tfulanchan / langchain-chroma

Search Your PDF App using Langchain, ChromaDB, and Open Source LLM: No OpenAI API (Runs on CPU)

dependencies required

pip install -r requirements.txt

python accelerate

pip install accelerate

sentence Transformers

pip install -U sentence-transformers

vicuna embeddings

minilm v6 sentence transformer resp

vector store

chroma db

pip install chromadb

streamlit

pip install streamlit

parquet python

pip install parquet

llm pipeline hugging face

inference langchain

closed-source alternatives

openai embedding pinecone

logic

langchain embeddings vector store chains by langchain llm / embeddings

knowledge point

vectore db vs sql db inbuilt algorithms semantic search levenstein jacquard

reasons

lower dimensional space inbuilt default algorithms cosine similarity

directory should contain

db docs lamini-t4-738m pycache .venv license requirements.txt .gitignore

llm

Get Started

code .

conda activate deeplearning

requirements.txt

langchain streamlit UI for citation transformers requests torch einops accelerate large models bitsandbytes pdfminer.six pi pdf bs4 sentence_transformers chromadb

constants.py app.py ingest.py

ingest.py

what's going on in ingest.py

import things

define directory

main function

recursive text splitter

create embeddings store in Chroma

python ingest.py

app.py

streamlit run app.py

validate response

client requirements

summarization and retrieval