Workspace
Managing Projects
Guide to projects in HoneyHive
Everything in HoneyHive is organized by projects.
A project is a workspace to develop, test & monitor a specific AI application.
Create your first project
Expected time: 1-2 mins
Steps
1
Navigate to the projects page
Navigate to the projects page.
2
Follow the interactive guide
Follow the steps as shown in the interactive guide below.
Congratulations, you have created your first project in HoneyHive.
Best-practices on organizing projects
Very often teams will have multiple components under the same application that they would like to develop & test independently as well.
For a complex pipeline, we recommend creating the following projects
- A production project - ex: Chatbot Production
- Goal: Source of truth for production
- All traces from production & CI tests are logged in this project.
- We want to make sure that the schemas in this project always maps to what we see in production & doesn’t get scrambled by team members trying different variations offline.
- One project per testable component - ex: Chatbot Retriever
- Goal: Source of truth for development
- All traces from offline evaluation runs & unit tests are pushed to this repo.
- Prompts & datasets for a particular piece of the pipeline can be managed in this project.
- We want to make sure as we try different things for a particular piece of our application that we can have a clear log of what we have tried & what worked. All of this can be tracked here.
- A catch-all/random project - ex: Chatbot Random
- Goal: Track random inspirations/ideas in one place
- Anything that’s not going to be referred to actively in future development, but would be good to be tracked somewhere.