Skip to main content

Distributed Tracing

Status

Owner: TBD · Status: draft · Canonical source: flezi.app.frontend.main/src/pages/GovernanceTracing.tsx

Who is this for?​

AI developers investigating agent hallucinations or failed tool calls.

What can you do?​

  • Search execution traces by conversation ID, agent name, or date range.
  • Inspect waterfall latency timelines showing context loading, LLM generation, and sandbox tool execution.
  • Review raw prompt inputs and model outputs.

Steps​

  1. In Governance, select Tracing.
  2. Filter or search for the target trace session.
  3. Click a trace to expand its hierarchical span tree.
  4. Click individual spans (e.g. chat_model_run, tool_call:bash) to view execution parameters and outputs.