Ok so here’s what happens next!
Module System
We want to keep the core of Convex small but extensible. This means we need to make it possible for third parties (and us) to build self-contained modules that can add functionality to a project without requiring you to copy-paste code to do so. This is going to be a really important building block for larger projects with many developers as well as opening the flood gates for feature development that doesn’t require changes to the Convex API.
Fun Run
This should be entirely transparent to you but we’re developing a new scale-out function runner service that will significantly increase the function parallelism in a single Convex project.
Open Source
Yes we’re finally getting around to this. We’re still defining exactly what this will look like but we at least want to get to the point where folks are able to run the Convex backend locally for testing/development. This will likely not be the entirety of the scale-out Convex infrastructure, which is a rather large and complex system, but will be the real source code with some parts stubbed out so it can be run as a single binary.
Snapshot Import/Export
Convex already has import/export but it’s not quite sufficient for backups or for bootstrapping a new dev deployment because it doesn’t provide referential integrity, i.e., if you import an csv/json export into Convex each document will have a new ID which doesn’t match existing document references. We want to provide the ability to import a true database snapshot into Convex.
Limits and enforcement
You may have already seen this but we just started actually charging for excess utilization on Pro accounts and are slowly starting to enforce limits on free accounts, albeit with some very generous grace thresholds. We’ll be tightening up some limit enforcement but not reducing any advertised limits. Reach out if you have a use-case that doesn’t fit our pricing model for some reason.
Search
We’re continuing to scale up vector search and are pretty happy and have successfully run workloads with millions of vectors in a single index. We’re also turning our attention back to full text search and finishing off some missing features there, especially prefix search.
Module SystemWe want to keep the core of Convex small but extensible. This means we need to make it possible for third parties (and us) to build self-contained modules that can add functionality to a project without requiring you to copy-paste code to do so. This is going to be a really important building block for larger projects with many developers as well as opening the flood gates for feature development that doesn’t require changes to the Convex API.
Fun RunThis should be entirely transparent to you but we’re developing a new scale-out function runner service that will significantly increase the function parallelism in a single Convex project.
Open SourceYes we’re finally getting around to this. We’re still defining exactly what this will look like but we at least want to get to the point where folks are able to run the Convex backend locally for testing/development. This will likely not be the entirety of the scale-out Convex infrastructure, which is a rather large and complex system, but will be the real source code with some parts stubbed out so it can be run as a single binary.
Snapshot Import/ExportConvex already has import/export but it’s not quite sufficient for backups or for bootstrapping a new dev deployment because it doesn’t provide referential integrity, i.e., if you import an csv/json export into Convex each document will have a new ID which doesn’t match existing document references. We want to provide the ability to import a true database snapshot into Convex.
Limits and enforcementYou may have already seen this but we just started actually charging for excess utilization on Pro accounts and are slowly starting to enforce limits on free accounts, albeit with some very generous grace thresholds. We’ll be tightening up some limit enforcement but not reducing any advertised limits. Reach out if you have a use-case that doesn’t fit our pricing model for some reason.
SearchWe’re continuing to scale up vector search and are pretty happy and have successfully run workloads with millions of vectors in a single index. We’re also turning our attention back to full text search and finishing off some missing features there, especially prefix search.
