that return 2560-dimensional vectors, including:

• perplexity/pplx-embed-v1-4b
• perplexity/pplx-embed-context-v1-4b

Motivation

The Agent component currently restricts vector dimensions to a predefined
allowlist. Although Convex vector indexes support 2560 dimensions, it is
missing from that list.

As a result, applications cannot use these embedding models at their native
dimension without maintaining a local patch or reducing the embedding size.

Changes

• Add 2560 to VectorDimensions
• Regenerate the Convex component types to include the new dimension

Testing

• Ran npm run build:clean
• Ran npm run test
• Ran npm run lint
• Ran npm run typecheck

I have also been using this change successfully through a local package patch
with perplexity/pplx-embed-v1-4b.

---

By submitting this pull request, I confirm that you can use, modify, copy, and
redistribute this contribution, under the terms of your choice.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: get-convex/coderabbit/.coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 6ff706ca-015c-4c52-a621-001c1dfa7322

📥 Commits

Reviewing files that changed from the base of the PR and between db083fe and bc9809a.

⛔ Files ignored due to path filters (1)

• src/component/_generated/component.ts is excluded by !**/_generated/**

📒 Files selected for processing (1)

• src/component/vector/tables.ts

---

📝 Walkthrough

Walkthrough

The supported vector dimensions now include 2560. Derived vector dimension types, validation, and dimension-based table names and identifiers consequently recognize the new dimension.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely summarizes the main change: adding support for 2560-dimensional embeddings.
Description check	✅ Passed	The description is directly related to the change and accurately explains the motivation, update, and validation steps.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

Warning

There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure.

🔧 ESLint

If the error stems from missing dependencies, add them to the package.json file. For unrecoverable errors (e.g., due to private dependencies), disable the tool in the CodeRabbit configuration.

ESLint install timed out. The project may have too many dependencies for the sandbox.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[ianmacartney]

thanks! Sounds like it's worth adding

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/@convex-dev/agent@298

commit: bc9809a

[ianmacartney]

I'll defer to @robelest whether we should wait to land it or replay the change on top of some big structural PRs he's working with, but npm i https://pkg.pr.new/@convex-dev/agent@298 should allow you to try it out (sounds like you have a patched version already though)