Fix a bug in demo code and add a demo for custom embedding_function

jdagdelen / hyperDB

A hyper-fast local vector database for use with LLM Agents. Now accepting SAFEs at $135M cap.

MIT License

1.36k stars 83 forks source link

Open conanak99 opened 1 year ago

conanak99 commented 1 year ago

The result of this line indemo/demo.py returns a list of tuple(document, similarity), so the demo code will not work correctly. This is fixed in this PR..

results = db.query("Likes to sleep.", top_k=5)

The current demo requires an OPENAI_API_KEY. Added demo/demo_custom_embed.py that makes use of SentenceTransformers/ to generate embedding, which can be run locally.

jdagdelen commented 1 year ago

Can this be run locally on cpu only?

jdagdelen commented 1 year ago

I would like to avoid adding dependencies just for the purpose of demos.

conanak99 commented 1 year ago

hi @jdagdelen , it can be run on CPU only, all-MiniLM-L6-v2 is a lightweight model.

I understand the concern about adding new dependencies. Fixed that, add a comment in the demo so people know which dependencies to install instead :D

One more thing to note is the embedding from local model is not as good as OpenAI, based on my test.