AIML Docs

Debugging and Observability

Monitor and debug your AI agentic workflows with built-in observability features AIML provides built-in observability features to help you monitor and debug your AI agents. By enabling logging, you can track agent interactions, performance metrics, and costs in real-time through the developer dashboard.

​

Getting Started

To observability is enabled by default for all agents.

Using IDE Plugins

Coming soon: AIML IDE plugins for VS Code, Cursor, Vim, Emacs and more!

Using the developer UI

To start the developer UI locally:

  1. Navigate to the aiml-ui directory:
cd apps/aiml-ui
  1. Copy the example environment file and add your Firework API key:
cp .env.example .env

Edit .env and add your Firework API key.

  1. Install dependencies and start the development server:
npm install
npm run dev

The developer UI will be available at http://localhost:3000.

Features

Logs

The developer UI provides comprehensive logging capabilities:

  • Real-time log streaming for all agent interactions
  • Log filtering by severity (INFO, WARN, ERROR)
  • Search and filter logs by timestamp, agent ID, or content
  • Log context expansion to see full interaction details
  • Export logs for offline analysis

Traces

Trace and analyze agent execution paths:

  • Detailed execution traces showing each step
  • Timing information for performance analysis
  • Input/output data at each step
  • Error traces with full stack information
  • Trace visualization with flame graphs

Dashboards

Monitor your agents with real-time dashboards:

  • Agent performance metrics
  • Token usage and cost tracking
  • Response time distributions
  • Error rate monitoring
  • Custom metric visualization

Workflow Visualization

Watch your agentic workflows in action:

  • Interactive workflow graph visualization
  • Real-time state updates
  • Step-by-step execution playback
  • Visual debugging tools
  • State inspection at any point
  • Branch and parallel execution visualization

What’s Being recorded

When observability is enabled, AIML automatically tracks:

  • Agent interactions and their outcomes
  • Model usage and token consumption
  • Cost metrics per interaction
  • Response times and latency
  • Evaluation steps and reasoning
  • Step executions and their results
  • Errors and failure cases ​

AIML Syntax

Learn about the core syntax features of AIML

IDE Plugins

Next Page

On this page