Each analysis by the Agent is based on the real data returned by tool invocations. The session interface fully displays the name, input parameters, and return results of each tool invocation, making it easy for you to verify the basis for the analysis.
View tool invocations
When the Agent invokes a tool before providing an answer, two nested blocks appear in sequence in the session:
Tool Invocations: Displays the number and names of tools invoked in the current round, for example,
2 tools: resolve_platform_product, ocp_cluster.Tool Output: Displays the invocation status (such as Completed) and a summary of the return result for each tool, for example,
ok: true,product_id: xxxx…,total: 3.
Click a nested block to expand or collapse its details. Click View Full Text below the result summary to view the complete return content of that tool.
Below Tool Output, the Token consumption and model used in the current round are displayed, for example, ↑40k · ↓65 · qwen3.5-plus.
Verify the analysis basis
The Agent references the key data returned by the tools when presenting conclusions. It is recommended to verify the following before adopting any remediation suggestions:
Expand Tool Output to confirm that the tool invocation status is Completed, not failed or timed out.
Use View Full Text to check whether the key metrics cited in the conclusion (such as event waiting ratio, sample count, resource utilization) match the values returned by the tools.
For change-related operations (such as executing a database change ticket), confirm that the target cluster, tenant, and database/table specified in the tool's input parameters match expectations before proceeding with the execution.
Note
The content generated by large models is not entirely correct. Please judge carefully. For operations involving production changes, parameter adjustments, index creation, etc., refer to the original data returned by the tools and execute them within the change window according to your enterprise's change process.
Manual approval
For tool invocations involving data writes or operational changes, the Agent requests confirmation in the session before execution. Please confirm the execution after verifying the tool name and input parameters. If the input parameters do not match expectations, you can reject the execution and add the correct constraints in the dialogue.