get_trace_stats
Analyze CUDA and host operation statistics to identify performance patterns, providing percentile metrics for small datasets and aggregated statistics for large datasets.
Instructions
Get CUDA and host operation statistics. Returns p50/p95/p99 for small DBs (≤500K events), count/avg/min/max from aggregates for large DBs.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| since | No | time range, e.g. 1m, 5m, 1h. Default: all data (0 = no time filter) | |
| tsc | No | telegraphic compression (default: true). Set false for verbose output. |
Implementation Reference
- The MCP tool 'get_trace_stats' is called via the MCPClient class, which forwards the request to the MCP server using the 'tools/call' JSON-RPC method.
def get_trace_stats(self, since: str = "10m") -> dict: """Get trace statistics (120s timeout — aggregation on large DBs).""" return self.call("get_trace_stats", {"since": since, "tsc": False}, timeout=120)