Tencent Cloud Observability API Routing Assistant

Problem

Tencent Cloud observability queries are spread across basic monitoring, alarms, APM, RUM, synthetic testing, load testing, Prometheus, and Grafana. Several modules share the tccli monitor service name, so a request like “APM alarms” or “instance metrics” can easily route to the wrong module, use the wrong Namespace, or miss the region, returning empty or irrelevant data.

How It Works

The skill identifies user intent, loads only the relevant references, and routes each request to the right module. For standard public actions, it prefers tccli; for console-internal APIs such as mongodb:DescribeDBInstanceSummaries, it uses the Python SDK CommonClient to reach the backend gateway. Simple scalar parameters are passed on the command line, while arrays or nested objects are written to a temporary JSON file via --cli-input-json. It confirms critical fields when ambiguous, including --region, Namespace, instance IDs, and time windows. Credential failures follow a refresh, interactive login, fallback flow without echoing SecretId or SecretKey.

Boundaries

The skill is read-only. When it detects Create*, Modify*, Delete*, or similar write actions, it does not construct commands and points users to the console instead. It depends on tccli and tencentcloud-sdk-python; Windows may need a python3 alias. It is best for querying monitoring data, troubleshooting alarms, and checking observability metadata, not for changing resources, starting load tests, or deleting configurations.