brianm / kcci

Keith's Index

Keith's Card Catalog Index

.plan

1. parse inputs in common formats to build candidate list
2. use openlibrary to get information about candidates
3. create embedding for each candidate
4. store embedding raw data with the candidate in sqlite or something (so we can rebuild index without recaclulating embeddings)
5. Build an index (annoy? faiss?) of the embeddings

References

• Library of Congress API
• OpenLibrary API
• WorldCat
• Google Books API
• SBERT Semantic Search
• Simon Willison's Explorations
• Semantic Search in Rust
• Annoy
• FAISS
  • To save the index: faiss.write_index(index, filename)
  • To load the index: index = faiss.read_index(filename)
• Textual terminal ui library