ai-cli-observability
Self-hosted OpenTelemetry observability stack for Codex, Claude Code, and Gemini CLI with Grafana dashboards. Track usage, latency, errors, tokens, tool activity, logs, metrics, and traces without sending data to another SaaS.
Ficha actualizada 23 sept 2026, 9:03 · upstream 23 sept 2026, 8:36 · release sin release detectada
Qué es ai-cli-observability
Descripción editorial construida a partir del repositorio, su documentación y sus señales de mantenimiento.
AI CLI Observability provides a self-hosted solution for monitoring AI coding assistants including Codex, Claude Code, and Gemini CLI. It aggregates telemetry from these tools into a single Grafana dashboard, enabling comprehensive visibility into token consumption, API costs, request latency, error rates, tool usage patterns, and trace-level observability. The stack is built on OpenTelemetry with VictoriaMetrics for metrics storage and Grafana for visualization, allowing teams to understand where their AI development time and resources are spent without relying on external providers.
- Monitor token consumption across multiple AI models to identify cost outliers
- Track request latency per model to optimize API usage patterns
- Detect tool activity anomalies that may indicate misconfiguration or misuse
- Visualize error rates and failure patterns for incident response
- Understand which tools are actually being used versus what is configured
- Single Compose stack with Docker Compose v2, requiring no additional infrastructure
- OpenTelemetry-native implementation supporting logsQL, Prometheus remote write, and VictoriaTraces
- Native Claude Code cost metrics without requiring external pricing APIs
- Live USD-to-EUR conversion for dashboard totals through the VictoriaMetrics exchange rate
- Git AI attribution dashboards with code-level tool activity visibility
- Pinned container images ensure reproducible builds across environments
- Prompt logging is disabled by default; model names, session IDs, and trace attributes may still leak operational context
- Storage backends (VictoriaMetrics, VictoriaLogs) must be configured within the Compose network for full access
- Remote agent deployments require a TLS-terminating reverse proxy at collector port 4318 with restricted network access
- Historical data retention is limited to what is exported; new telemetry may not be fully recoverable from old exports
Perfil de infraestructura
Estimación inicial derivada del repositorio y de los servicios detectados.
Despliegue detectado
Qué señales reales de instalación hemos encontrado.
Deploy via Docker Compose v2. Use the .env.example template and generate two secrets: OTEL_AUTH_TOKEN for collector authentication and GRAFANA_ADMIN_PASSWORD for Grafana access. The stack requires Python 3.11+ only for smoke testing; production use requires a full environment with all dependencies. For remote agent deployments, place a TLS-terminating reverse proxy in front of port 4318 and restrict VictoriaMetrics, VictoriaLogs, and VictoriaTraces to the same network segment as the collector. Always verify Grafana dashboard panels against sanitized data before publishing.
Stack y dependencias
Componentes inferidos a partir de manifests, configuración y documentación.
Puede sustituir a
Relaciones detectadas por el análisis del repositorio y su categoría.
Experiencia real de la comunidad
Ayuda a contrastar la estimación automática con instalaciones reales.
Todavía no hay datos comunitarios para este proyecto.