Skip to main content
Interact directly with an assistant inside Devic to test its behavior, see which tools it uses, and perform troubleshooting. This is not the final interface for end users or clients —that is handled through an API endpoint or an embedded widget, explained in the sections API Execution and Other options respectively.
This console is designed for development and fine-tuning environments.

​
Accessing the Console

To access it:
  1. Open the assistant from Devic’s sidebar.
  2. Navigate to the Conversations tab.
  3. Select an existing conversation or create a new one.
Access to the assistant's conversation console Each conversation displays the message history exchanged between the user and the assistant, along with the technical details of each interaction.

​
Main Interface

The console provides a complete view of the conversation and all events generated during it. Main view of the conversation console From here you can:
  • Send test messages to the assistant.
  • Observe how it responds and which tools it uses.
  • Analyze the model’s internal processes in detail.
  • Access the Chat Log to see system events.

​
Tool Calls

When the assistant performs an action (for example, querying a database, performing a web search, or sending an email), Devic displays it through a Tool Call Card.
These cards show which tool was invoked, with what arguments, and what the response was. Example of tool calls executed by the assistant You can expand each card to see the full event information, including the arguments sent and the results returned.

​
Logs and Debugging

The Logs tab provides a detailed view of all events recorded during the conversation, including model API calls, tool calls, partial responses, and errors. Chat Log view with tool call tracking

​

Chat Log view with tool call tracking From this view you can:
  • Filter by event type (LOG, WARN, ERROR, DEBUG).
  • Follow the chronological order of execution.
  • Analyze response times, internal messages, and processed content.
  • View the exact parameters of each tool call.
This helps easily identify:
  • Why a tool didn’t return the expected result.
  • Whether the assistant is correctly interpreting the prompt.
  • When the LLM intervenes and when an external tool is used.

​
Message Analysis

You can inspect every message exchanged (both from the user and the assistant) to review its full JSON structure, including inputs, outputs, and metadata. Message detail with a tool response This feature is essential to understand how the assistant:
  • Processes the received context.
  • Decides which tool to use.
  • Generates the final response shown in the conversation.

​
Troubleshooting Best Practices

  • Review the Logs after every test to identify possible errors in tool calls or MCP configurations.
  • Validate the arguments sent to each tool before deploying to production.
  • Use short and direct conversations during initial tests.
  • Evaluate performance: monitor response times and token usage in each execution

​
Next Steps

Other Options

Explore advanced configurations such as contextual settings, memory, or specialized behaviors.

API Execution

Discover how to launch and manage interactions with your assistants directly through Devic’s public API.