Skip to main content

What is a Trace?

A trace represents a single AI interaction captured from your systems. Every time your AI makes a request, returns a response, or triggers an event, it creates a trace. Traces are the foundation of monitoring in Velatir. They give you visibility into what your AI systems are doing and provide the data for policy evaluation and workflow execution.

Trace Properties

PropertyDescription
Session IDGroups related traces together
DirectionInlet, Response, or Signal
SourceThe system or service that generated the trace
Function NameWhat action was attempted
ArgumentsInput data for the action
MetadataAdditional context
StatusCurrent state (Received, Assessed, Completed, Rejected)

Trace Directions

Traces have three directions that describe the type of interaction:

Inlet

An incoming request to an AI system. This is typically a user input or external trigger. Examples:
  • User sends a message to a chatbot
  • Query sent to an AI search system
  • Request to generate content

Response

An outgoing response from an AI system. This is what the AI produces. Examples:
  • Chatbot’s reply to a user
  • Generated content returned to the client
  • AI’s decision or recommendation

Signal

A passive event or notification that doesn’t require a response. Useful for tracking tool calls, system events, or logging. Examples:
  • AI calling an external tool
  • System health check
  • Audit log entry

Trace Status

StatusDescription
ReceivedTrace just arrived, not yet processed
AssessedPolicies have evaluated the trace
CompletedApproved and finished
RejectedWorkflow rejected the trace

How Traces Are Created

Traces are created automatically through Velatir integrations:
  • Browser Extension - Captures LLM interactions in the browser
  • n8n Node - Creates traces from workflow automation
  • LangChain Middleware - Captures agent tool calls and responses
  • MCP Server - Traces from AI agent interactions
  • SDKs - Programmatic trace creation from your applications

Viewing Traces

In the Velatir dashboard:
  1. Go to Traces to see all traces
  2. Filter by session, source, direction, or status
  3. Click a trace to see full details including:
    • Arguments and metadata
    • Policy assessments
    • Workflow execution history
    • Related traces in the same session

What Happens After Creation

When a trace is created:
  1. Policy Assessment - Active policies evaluate the trace
  2. Workflow Check - If a workflow exists, triggers are checked
  3. Workflow Execution - Matching triggers execute their nodes
  4. Status Update - Trace is marked Assessed, Completed, or Rejected