DocumentationHistory & MonitoringFiltering and Searching Requests

Filtering and Searching Requests

The History page includes filtering capabilities that help you narrow down classification requests to find exactly what you’re looking for. Whether you need to review a specific request, analyze patterns by source, or examine all failed classifications, filters let you quickly isolate relevant results.

Available Filters

The filter bar appears at the top of the History page, above the request table. All filters work together—you can combine multiple filters to create precise queries.

History page showing the filter bar with all filter options visible

Request ID

Search for a specific classification request by its unique identifier.

How to use: Enter the complete request ID in the search field. Request IDs are available in the “Categorization details” dialog when you click View on any request in the History table.

When to use:

  • Troubleshooting a specific classification issue
  • Following up on a request reported by your team
  • Reviewing details of a particular product classification

Example: Search for req_abc123xyz to view that single request.

Source

Filter requests by where they originated.

Options:

  • All Sources (default) - Shows requests from all sources
  • Playground - Only web interface classifications
  • API - Only programmatic requests
  • MCP server - Only Model Context Protocol requests

When to use:

  • Monitoring API vs Playground usage patterns
  • Debugging issues specific to one integration
  • Analyzing which systems generate the most classifications
  • Reviewing test classifications separately from production requests

Example: Select “API” to see only production classifications, excluding any testing done in the Playground.

Status

Filter by classification outcome to focus on specific result types.

Options:

  • All Statuses (default) - Shows all classifications
  • Categorized - Only requests where a category was found (green and yellow indicators)
  • No Category - Only requests where no category was returned (black indicators)
  • Error - Only requests that failed with an error (red indicators)

For detailed explanations of what each status means, see Understanding Status Indicators.

When to use:

  • Identifying products that need better descriptions (No Category results)
  • Reviewing lower-confidence classifications (Categorized status includes yellows)
  • Troubleshooting API errors (Error status)
  • Measuring successful classification rate

Example: Select “No Category” to review all products that returned black indicators, then improve their descriptions or add custom instructions.

Date Range

Filter requests within a specific time period.

How to use: Click the date range picker to select start and end dates. The History page retains data for 45 days, so you can select any range within that window.

Default: Last 7 days

When to use:

  • Comparing classification patterns over time
  • Reviewing yesterday’s bulk import
  • Auditing a specific deployment or configuration change
  • Analyzing recent error rates

Example: Select “Yesterday” to review all classifications from your latest batch import.

Date range picker showing calendar interface with a selected range

How Filtering Works

Immediate Results

Changing any filter immediately reloads the History table with matching requests. No “Apply” button is needed—results update as soon as you adjust filters.

Combined Filters

All active filters work together using AND logic. Each filter further narrows your results.

Example combined filter:

  • Source: API
  • Status: No Category
  • Date range: Last 24 hours

This shows only API requests from the last 24 hours that returned no category. Use this to identify recent API classifications that need attention.

Resetting Filters

To reset to the default view:

  1. Change each filter back to its default value
  2. Source: All Sources
  3. Status: All Statuses
  4. Date range: Last 7 days

Or refresh the page to reset all filters at once.

Common Filtering Scenarios

Reviewing Failed Classifications

Goal: Find products that need better descriptions

Filters:

  • Status: No Category
  • Source: All Sources
  • Date range: Last 7 days

Review these results and improve product descriptions or add custom instructions for recurring patterns.

Monitoring API Errors

Goal: Identify and troubleshoot API issues

Filters:

  • Status: Error
  • Source: API
  • Date range: Last 24 hours

Check for patterns in error messages, authentication issues, or rate limit problems.

Analyzing Test vs Production

Goal: Separate testing activity from production classifications

Filters for testing:

  • Source: Playground
  • Date range: Today

Filters for production:

  • Source: API
  • Date range: Last 7 days

Compare results to ensure your Playground testing translates well to production.

Audit Trail for Specific Request

Goal: Find complete details for a reported issue

Filters:

  • Request ID: [paste ID from support ticket]

This returns a single request with all details. Click View to see the complete classification information.

Measuring Classification Quality

Goal: Understand your success rate

First query (successful):

  • Status: Categorized
  • Date range: Last 7 days

Note the count of results.

Second query (failed):

  • Status: No Category
  • Date range: Last 7 days

Compare the counts to calculate your classification success rate.

Tips for Effective Filtering

Start Broad, Then Narrow

Begin with broader filters and add more specific ones as you identify patterns:

  1. Status: All Statuses
  2. Notice many No Category results from yesterday
  3. Add Date range: Yesterday
  4. Add Source: API
  5. Now reviewing yesterday’s failed API classifications

Use Request ID for Debugging

When investigating a specific issue:

  1. Get the request ID from your logs or the View dialog
  2. Filter by that request ID
  3. Review complete classification details
  4. Check product description, settings used, and result

Review Yellow Results Regularly

Yellow indicators (RelevantCategory confidence level) represent lower-confidence classifications. Filter for these to identify products that might benefit from:

  • More detailed descriptions
  • Custom instructions
  • Manual verification

Note: To see yellow vs green results, you’ll need to view individual requests since the Status filter only distinguishes Categorized vs No Category vs Error.

Track Patterns Over Time

Use date ranges to spot trends:

  • Compare error rates week over week
  • Identify if new products cause more No Category results
  • See if configuration changes improved success rates

Limitations

45-Day Retention: Only requests from the last 45 days appear in History. Older requests are automatically deleted.

No Text Search: You cannot search within product descriptions or category names. Use Request ID to find specific requests.

Status Filter Granularity: The Status filter distinguishes between Categorized (green and yellow) and No Category, but not between green (AI-selected) and yellow (top-ranked fallback) indicators. Click View on individual requests to see the specific confidence level.

Next Steps

Understanding Classification HistoryViewing Request Details