search

run-llama / llama_index

LlamaIndex is a data framework for your LLM applications
https://docs.llamaindex.ai
MIT License
36.9k stars 5.28k forks source link

[Documentation]: Postgres Vector Store example notebook outputs a lot (4 MB) of vector embeddings #16800

Open jilljenn opened 2 weeks ago

jilljenn commented 2 weeks ago

Documentation Issue Description

I was a bit surprised; there are 4 MB of vectors in the output of this notebook.

https://github.com/run-llama/llama_index/blob/main/docs/docs/examples/vector_stores/postgres.ipynb

Also live: https://docs.llamaindex.ai/en/stable/examples/vector_stores/postgres/

Documentation Link

https://docs.llamaindex.ai/en/stable/examples/vector_stores/postgres/

dosubot[bot] commented 2 weeks ago

Hey @jilljenn! 👋 I'm here to help you with bugs, questions, and becoming a contributor. Let's tackle this together!

The Postgres Vector Store example notebook demonstrates how to use PostgreSQL and pgvector for vector searches in LlamaIndex. The 4 MB of vector embeddings are likely due to the default settings, such as the embedding dimension, which is set to 1536. To reduce the size of the output, you can adjust configuration options like reducing the embedding dimension, disabling text storage, or using JSONB for metadata storage [1].

To continue talking to Dosu, mention @dosu.