Skip to main content
After tracing your application, the next step is to look at your data to understand system behavior and identify issues. Because there can be a lot of spans to navigate, it’s helpful to query traces by specific attributes or narrow results to a particular time range to focus your investigation. You can also export trace data for further analysis outside the platform.

​
Query Traces

Querying traces allows you to focus your analysis on specific system behaviors. Start by defining your goals, such as identifying performance bottlenecks or isolating spans that contain errors.
Image

​
Construct Queries

To start querying your traces, click on the filter bar. You can type, select, or use Alyx to help you craft filters that express your question clearly.

​
General Syntax

  • Format: "dimension name" = 'value'
  • Enclose the dimension in double quotes (") and the value in single quotes (').

​
Examples:

  • Equals: "dimension_name" = 'value'
  • Not Equals: "dimension_name" != 'value'
  • Contains: "dimension_name" contains 'substring'
  • Null: "attributes.input.value" = null or "attributes.input.value" is null
  • IN: "dimension_name" IN ('value_1', 'value_2')
  • AND/OR Support: Combine multiple conditions using AND or OR.

​
Operators:

  • = : Equals
  • != : Not equals
  • < : Less than
  • > : Greater than
  • IN : Includes

​
Functions:

  • Contains: Use contains to check if a string column contains a substring
✨ You can use AI Search to automatically generate query filters! Example: “Filter for LLM spans”

​
Saved Filters

When you find yourself using the same filters repeatedly — for example, to track specific spans or error types — you can save them. Saved filters act like shortcuts to your go-to queries, so you can eliminate the need to recreate complex filter queries repeatedly. To save a filter for future use:
  1. Configure your desired filter using the query filter syntax
  2. Click the Save button in the filter interface
  3. Choose one of the following options:
    • Add a new saved filter view: Create a completely new saved filter with a custom name
    • Update an existing filter: Modify an existing saved filter with your current configuration
  4. You can save and pin up to 7 filters at a time for quick access. (Note: Filters cannot currently be deleted from the pinned view)

​
Time Filters

Time filters let you focus analysis on the period that matters. The Time Selector makes it quick to define that window. Choose from presets, use the calendar picker, or type a custom range directly.
Image
  • Typing Custom Range: You can define exact time windows by typing directly into the selector. Supported formats:
    • Dates: 4/1 or 2025-04-01
    • Date and time: 4/1 3:00 AM or 2025-04-01 15:00
    • Ranges: 4/1 - 4/6 or 4/1 3:00 AM - 4/6 5:00 PM
  • Relative Time Shortcuts: For rolling time windows, use shorthand notation in the selector.
    • 15m, 15min, 15minutes → last 15 minutes
    • 1 h, 1 hr, 1 hour, 1 hours → last hour
    • 10d, 10day, 10days → last 10 days
  • Time Zone Support: You can set your preferred timezone, and the selection persists across sessions to ensure data is always shown in your local context.

​
Export Traces

It can be helpful to export the trace data from Arize for a variety of reasons. Common use cases include:
  • Testing out evaluations with a subset of the data
  • Create a dataset of few-shot examples programmatically
  • Augment trace data with metadata programmatically
  • Fine-tune a smaller model with production traces from Arize
Arize AX supports Exporting to a notebook and Exporting to a CSV from the UI:
Image

​
Export to Notebook

You can export data into code from the Tracing page by selecting Export to Notebook from the export dropdown menu. You will be provided with a prefilled function to export your trace data to a dataframe. 
# Install Arize SDK
! pip install "arize[Tracing]>=7.1.0"

import os
os.environ['ARIZE_API_KEY'] = 'YOUR_API_KEY'

from datetime import datetime
from arize.exporter import ArizeExportClient
from arize.utils.types import Environments

client = ArizeExportClient()

print('#### Exporting your primary dataset into a dataframe.')

primary_df = client.export_model_to_df(
    space_id= '',  # pre-filled space_id
    model_id='', # pre-filled model_id
    environment=Environments.TRACING,
    start_time= '', # pre-filled start_time
    end_time '', # pre-filled end_time
    # Optionally specify columns to improve query performance
    # columns=['context.span_id', 'attributes.llm.input']
)