Traces
Traces allow you to add some meaningful structure to the important steps in your GenAI workflow.
If you’re using a framework like Langchain, or LlamaIndex - Puzzlet will automatically detect the framework and annotate your traces. You do not need to follow the steps outlined here.
Puzzlet provides a few options for organizing your GenAI workflows with traces, components, and more.
Traces and Components
A trace serves as the central node of your observability workflow, ideal for consolidating your LLM-related functionality into a single view.
A component enables you to organize subsets of your trace into distinct, structured sub-views.
These can be executed using either decorators or functions. Note, if you use TypeScript decorators, you must enable
"experimentalDecorators": true
in your tsconfig.
You can run raw OpenAI, or other model provider calls here. It doesn’t have to be from our Prompt Management SDK. Each feature we have is incrementally adoptable. You can use tracing without our prompt management.
Questions or Feedback?
Have any questions or feature requests? Reach out to us at hello@puzzlet.ai