AI Observability
Langfuse-style tracing for every LLM call — tokens, cost, latency, fallbacks, evaluation and compliance scores, correlated to tenant and job.
8 traces
Total cost (traces)
$1.11
Avg latency
13.9s
Fallbacks used
1
Errors
1
Cost by provider
Spend attributed per AI provider
Latency by model
Average response time (ms)
Error rate by provider
% of failed calls
Hallucination risk over time
Lower is better (7-day trend)
Compliance score by tenant
Weighted compliance evaluation
Token cost composition
Provider share of monthly token spend
Trace explorer
Every LLM invocation, end to end.