ai-knowledge-base-sample

Pre-Requisites

- A mechanism for running Docker containers locally, perhaps Docker with Docker Desktop, Rancher, Podman or a preferred alternative
- The `docker-compose` command being available from your terminal. Depending on your Docker services this maybe built-in or not.
- Basic familiarity with running Docker containers & Docker-Compose services

Workshop Tasks

Setup your local environment

• Run docker compose up will start two containers
  • ElasticSearch which is a search and analytics engine, we will be using it as a 'vector store' for our AI
  • Our Knowledge retrieval AI app, which is a simple Python Flask interface.
• Open the web ui

NOTE Please be aware that in the interest of keeping this workshop accessible and achievable in a limited time frame the ElasticSearch best practice and security mechanisms have been disabled. This should never be run anywhere but this local development scenario.

Introducing OpenAI LLM to our base application

• Update the .env file with the provided Azure OpenAPI Keys provided
• Modify the ai.py file.
  • Modify the search function such that it takes the raw matching text from our vector store, and queries the LLM to get a natural language response: References

Further expansion

If you were unable to complete the previous task or ran out of time, we have a branch with the task completed available

Discuss what you would do next working towards an enterprise solution.

• Add styling though UI, UX?
• Add support for a conversational workflow?
• Batch uploading of knowledge?
• ...