Logging traces

Another method to upload datasets is by logging traces of your application. This involves setting up a tracer to monitor your application. Follow these steps:

  1. Set up a tracer in your application:

from RagaAICatalyst import Tracer

tracer = Tracer(
        "llm_model": "llm_model_name",
        "vector_store": "vector_store_name",
        "embed_model": "text-embedding-ada-002",
    metadata={"use-case": "YourUseCase", "stage": "testing-stage"}


# Your RAG Langchain application code here

  1. Ensure that the tracer is properly configured with the correct project name, tracer type, pipeline details, and metadata.

  2. Run your application with the tracer enabled to log the necessary data.

Once the traces are logged, you can view the logged dataset by navigating to the Logs tab within your project:

  1. Go to Projects.

  2. Select your project.

  3. Navigate to the Logs tab to view the logged dataset.

This method provides an automated way to log and upload datasets based on the activity within your application.

Creating Sub-Datasets from Logs

After logging traces, you can create a sub-dataset using filters. This allows you to experiment on specific segments of your logged data. Follow these steps:

  1. In the Logs tab, click the Create Dataset button.

  2. Enter a name for the sub-dataset.

  1. Select filters to refine your sub-dataset. You can filter by metadata, filter type, prompt length, time, etc.

  2. Click Save to create the sub-dataset.

Your sub-dataset is now created and can be used for further experiments and analysis within RagaAI Catalyst.

Last updated