Understanding Observability

About

LLM observability is the practice of capturing, structuring, and analyzing everything that happens inside your AI application. Every LLM call, retrieval, tool execution, and agent decision is recorded as structured data that you can search, filter, score, and alert on.

Future AGI’s observability stack is built on OpenTelemetry. Your application sends traces to the platform, and everything else (dashboards, evals, sessions, alerts) runs on top of that traced data. Without tracing, there is nothing to observe.

---

The Tracing Pipeline

Your app emits spans (LLM calls, tool calls, chain steps) via OpenTelemetry or the traceAI SDK. The backend receives them over HTTP or gRPC, groups them into traces, and stores them by project.

Your App → traceAI / OpenTelemetry SDK → OTLP (HTTP or gRPC) → Future AGI Backend → Observe Dashboard

Each trace is one request or execution. Each span is one operation (LLM, tool, retriever, etc.) with input, output, timing, and optional cost and tokens. That data powers the entire UI: trace list, span detail, sessions, evals, and alerts.

---

Key Concepts

Concept	What it is	Learn more
Traces	A group of spans representing one complete request flow from input to output.	What are Traces?
Spans	A single operation (LLM call, retrieval, tool execution). Records inputs, outputs, timing, and errors.	What are Spans?
OpenTelemetry	The open standard used to collect and export trace data.	What is OpenTelemetry?
traceAI	Future AGI’s instrumentation library that wraps OpenTelemetry for LLM-specific spans.	What is traceAI?

---

How It Works

1. Instrument your app: Add a traceAI instrumentor (or use manual spans) to capture LLM calls automatically
2. Traces flow to the platform: Data is exported via OTLP to Future AGI’s backend
3. Everything is available in the dashboard: Trace list, span detail, sessions, evals, and alerts all run on top of traced data

---

Next Steps