langchain-samples

Python code for samples at LangChain Chat with Your Data course of https://learn.deeplearning.ai/ See original github repo https://github.com/ksm26/LangChain-Chat-with-Your-Data/

prerequisites on environment

(Tested on macOS 14.2 Sonoma)

xcode-select --install

brew install python

pip3 install langchain (https://pypi.org/project/langchain/)
pip3 install langchain-community
pip3 install openai

Obtain OpenAI API Key http://www.openai.com Add OpenAI API Key OPENAI_API_KEY in .zshenv file

prerequisites on pdf sample

pip3 install python-dotenv
pip3 install langchain-community

prerequisites on youtube sample

brew install fprobe
brew install ffmpeg
pip3 install yt_dlp
pip3 install pydub

The 429 error {'error': {'message': 'You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.', 'type': 'insufficient_quota', 'param': None, 'code': 'insufficient_quota'}} is avoid if account for OpenAI API is charged with some credit.

prerequisites on HTML sample

pip3 install bs4

prerequisites on Text Splitter sample

No prerequisites

Prerequisies on Vector Stores sample

pip3 install -U langchain-openai
pip3 install tiktoken
python3 -m pip install tiktoken
python3 -m pip install chromadb
Add on PATH '/Library/Frameworks/Python.framework/Versions/3.11/bin'

Prerequisites on Retrieval sample

pip3 install lark
pip3 install scikit-learn

Prerequisites on Question Answering sample

No prerequisites

Prerequisites on Conversational chatbot sample

pip3 install pydantic==1.10.8  (it does not work with last pydantic version)
pip3 install panel
pip3 install docarray

Prerequisites on testPretrained sample

pip3 install torch torchvision
pip3 install text-generation
BUILD_CUDA_EXT=0 pip install auto-gptq ( reember disable CUDA if non-nvdia GPU)
brew install poppler (prerequisite pdf2image)
pip3 install pdf2image
pip3 install -U sentence-transformers==2.2.2 (important force version to 2.2.2 or we will face : TypeError: INSTRUCTOR._load_sbert_model() got an unexpected keyword argument 'token')
pip3 install InstructorEmbedding
pip3 install einops
(When running on CPU => python3 testPretrained.py --autolaunch --skip-torch-cuda-test --disable-nan-check --no-half --use-cpu all)