GPT4All

GPT4All is a free-to-use, locally running, privacy-aware chatbot. There is no GPU or internet required. It features popular models and its own models such as GPT4All Falcon, Wizard, etc.

This notebook explains how to use GPT4All embeddings with LangChain.

Install GPT4All’s Python Bindings

%pip install -qU  gpt4all > /dev/null

Note: you may need to restart the kernel to use updated packages.

from langchain_community.embeddings import GPT4AllEmbeddings

gpt4all_embd = GPT4AllEmbeddings()

100%|████████████████████████| 45.5M/45.5M [00:02<00:00, 18.5MiB/s]

Model downloaded at:  /Users/rlm/.cache/gpt4all/ggml-all-MiniLM-L6-v2-f16.bin

objc[45711]: Class GGMLMetalClass is implemented in both /Users/rlm/anaconda3/envs/lcn2/lib/python3.9/site-packages/gpt4all/llmodel_DO_NOT_MODIFY/build/libreplit-mainline-metal.dylib (0x29fe18208) and /Users/rlm/anaconda3/envs/lcn2/lib/python3.9/site-packages/gpt4all/llmodel_DO_NOT_MODIFY/build/libllamamodel-mainline-metal.dylib (0x2a0244208). One of the two will be used. Which one is undefined.

text = "This is a test document."

Embed the Textual Data

query_result = gpt4all_embd.embed_query(text)

With embed_documents you can embed multiple pieces of text. You can also map these embeddings with Nomic’s Atlas to see a visual representation of your data.

doc_result = gpt4all_embd.embed_documents([text])