Making a way to have LLMs autonomously construct, query and update an external database from raw data.
Measure performance of system:Create library (Pdfs, textbooks, documentation, code etc)Create queries corresponding to a chunk of text (To test performance of external database)
- Write cloud functions to extract text from raw data like:
PDFs- Codebases on github
- Documentation on web
- Implement system:
Create binary tree with LLMs using sumarization(Constructing tree) and comparison(Traversing tree) first- Add features:
- Try optimizing for tokens, aka, not strictly a binary tree
- Allow for clone nodes, implement cleanup operation