This technical documentation covers access and review of granular FastGPT application evaluation results, as part of the official FastGPT build evaluation workflow.
Accessing Evaluation Detail Pages
Follow these steps to open the full evaluation detail view:
- Navigate to the FastGPT evaluation management dashboard within your workspace.
- Locate the target saved or completed evaluation task from the listed runs.
- Click the explicitly labeled "View Details" button associated with the selected task. The reference interface asset
shows the initial dashboard layout where this control is located.
Task Overview Panel
The top section of the opened detail page is the Task Overview panel. This section displays consolidated high-level metadata for the evaluation run, including two core data sets: the full evaluation configuration parameters used to launch the task, and aggregated summary statistics across all evaluated QA pairs. This panel allows technical stakeholders to quickly assess the evaluation’s scope and overall performance without reviewing individual test cases.
Detailed Results Panel
The lower section of the page is the Detailed Results panel, which provides a line-item breakdown of every QA pair included in the evaluation, paired with its calculated performance score. Each listed entry includes four critical data points: the original end-user question submitted to the FastGPT application, the predefined expected output defined in the evaluation dataset, the actual response generated by the FastGPT application during the evaluation run, and the assigned performance score for that specific pair. This granular view enables targeted debugging of individual application responses, validation of specific QA pairs against expected outcomes, and analysis of recurring performance patterns across the evaluated dataset.
Source: FastGPT official source
Applicability and version scope
Use this page for the documented Tutorials scenario. Confirm the FastGPT, dependency, API, and deployment versions in the official source before applying a change.
Safety guardrails
Use [REDACTED_CREDENTIAL] for credentials and private data. Confirm the documented environment and version before review.
Rollback guidance
Restore the prior technical-content authority snapshot. Restore saved configuration and data snapshots, then repeat the smallest verification scenario.