โ Back to Projects
P-010AI Observability Platform
Build a self-hosted AI observability platform that traces LLM calls, tracks costs, monitors latency, evaluates quality, and alerts on anomalies. Integrates with OpenTelemetry for vendor-neutral instrumentation and provides a dashboard for real-time monitoring. Think of it as a mini Langfuse โ purpose-built for your LLM applications.
โฑ๏ธ 6h 40m โ 9hโญ 600 XP๐ ai observability
Skills
Distributed tracingCost trackingLatency analysisQuality evaluationDashboard developmentOpenTelemetryAlerting systemsFull-stack development
Tech Stack
Next.jsReactTypeScriptPythonSQLiteChart.jsTailwind CSS
Deploy To
๐ Docker๐ Railway๐ Local
What You'll Learn
- โBuild a trace ingestion API that accepts OpenTelemetry spans
- โImplement per-request cost calculation with model pricing tables
- โCreate a real-time monitoring dashboard with cost, latency, and quality metrics
- โBuild an automated evaluation pipeline using LLM-as-judge
- โSet up configurable alerts for cost spikes, latency regressions, and quality drops
- โDesign a trace viewer with span hierarchy visualization
Prerequisites
๐LessonโNot started
Langfuse Deep Dive
intermediateโฑ๏ธ 22mai observability
๐LessonโNot started
Metrics, Dashboards, and Alerting
intermediateโฑ๏ธ 18mai observability
๐LessonโNot started
OpenTelemetry for LLMs
advancedโฑ๏ธ 20mai observability
๐งWorkshopโNot started
W-020LLM Observability Dashboard Builder
intermediateโฑ๏ธ 50mโ1h 15mai observability
Rank 08