Aden TUI Dashboard

The Aden TUI (Terminal User Interface) Dashboard provides a real-time, interactive environment to monitor, debug, and interact with your agents. Built using the Textual framework, it transforms your terminal into a command center for autonomous agent execution.

Launching the Dashboard

The TUI is typically launched via the agent's CLI. If you are using a standard template like the deep_research_agent, you can start the dashboard with:

python -m your_agent_name tui

Note: The TUI requires the textual package. If it is not installed, you can add it via pip install textual.

Interface Layout

The dashboard is divided into three primary functional areas designed to give you total visibility into the agent's "thinking" and "doing":

Graph Overview (Top Left): A live visualization of the agent's node graph.
Log Pane (Bottom Left): Real-time execution logs, including tool calls and internal state changes.
Chat REPL (Right): An interactive console for human-in-the-loop (HITL) communication and agent prompting.
Status Bar (Top): Displays the current graph ID, execution state (idle/running/completed), and elapsed time.

Visualizing the Agent Graph

The Graph Overview renders the agent’s logic as an ASCII Directed Acyclic Graph (DAG). It updates dynamically as the runtime moves through different nodes.

Node Status Indicators

To help you track progress at a glance, nodes use specific symbols and colors:

● (Bold Green): The currently active node being executed.
✓ (Dim): A node that has successfully completed its execution.
■ (Yellow): A terminal node (the end of a specific path).
○ (White): A pending node that has not yet been reached.

Edge Connectivity

Edges represent the transitions between nodes. If a node has multiple outgoing edges (a fan-out), the dashboard displays the branch conditions (e.g., on_success, always) to show you exactly why a specific path was chosen.

Human-in-the-Loop Interaction

The Chat REPL is your direct line to the agent. When an agent enters a node requiring human intervention (such as an approval step or a request for clarification), you can use this pane to:

Provide Context: Answer questions asked by the agent during execution.
Approve Actions: Confirm or reject proposed steps in the workflow.
Inject Commands: Send instructions that influence the agent's next move.

Real-Time Monitoring and Debugging

The Log Pane provides granular visibility into the agent's internal operations. While the Graph View shows where the agent is, the Log Pane shows what it is doing.

Tool Execution: See raw tool inputs and outputs as they happen.
LLM Streams: View the thought process and generated text from the underlying LLM.
Error Reporting: If a node fails, the dashboard captures and displays the error and stack trace immediately, allowing for rapid debugging without digging through flat log files.

Keyboard Shortcuts

Navigate the dashboard efficiently using these global hotkeys: