Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.monobot.ai/llms.txt

Use this file to discover all available pages before exploring further.

Debug Mode allows developers and operators to monitor the internal bot execution process in real time directly from the platform interface. While the assistant processes user requests, logs are streamed into the Live Debug Feed, showing step-by-step execution details including tool calls, actions, execution time, responses, warnings, and errors.
Debug Mode

​
Enabling Debug Mode

Debug Mode can be enabled directly from the bot Flow editor. Navigate to: 
Bot Configuration → Flow
Enable the Debug mode checkbox in the top-right corner of the Flow editor interface.
Debug logs will start appearing in the Live Debug Feed only after Debug Mode is enabled.

​
What is Debug Mode

Debug Mode is a real-time debugging interface that helps monitor:
  • Active flows
  • Tool execution
  • Action execution
  • LLM requests and responses
  • Errors and warnings
  • Custom logs from Function actions
  • Execution timing and performance
Logs are streamed directly from the backend and displayed live without refreshing the page.

​
Why Use Debug Mode

ScenarioWhat Debug Mode Provides
Bot gives unexpected responseShows active flows, tools, and passed arguments
Action works slowlyShows execution time for every action
Need to inspect action resultShows full JSON result returned by actions
LLM behaves unexpectedlyShows model name, API version, and full response
Tool or action errorShows [ERROR] and [WARNING] messages
Developing custom Function actionsShows custom logs created through interaction["_LOGGING"]

​
Live Debug Feed

The Live Debug Feed displays all interaction logs in chronological order. Each log entry contains:
FieldDescription
TimestampTime in HH:MM:SS.mmm format
Level[INFO], [WARNING], [ERROR], [DEBUG]
Event Type[CHAT], Action, Tool, or system event
MessageShort description of the event
Clicking a log entry opens additional details including full JSON payloads, arguments, and execution results.

​
Session Start Example

16:37:02.024  [INFO]  [CHAT]  Session started
16:37:02.029  [INFO]          Active flow: "Root Flow"
Shows the interaction start and currently active flow.

​
User Request Example

16:37:12.196  [INFO]  [CHAT]  Request  "propan"
                              Tools (1): "commodity_code_search"
Shows the incoming user request and available tools provided to the model.

​
Tool Execution Example

16:37:13.589  [INFO]  [CHAT]  Tool  Running tool "commodity_code_search"
                              Arguments: product_name: Liquefied propane...
                              Tool description: Use this tool to search for...
Displays:
  • Tool name
  • Arguments passed by the LLM
  • Tool description
  • Tool execution start

​
Action Execution Example

16:37:13.946  [INFO]  Action  Action "reasoning_rag"
                              Condition: "PASS"
                              Execution time: 14.106 sec
                              Result: {"answers": [...], "found_digit_code": null}
Each action execution contains:
FieldDescription
ConditionWhether execution condition passed (PASS or SKIP)
Execution timeTime spent executing the action
ResultFull JSON response returned by the action

​
Tool Completion Example

16:37:28.294  [INFO]          Tool execution finish: 14.705 sec
Shows total execution time for the entire tool execution chain.

​
Bot Response Example

16:37:29.183  [INFO]  [CHAT]  Response  "Which best describes..." in 16.982 sec
                              API version: Responses
                              Model name: Openai "gpt-5.2 Chat"
Displays:
  • Final assistant response
  • Total generation time
  • API provider version
  • Model name

​
Custom Logs in Function Actions

Function actions can send custom logs directly into Debug Mode using interaction["_LOGGING"].

​
Example

def my_function(tool_params, interaction):
    interaction["_LOGGING"].info("Starting product lookup")
    interaction["_LOGGING"].warning("Product not found, using fallback")
    interaction["_LOGGING"].error("External API returned 500")
    interaction["_LOGGING"].debug("Raw response: " + str(raw))
These logs appear together with system logs inside Live Debug Feed.

​
Logging Levels

LevelDescription
INFOStandard execution information
WARNINGNon-critical issue or fallback
ERRORCritical execution problem
DEBUGDetailed developer debugging information

​
Full Debug Flow Example

16:37:02.024  [INFO]  [CHAT]  Session started
16:37:02.029  [INFO]          Active flow: "Root Flow"

16:37:12.196  [INFO]  [CHAT]  Request  "propan"
                              Tools (1): "commodity_code_search"

16:37:13.589  [INFO]  [CHAT]  Tool  Running tool "commodity_code_search"

16:37:13.946  [INFO]  Action  Action "reasoning_rag"
                              Condition: "PASS"
                              Execution time: 14.106 sec

16:37:28.294  [INFO]          Tool execution finish: 14.705 sec

16:37:29.183  [INFO]  [CHAT]  Response  "Which best describes your propane shipment?"
                              Model name: Openai "gpt-5.2 Chat"

​
Technical Details

  • Logs are streamed through Redis pub/sub.
  • Events are stored in events:{interaction_uuid} queue with TTL.
  • Each log contains:
    • id
    • timestamp
    • level
    • message
    • optional details
    • optional nested children
  • Nested logs are displayed in UI as expandable trees.
  • Logs remain available during TTL lifetime after interaction completion.

​
Notes

  • Debug Mode is intended for development and troubleshooting.
  • Full responses and action results may contain sensitive information.
  • Excessive debug logging may affect readability during large interactions.
  • Custom Function logs are supported only inside Function-type actions.