RAG Component

Convex now has a Retrieval-Augmented Generation (RAG) component. You can now easily chunk, namespace, filter, and do surrounding data retrieval when working with large documents in your AI apps.

Use the component

Convex on PlanetScale Postgres

Convex is switching to PlanetScale from AWS Aurora for its underlying data storage. Every new Convex project now runs on PlanetScale, and we’ll be rolling the change out to all projects soon. This should be a transparent change for you; things will simply run faster.

See Jamie’s post

Convex 1.25

Convex 1.25 was released last week with several important improvements, including enhancements to bundling, queuing for the ConvexHttpClient, and more. Upgrade now: npm i convex@latest

Be sure to read the changelog

Better scheduler logs

Now you can get statistics on scheduled functions in log streams. For example, graph scheduled functions by count in Axiom:

dataset
| where ['data.topic'] == "function_execution" 
| where isnotnull(['data.scheduler_info.job_id'])
| summarize count() by ['data.function.path']

Also, for example, graph the number of concurrent jobs

dataset
| where ['data.topic'] == "scheduler_stats"
| summarize max(['data.num_running_jobs']) by bin(_time, 1m)

See the docs

Self-hosting Axiom, Datadog, Sentry, and other integrations

We’ve open sourced our various log and data streams integrations. As more people start using self-hosted Convex for real-world applications, we want to provide you with additional tools to manage and scale.

See the latest self-hosted changelog