Save and Load Traces

How to manually save and load traces

In addition to persisting to a database, Phoenix allows you to save and load your trace data to and from external files.

Connect to Phoenix

Before accessing px.Client(), be sure you've set the following environment variables:

import os

os.environ["PHOENIX_CLIENT_HEADERS"] = f"api_key=..."
os.environ["PHOENIX_COLLECTOR_ENDPOINT"] = "https://app.phoenix.arize.com"

If you're self-hosting Phoenix, ignore the client headers and change the collector endpoint to your endpoint.

Saving Traces

The initial step involves saving the traces present in a Phoenix instance to a designated location.

my_traces = px.Client().get_trace_dataset().save()

Loading Traces

To facilitate the retrieval of these saved traces, one can execute the prescribed commands upon starting Phoenix.

px.launch_app(trace=px.TraceDataset.load(my_traces))

Note the above will save to a default phoenix trace directory, to save in another directory, use the following example below.

Saving Traces to a Specific Directory

You can specify the directory to save your traces by passing adirectory argument to the save method.

import os

# Specify and Create the Directory for Trace Dataset
directory = '/my_saved_traces'
os.makedirs(directory, exist_ok=True)

# Save the Trace Dataset
trace_id = px.Client().get_trace_dataset().save(directory=directory)

This output the trace ID and prints the path of the saved file:

💾 Trace dataset saved to under ID: f7733fda-6ad6-4427-a803-55ad2182b662

📂 Trace dataset path: /my_saved_traces/trace_dataset-f7733fda-6ad6-4427-a803-55ad2182b662.parquet

Loading Traces from a Specific Directory

To load the previously saved trace dataset, use the trace ID and the specific directory path where the trace was stored.

NOTE: Make sure to trim down to ONLY the UUID to load. Include path strings for directory only and note to the dataset ID

px.launch_app(trace=px.TraceDataset.load('f7733fda-6ad6-4427-a803-55ad2182b662', directory="/my_saved_traces/"))