Skip to main content

Logs and Events

  • Updated

Index 

  1. Introduction 
  2. How to access Logs and Events 
  3. How does it work? 
  4. Activation 
  5. Contacts and contexts 
  6. Tracking Mode 
  7. Number of tokens

 

Introduction 

The feature was developed with the goal of assisting Blip users in debugging their flows. With "Logs and Events," it is possible to observe the current flow of the smart contact, starting from the user's input.

For example: You can analyze a specific user and follow their entire journey: contact, context, time elapsed in each action, and the complete raw log of the actions performed at each step.   

 

How to access Logs and Events 

On the Studio screen: 

  • In the upper right corner of the screen, click the icon . 

  • In the menu that opens, select Logs and events.

The following screen will open:

How does it work? 

"Logs and Events" present the following functionalities:

  • Contact tracing

  • Routing verification

  • Error tracking

  • Content of context variables

  • Clear the context of a specific user

This feature is supported on any channel that has an integration with the Blip platform.  

 

Activation 

By activating ‘Logs and Events,’ you will have access to the monitoring of messages handled by the bot; this way, it is possible to view the content of each one.

Usage examples 

On the Logs and Events screen, it will be possible to view all messages handled in the contact where Tracing is active. To debug a flow, still on the logs and events screen, you can go to the laboratory icon, then to test published chatbot or chatbot under construction. In the message monitoring, it is possible to apply some filters:

  • Channel

  • User Id

From there, you will see the monitoring screen:

By clicking on a block, it is possible to view all the flow execution information such as:

  • Script execution information;

  • Event Registration (Trackings);

  • HTTP Requests;

  • Model, prompt, message history, and tools used (agent block)

The Logs and Events functionality is a powerful tool for understanding and enhancing your agent's behavior. With it, you can see in detail the context that reaches the agent, the tools available at the moment, the prompt used, the response generated by the agent, and the return from the triggered tools. This comprehensive view allows for identifying improvement points and adjusting the agent for more effective performance.

 

Contacts and contexts 

By clicking on the bubble containing the message, it is possible to view and edit all user information, context variables, and contact variables.

Creation, editing, and deletion of context variables

Tracking Mode 

Tracking mode is a functionality that allows you to view trackings (or TrackEvents) in an easy way, without having to search through the JSON of the entire process. To activate Tracking Mode, simply click the button labeled Tracking Mode on the top bar of the screen. When the messages appear, you will see the information below each block.

After activating, you will be able to view the trackings for each block and have access to the category, action, and extra variables. 

It is also possible to hide blocks that do not have trackings by clicking the button with the question mark icon in the top right of the screen.

 

Number of tokens 

Understanding Token Usage in Logs 

To help you monitor the performance and cost of your agent, Studio displays token consumption on the Logs and Events screen.

What are Tokens? 

Think of tokens as "puzzle pieces" or the syllables of a word. For Artificial Intelligence, everything it reads and writes is transformed into these small units so that it can process the information.

Where to view? 

You can find this information directly in the agent block or in details by clicking on the block and accessing the "usage" field.

Detailed Information Breakout 

When analyzing consumption, you will see four main indicators:

  • Input Tokens: This is everything the agent needs to "read" to understand what it should do. This includes its system instructions, available tools, and the conversation history (short-term memory).

  • Input Cached: This is like a "quick memory." It is information that the model has already processed recently and saved to save time and effort, making the response faster.

  • Output Tokens: This is the amount of text the agent "wrote" to respond to the user. You can limit this size in the agent settings via the Max Tokens parameter.

  • Total Tokens: This is the sum of everything processed (Input + Cached + Output).

Practical Tip: As the user chats with the agent, the number of tokens tends to grow because the agent needs to "reread" the accumulated history to maintain the conversation context.   

How to analyze in detail? 

If you want a complete technical view, click on the agent block within the Log and look for the usage section. There, the term "completion_tokens" refers exactly to the Output Tokens (the generated response).

 

For more information, visit the discussion on the subject at our community or videos on our channel. 😃

Related articles