Learning from past mistakes

[jlewi]

This PR gives Foyle the ability to learn from implicit human feedback.

Logs are mined for examples where the AI generated response was edited by the human before it was corrected. These are turned into examples (represented by the Example proto) of (Query, Answer). These examples are then stored in memory and retrieved via similarity search. They are then used to prompt the model with few shot prompting.

Key Changes:

1. Protobuf Adjustments: Introduction of the Example message in trainer.proto
   • An example represents (query, answer) pair that should be used in few shot prompting to try to teach the AI
2. Logging Module Enhancements: Modifications in the logging mechanism to better capture and log the corrections made by users, coupled with the generation of unique identifiers (UUIDs) for tracking.
3. Add a learn command in Foyle's CLI to process logs, identify corrected commands, and prepare data for training. This feature automates the learning process, making it more efficient and user-friendly.
4. InMemoryDB implements a simple in memory; brute force vector search

[netlify]

✅ Deploy Preview for foyle ready!

Name	Link
🔨 Latest commit	1e3b2cd
🔍 Latest deploy log	https://app.netlify.com/sites/foyle/deploys/662d6b07f62a440008b0d6c6
😎 Deploy Preview	https://deploy-preview-83--foyle.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.