agent-metrics-osiris
par nantes
DevOps & CloudRésumé
Observability and metrics for AI agents - track calls, errors, latency.
Description
Ce que fait ce skill
A Python/PowerShell tool that tracks API calls, errors, latency, and system resource usage for AI agents. Outputs a terminal dashboard and exports metrics to JSON.
Captures granular per-label metrics locally without requiring an external observability platform.
Quand l'utiliser
- Tracking error rates across multiple agent runs
- Comparing latency between different API providers
- Monitoring CPU and memory usage during long agent tasks
- Exporting metrics snapshots for post-run analysis
- Auditing call volume to stay within API rate limits
Exemple de workflow
User asks: show me a summary of my agent's performance after a long run
- 1
Record each API call with
record --type call --label api_openai - 2
Record any errors with
record --type error --label api_error --details ... - 3
Record task latency with
record --type latency --label task_process --value 1500 - 4
Run
dashboardaction to view totals, error rate, and avg latency - 5
Run
export --format jsonto save metrics to a file
Terminal dashboard showing 1,247 calls, 1.84% error rate, 234ms avg latency, plus a metrics.json file
Installation
clawhub install nantes/agent-metrics-osirisExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.