Overall this is dead simple, throw your prompt into the prompt.txt file and run the app. It will generate a chain message for you using whatever extra information that it can find in the knowledge base.
I have left out the extra data that I have trained on because it is not my data to share. If you would like to update your own model
you can aggregate your own data and train it by running the service and using the /reload endpoint, providing an
optional test_prompt parameter to see how the extra data may have affected the models output.
You can do two simple things here, update the models knowledge base and post a query to generate a chain message. Both require the presence of an OpenAI API key. You can get one by signing up for the beta at https://beta.openai.com/signup/.
This must be set in your ENV to OPENAI_API_KEY.
This app is run using poetry, a python package manager. To install dependencies using poetry, run the following command:
% poetry install && poetry update
Lint Python
% poetry run black . && poetry run isort .
Lint Typescript
% cd my-app && npx prettier --write .
If you have not already, generate the Open API client code for the front end using the following command:
openapi-generator generate -i http://localhost:8000/openapi.json -g typescript-fetch -o my-app/src/api
Run backend:
poetry run uvicorn service:app --reload
Run frontend
cd my-app && npm start
To view the API documentation, navigate to http://localhost:8000/docs