Skip to main content
Usually these will be traces you’ve previously saved using Save All Traces.

​
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.

​
Importing Traces to an Existing Phoenix Instance

import phoenix as px

# Re-launch the app using trace data
px.launch_app(trace=px.TraceDataset(df))

# Load traces into an existing Phoenix instance
px.Client().log_traces(trace_dataset=px.TraceDataset(df))

# Load traces into an existing Phoenix instance from a local file
px.launch_app(trace=px.TraceDataset.load('f7733fda-6ad6-4427-a803-55ad2182b662', directory="/my_saved_traces/"))

​
Launching a new Phoenix Instance with Saved Traces

You can also launch a temporary version of Phoenix in your local notebook to quickly view the traces. But be warned, this Phoenix instance will only last as long as your notebook environment is running 
# Load traces from a dataframe
px.launch_app(trace=px.TraceDataset.load(my_traces))

# Load traces from a local file
px.launch_app(trace=px.TraceDataset.load('f7733fda-6ad6-4427-a803-55ad2182b662', directory="/my_saved_traces/"))