This project is an exploration of OpenAI's Assistant API - trying to evaluate the strengths and weaknesses.
- Is an Assistant easier to use than Llama Index/Langchain?
- Does an Assistant provide more accurate answers than Llama Index/Langchain?
This project is based off the existing Knowledge Base Demo.
- Our external Notion wiki
- Our public website: Focused Labs
- A Open AI API account (api key). You can sign up at Open AI's website.
- Python (and your favorite IDE). We are using python v3.10.7.
OPENAI_API_KEY = "<Open AI API Token>"
Run the code blocks in sandbox.ipynb
This will result in an accuracy_test_results.csv
file. You can open this and evaluate the accuracy of the answers.