Skip to content

AetherGraph UI Overview

The AetherGraph UI gives you a visual control panel for your graphs, agents, and runs:

  • Launch apps from an App Gallery (one-click flows).
  • Chat with agents that orchestrate graphs behind the scenes.
  • Inspect runs, sessions, and artifacts.
  • Resume failed or partial runs (for graphify graphs).

This section of the docs covers:

  1. Start the server with the built-in UI
  2. Expose agents and apps to the UI
  3. Important notes and troubleshooting tips

If you only use AetherGraph as a Python library (no UI), you can ignore this whole section and just import and run graphs from your own code.

What the UI actually does

At a high level:

  • The server exposes:

    • A REST API: /api/v1/…
    • A static UI bundle: /ui

  • The frontend reads metadata about:

    • Graphs (@graphify) and graph functions (@graph_fn)
    • Anything marked with as_app or as_agent

  • The UI then lets you:

    • Start runs (apps)
    • Start chat sessions (agents)
    • Inspect runs, sessions, memory, and artifacts

You control what appears in the UI by how you define your project module, apps, and agents. The next pages walk you through:

When should I use the UI?

Good use cases for the UI:

  • You want non-engineers (or future you) to use your graphs without reading code.
  • You want a visual “lab notebook”: runs, artifacts, and memory all tracked in one workspace.
  • You want to debug complex flows with better observability and resumption.

Cases where the UI is optional:

  • Fully automated backends where you call AetherGraph purely as a library.
  • One-off scripts or small experiments where a CLI is enough.

If you’re unsure, start with the UI. It’s usually easier to wire up a graph once and then reuse it from both UI and code.