Pipeline Logs
Learn how to inspect pipeline and agent execution logs in Digibee, filtering by pipeline, agent, or log content to troubleshoot and monitor your pipelines in real time.
Pipeline logs are detailed records that let you inspect what is happening at each step of a pipeline's execution cycle. In the Pipeline Logs tab in the Monitor area, you can track the event logs recorded during pipeline execution, either at the pipeline level or at the agent level.
Environment
Select the desired environment in the upper left corner of the page. When you select an environment, the page refreshes to display the corresponding data.
Pipelines and Agents views
Use the toggle at the top of the Pipeline Logs tab to switch between two views:
Pipelines: Logs generated during pipeline execution.
Agents: The same logs as Pipelines, filtered to show only the logs generated by agents created with the Agent Component.
Pipelines
Search fields
Use the fields below to filter pipeline logs. The results list displays logs that match the selected criteria.
Log message
The information sent by a connector that returns logs during pipeline execution. Search for whole words only.
Pipeline name
The pipeline name as defined by the user. Type the pipeline name and select it from the dropdown list to filter the results.
Pipeline version
The pipeline version number. Must be used together with the pipeline name, otherwise the search returns no results.
Pipeline execution key
The unique identifier of a pipeline execution.
Filters
Severity
Filter results by log classification:
All severities: Any kind of log information.
Info: Information about ordinary events during pipeline execution.
Error: Information about errors during pipeline execution.
Warn: Information about possible problems during pipeline execution.
Source
Filter results by log source:
All sources: Any kind of source.
Digibee internal: Only internal platform logs.
Log connector: Only logs explicitly added by the user in the pipeline.
Time period
Filter results by execution time. Select from the last 5, 15, 60 minutes, 1 day, or define a custom time interval.
Logs table
The logs table displays results according to the selected filters. Click a log to open its details in a sidesheet.
Agents
The Agents view draws from the same log data as the Pipelines view, but only displays the logs generated by agents created with the Agent Component.
Search fields
Use the fields below to filter agent logs. The results list displays logs that match the selected criteria.
Pipeline
The pipeline name that contains the agent.
Agent name
The agent name as defined in the Agent Component.
Content
The message content of the agent log.
You can also filter by time period (last 5, 15, 60 minutes, or a custom interval), the same way as in Pipelines.
Logs table
The logs table displays one row per agent execution, with the following columns:
Timestamp: Date and time when the agent execution occurred.
Agent name: The Agent Component name.
Agent execution key: The unique identifier of an agent execution. A pipeline key can have multiple agent execution keys, from different agents or from one agent running in a loop.
Pipeline name: The pipeline name and version that contains the agent.
Pipeline key: The unique identifier of a pipeline execution. Included in each agent execution row so you can trace it back to the pipeline that ran it.
Content preview: A preview of the message returned by the agent.
Click a log to open the Agent log details panel.
Agent log details
The Agent log details panel shows the execution ID, time, and duration of the selected agent execution, followed by two collapsible groups — Execution logs and Output — and an action to save the execution as a test case.
Execution logs
Expand each section to inspect a specific stage of the agent execution:
Configuration: Provider and model settings.
Input: Context, tools configuration, and retrieval status.
System message: The system prompt sent to the model. You can view it as Text, Markdown, or JSON.
User message: The user prompt sent to the model, such as the claim context pack or other contextual data. You can view it as Text, Markdown, or JSON.
Tool call: Tool calls, arguments, and results.
Input/Output Guardrail: Applied guardrails and their impact.
Use these logs to validate behavior and troubleshoot issues.
Output
The Output section shows the final result returned by the agent execution, formatted as JSON. Click the copy icon to copy the full output. You can also use the Search by JSONPath field to look for specific information in the output.
Save this execution as a Test Case
At the bottom of the panel, click Create test case to turn this execution into a Test Case instead of building a test scenario from scratch. This lets you validate your agent against inputs that actually happened in production, rather than simulated ones.
Select the Dataset you want to save this execution to, then map the pre-filled variables.
Use the Evaluations tab to add or edit the values for the evaluations already configured on the Dataset. You can't create new evaluations from this tab; to add a new evaluation, configure it directly on the Dataset inside the Agent Component.
Click Save to complete the Test Case.
The new Test Case is added to the selected Dataset and becomes available the next time you run your Evaluations on the Agent Component inside the pipeline. For details on Datasets, Test Cases, and Evaluations, read Build Agent tests: Datasets, Test Cases, and Evaluations.
Last updated
Was this helpful?