Langchain cannot create index when running inside Django server

Question

I have a simple Langchain chatbot using GPT4ALL that's being run in a singleton class within my Django server.

Here's the simple code:

gpt4all_path = './models/gpt4all_converted.bin'
llama_path = './models/ggml_model_q4_0.bin'

embeddings = LlamaCppEmbeddings(model_path=llama_path)

print("Initializing Index...")
vectordb = FAISS.from_documents(docs, embeddings)
print("Initialzied Index!!!")

This code runs fine when used inside the manage.py shell separately but the class instantiation fails to create a FAISS index with the same code. It keeps printing the llama_print_timings 43000ms with the ms increasing on every print message.

Can someone help me out?

first does the same thing workds outside django for example in script? if it works in script then you should be using *Celery* for this such big task when dealing in django to. https://docs.celeryq.dev/en/stable/django/first-steps-with-django.html — umair mehmood, Jun 16 '23 at 07:06
Celery is not what I needed for this... I just changed it to Chroma from FAISS and it worked fine. — Rocky C., Jul 01 '23 at 16:01

score 0 · Accepted Answer · answered Jul 01 '23 at 16:05

The answer was to use Chroma instead of FAISS for some reason. I still don't understand it because it honestly doesn't make sense but it works. I'll update this answer when I can and when I really figure it out but for those who run into something similar in the future and can't find a solution, try Chroma instead of FAISS.

Langchain cannot create index when running inside Django server

1 Answers1