jamesliangg / PrismBot

Get factual information grounded from reliable sources

Getting Started

Folder Structure

.
├── .github                                             # Github Actions
│   └── workflows                                       # Workflows
│       └── gcp_deploy.yaml                             # The cooler package
├── .gitattribute                                       # Git attributes
├── .gitignore                                          # Git ignore
├── Dockerfile                                          # Dockerfile
├── Guidelines-FINAL-4TH-EDITION-With-2023-Updates.pdf  # Sample PDF
├── README.md                                           # Readme
├── chat.py                                             # Chatbot
├── lit.py                                              # Streamlit app
├── requirements.txt                                    # Python requirements
└── vectorstore.pkl                                     # Vectorstore configuration

Redis Database

Locally

• Run docker run -d -p 6379:6379 -p 8001:8001 redis/redis-stack:latest
• Connection URL = redis://localhost:6379
• If hosting on a cloud service, can use ngrok to connect to Redis
  • Run ngrok tcp 6379
  • Example output is tcp://2.tcp.ngrok.io:17000 use 2.tcp.ngrok.io as host and 17000 as port
  • Redis url becomes redis://2.tcp.ngrok.io:17000

Cloud

All you need is the connection URL to your database

.env File

Create a .env file in the root directory with the following content:

REDIS_URL=<REDIS_URL>
GOOGLE_APPLICATION_CREDENTIALS=GOOGLE_APPLICATION_CREDENTIALS.json

You'll also need to create a service account key for Google Cloud and save it as GOOGLE_APPLICATION_CREDENTIALS.json in the root directory

Service Account Scopes

• Compute Engine API
• Cloud Run Admin API
• Vertex AI API

Starting Streamlit

• Start Streamlit with streamlit run lit.py --server.port=8080
• Access the app at http://localhost:8080

Useful Commands

Delete all keys in Redis DB redis-cli flushall

Bash Script to Import

Run the following script in your terminal to import the data into Redis from resources.md

chmod +x process_links.sh

./process_links.sh

Document Sources

• https://www.rainbowhealthontario.ca/
• https://genderdysphoria.fyi/en
• https://www.the519.org/about/

References

• https://python.langchain.com/docs/integrations/vectorstores/redis
• https://python.langchain.com/docs/integrations/providers/cohere
• https://python.langchain.com/docs/use_cases/question_answering/quickstart
• https://api.python.langchain.com/en/latest/vectorstores/langchain_community.vectorstores.redis.base.Redis.html#
• https://python.langchain.com/docs/modules/data_connection/document_loaders/pdf
• https://docs.streamlit.io/library/cheatsheet
• https://python.langchain.com/docs/integrations/llms/google_vertex_ai_palm
• https://python.langchain.com/docs/integrations/text_embedding/google_vertex_ai_palm
• https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/overview
• https://cloud.google.com/blog/products/databases/memorystore-for-redis-vector-search-and-langchain-integration
• https://github.com/google-github-actions/example-workflows/blob/main/workflows/deploy-cloudrun/cloudrun-docker.yml
• https://cloud.google.com/run/docs/securing/managing-access