-
Notifications
You must be signed in to change notification settings - Fork 60
Description
Hey I went through the examples of Convex Agent component, what would be the best design/architecture and entry point to integrate an "artifacts files" feature (similar as Gemini chat canvas). E.g. instruct the LLM to use special <actifact_file>
tags in its text output (similar as <BoltArtifact>
) or provide tools to the LLM such as create_file
and edit_file
tools (similar as Codex/Cursor CLI)... the artifacts to be created need not be a full directory of files, just a specific single file/artifact (maybe with version history) would be enough.
So what is the best entry point and design/architecture to integrate such "artifact files"/canvas feature to Convex Agents component? I mean, the implementation should be easy, but what is the best design/architecture in the context of Convex dev?
Question 2: I also went through the source of Convex Chef/Bolt.diy and it seems that it is not using the Convex Agent component... why? I understand it is different people on different projects, but could we now settle this Jordan Hunt vs Ian Macartney, lol