Build applications that make decisions (chatbots, agents, simulations, etc...). Monitor, trace, persist, and execute on your own infrastructure.
Email Assistant
This example shows how to create a basic email assistant. This assistant takes in an email and will write a response. It allows you to prompt it with:
- An email
- Directions on how to respond
Then it generates questions for you for clarifications, and uses the answer to write a draft. It will go back and forth with you until you are satisfied with the response.
This demonstrates how to do multi-shot modeling interactively with an LLM. You'll learn:
- How to move in/out of program control -> user control
- How to process inputs at multiple points through the process
- How to involve multi-shot modeling with an LLM
How to run
To run this example, you need to install the following dependencies:
pip install -r requirements.txt
Then if you run jupyter lab you can open/execute the notebook.
This simulates an interactive web app. The notebook is the best way to get started.
Note we will be adding two things to this demo:
- An integration with the burr web app
- a standalone server example with a walkthrough
Running the UI with Email Server Backend in a Docker Container
We have added Docker support along with a compose file and nginx as a proxy for the API and Burr UI. To run the email assistant app with these new features:
Run the following command:
docker compose up --buildThis will start the email assistant app along with nginx, which allows you to access:
- BURR UI on
telemetry.localhost - API on
api.localhost
- BURR UI on
If you prefer not to use subdomains powered by nginx, you can also access:
- Email server running Burr: Navigate to
localhost:7242/docs - Burr UI: Navigate to
localhost:7241
- Email server running Burr: Navigate to
If you don't have a UI, go to demos and play with the email-assistant. Otherwise, connect to the Burr email server with your UI code.
Note: This setup does not mount a persistent volume, so state is lost once the container goes down.
Additional Information
We will be adding two things to this demo:
- An integration with the Burr web app
- A standalone server example with a walkthrough