> ## Documentation Index
> Fetch the complete documentation index at: https://docs.promptlayer.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Search

> Find request logs and traces with structured filters, date ranges, and request-content search.

Search narrows [request logs](/features/observability/request-logs) and [traces](/features/observability/traces) to the data you need. The same query also controls the corresponding Analytics view, so you can move from individual records to aggregate trends without rebuilding your filters.

PromptLayer's current structured search is backed by OpenSearch. The dashboard automatically uses a compatibility mode when your date range includes data from before your workspace's OpenSearch cutover.

## Search in the Dashboard

1. Open [**Request Logs**](/features/observability/request-logs) or [**Traces**](/features/observability/traces) from the PromptLayer sidebar.
2. Set the date range and timezone.
3. Select the search bar, then choose a field, operator, and value.
4. Add more filters to narrow the results.
5. Stay on the Requests or Traces tab to inspect matching records, or switch to **Analytics** to chart the same query.

Multiple structured filters are combined to narrow the result set. Suggested values are based on the selected field and data available in your workspace.

## Search Request Logs

Request-log search supports a free-text query across prompt input and model output, plus structured filters in these categories:

| Category            | Example fields                                                                 |
| ------------------- | ------------------------------------------------------------------------------ |
| Prompt and user     | Prompt, User ID                                                                |
| Model               | Model, Provider                                                                |
| Labels and metadata | Tags, Metadata                                                                 |
| Content             | Input Text, Output Text, Input Variable, Output, User Intent, Agent Intent     |
| Usage               | Input Tokens, Output Tokens, Cost, Latency                                     |
| Request             | Request ID, Request Status, Error Type                                         |
| Format              | Response is JSON, Response is a Tool Call, Tool Name, Request is Part of Trace |

Nested fields such as metadata, structured output, and input variables let you choose a key before matching its value. Use an exact operator for identifiers and short structured values; use a contains operator for longer text. See [Search Data Model](/features/search-and-analytics/search-data-model) for indexing and operator details.

## Search Traces

Trace search separates whole-trace fields from span-level fields:

| Scope | Example fields                                                                                                  |
| ----- | --------------------------------------------------------------------------------------------------------------- |
| Trace | Name, status, ID, user, duration, cost, tokens, models, tools, prompts, workflows, span count                   |
| Span  | Name, status, kind, duration, cost, tokens, model, prompt, workflow, tool, input, output, exception, attributes |

A span-level filter returns traces containing a matching span. This makes it possible to find an end-to-end run by a tool call, model, exception, or custom span attribute anywhere in its hierarchy.

## Search Older Data

Each workspace has a search cutover date. When the selected date range starts before that date, the dashboard switches to older-data compatibility mode:

* Filters that are unavailable for older data are marked as unavailable.
* Incompatible active filters are not applied to the older-data query.
* The Analytics tab is unavailable for that date range.

Move the start of the date range to the displayed cutover date or later to use the complete OpenSearch filter set and Analytics.

## Continue from Search Results

* Select a request or trace to inspect its details.
* Switch to [Analytics](/features/search-and-analytics/analytics) to aggregate the current query.
* Select matching request logs or traces and import them into a [Table](/features/tables/overview#import-data).
* Use the [Search Request Logs API](/reference/search-request-logs) to run structured request-log searches outside the dashboard.