Skip to main content

Observability Overview

Taimoe Enterprise AI Gateway provides comprehensive observability into all LLM API calls, agent reasoning steps, and conversation streams routed through your organization. This suite ensures full transparency for cost control, performance optimization, and issue debugging.

The Observability suite comprises five specialized modules accessible under OBSERVABILITY in the Aegis Console:

Observability Modules​

1. Usage & Cost​

Macro-level analytics on API request throughput, token consumption, failure rates, and spend attribution sliced across Teams, Agents, Virtual Keys, and Runtimes.

2. Requests​

Real-time data plane log stream recording every individual LLM API call, including HTTP status codes, latency, model usage, and calculated cost.

3. Conversations​

Full transcript viewer and audit history for multi-turn chat sessions across API, Playground, and Teams channels, with Guardrail blocked-turn indicators and CSV/JSON export options.

4. Traces​

OpenTelemetry-compliant distributed tracing for multi-step agent reasoning, tool calls, and vector retrievals, complete with interactive span waterfall timelines.

5. Metrics​

Real-time visual dashboards charting request throughput, latency trends, token ratios (Prompt vs Completion), and top active agents.

Core Performance Indicators (KPIs)​

Across Observability dashboards, key performance metrics provide immediate operational visibility:

  • Total Requests / Traces: Total volume of API calls or agent executions handled by the gateway.
  • Fail Rate (%): Percentage of failed operations. Healthy systems maintain this near zero.
  • Avg Latency (ms): Average end-to-end processing time in milliseconds.
  • Total Tokens & Cost: Aggregate prompt and completion tokens processed and their estimated USD cost.