CopilotKit is a typescript library for adding a hackable copilot to any react app.
You can let the copilot interact with your app via plain typescript closures, and give it (explicit) read access to app data.
An example user interaction could look like:
- "Which of these travel destinations has a rich architecture history?"
- [Copilot answers]
- "Great. Add these to my august trip folder, except the ones where it's typically rainy then"
- [Added]
Recursive agent integration (via Langchain) is in the works - if you have ideas I'd love to hear them here or on the discord.
LLMs (and agents) can often do 90% of the heavy lifting, but those 10% failures (or just weak results) can unpredictably pop up anywhere. Even 1% failures can sometimes render an AI solution uneconomical (see: self-driving cars).
The reasonable path to AGI is to have a tight AI/human feedback loop, and take more and more off of the human's plate over time. That way you can deliver actual value to users right now (not just a cool demo), and provide more and more value as the AI ecosystem matures.