The code for delivering the ChatBot project for PS
pip install -r requirements.txt
Put your documents into ./docs/[cat_name]/ for separate folders. The names of [cat_name] will appear as tickboxes so that you can choose one or multiple sources for chatting
Create a questions.txt file for each category of documents and put each question on a new line. Put your example questions in the ./docs/[cat_name]/ folder containing the relevent documents for that set of questions.
See below regarding where to find your neo4j, OpenAI and Pinecone keys.
# This is the options that if you want to talk to your graph
NEO4J_URL="xxxxx"
NEO4J_USER="xxxxx"
NEO4J_PASSWORD="xxxxx"
# Your openai key
OPENAI_KEY="xxxxx"
# The pinecone of storing the embeddings of your pdf documents
PINECONE_KEY = "xxxxx"
PINECONE_ENV = "xxxxx"
PINECONE_INDEX = "xxxxx"
# The directory of storing your documents
PDF_DOC_DIR = "./docs"
# Upload parameters change as you like
CHUNK_SIZE = 1000
CHUNK_OVERLAP = 0
# Define the namespaces for customization. (You can ignore it)
PINECONE_PDF_NAMESPACE = "my-pdf"
PINECONE_GRAPH_NAMESPACE = "my-graph"
- Visit OpenAI to retrieve API keys and insert into your .env file.
- Visit Pinecone to create and retrieve your API keys, and also retrieve your environment and index name from the dashboard.
- Visit neo4j to retrieve URL
This could take a while if you have multiple documents. This function will upload the embeddings of your PDF documents and graph database to the cloud vector database. You can uncomment the graph ingesting if you only want to upload a new pdf.
python ingest_data.py
After ingesting your data, simply run main.py for local deployment and debugging.
python main.py
gunicorn -w 2 -b :8000 main:app --timeout=120
Please refer https://ngrok.com/docs/getting-started/