Skip to content
Karyam

AI Flow Logs

AI flow Logs provide complete visibility into how an AI Flow executes.

Every flow execution generates a sequence of structured events that allow teams to understand:

  • Which step is currently running
  • Which tools were executed
  • Whether approvals were requested
  • Why execution paused
  • When execution resumed
  • Why a AI flow failed

This makes AI flows observable, debuggable, and auditable.

AI flow Event Lifecycle

Section titled “AI flow Event Lifecycle”

A typical execution may look like this:

REQUEST
    ↓
TOOL_EXECUTION_START
    ↓
TOOL_EXECUTION_END
    ↓
STEP_FINISH
    ↓
APPROVAL
    ↓
PAUSED
    ↓
RESUME
    ↓
RESPONSE

AI flow Logs

Section titled “AI flow Logs”

AI flow Logs provide complete visibility into how an AI Flow executes.

Every flow execution generates a sequence of structured events that allow teams to understand:

  • Which step is currently running
  • Which tools were executed
  • Whether approvals were requested
  • Why execution paused
  • When execution resumed
  • Why a AI flow failed

AI flow Logs can be accessed from:

OPERATE
 └── Runs
      └── Logs

This makes AI flows observable, debuggable, and auditable.

Supported Events

Section titled “Supported Events”

REQUEST

Section titled “REQUEST”

Indicates the beginning of a AI flow execution.

Example:

REQUEST
Flow: VPN Access Restoration
Run ID: flow_01HABC123
Status: RUNNING

TOOL_EXECUTION_START

Section titled “TOOL_EXECUTION_START”

Generated when a AI flow step starts executing a tool.

Examples include:

  • PostgreSQL Queries
  • Email Notifications
  • Webhooks
  • MCP Tools
  • Vector Search

Example:

TOOL_EXECUTION_START
Tool: PostgreSQL
Operation: Employee Lookup
Status: EXECUTING

TOOL_EXECUTION_END

Section titled “TOOL_EXECUTION_END”

Generated when tool execution completes.

Example:

TOOL_EXECUTION_END
Tool: PostgreSQL
Duration: 245ms
Status: SUCCESS

STEP_FINISH

Section titled “STEP_FINISH”

Generated when a AI flow step completes.

Example:

STEP_FINISH
Step: Verify Employee Eligibility
Status: SUCCESS

AUTH_REQUIRED

Section titled “AUTH_REQUIRED”

Generated when a connected integration requires authentication before execution can continue.

Example:

AUTH_REQUIRED
Integration: GitHub MCP
Reason: OAuth token expired
Status: WAITING

Human Approval Events

Section titled “Human Approval Events”

APPROVAL

Section titled “APPROVAL”

Generated when a AI flow requests human approval.

Example:

APPROVAL
Approval Type: Manager Approval
Requested For: VPN Restoration
Status: REQUESTED

PAUSED

Section titled “PAUSED”

Generated when AI flow execution pauses while waiting for approval.

Example:

PAUSED
Reason: Waiting for manager approval
Status: WAITING_APPROVAL

RESUME

Section titled “RESUME”

Generated when AI flow execution resumes after approval.

Example:

RESUME
Approved By: John Doe
Status: RESUMED

Approval Outcomes

Section titled “Approval Outcomes”

Approvals may result in:

APPROVED
REJECTED

Example:

APPROVAL
Status: APPROVED

or

APPROVAL
Status: REJECTED

AI flow Statuses

Section titled “AI flow Statuses”

Each AI flow execution maintains a current status.

Execution Statuses

Section titled “Execution Statuses”
Status Description
INFO Informational event
SUCCESS Operation completed successfully
FAILED Execution failed
REQUESTED Action has been requested
EXECUTING Currently executing
RUNNING AI flow is active
WARNING Non-blocking issue detected
WAITING Waiting for external input
TRANSFERRED Execution transferred
CANCELLED AI flow cancelled
WAITING_APPROVAL Waiting for human approval

Approval Statuses

Section titled “Approval Statuses”
Status Description
APPROVED Request approved
REJECTED Request rejected
RESUMED AI flow resumed
PAUSED AI flow paused

Example AI flow Timeline

Section titled “Example AI flow Timeline”
10:01:02 REQUEST


10:01:03 TOOL_EXECUTION_START
Employee Lookup


10:01:03 TOOL_EXECUTION_END
SUCCESS


10:01:04 STEP_FINISH
Employee Verification


10:01:05 APPROVAL
Manager Approval Requested


10:01:05 PAUSED
Waiting for Approval


10:03:20 RESUME
Manager Approved Request


10:03:21 TOOL_EXECUTION_START
Create Ticket


10:03:22 TOOL_EXECUTION_END
SUCCESS


10:03:23 RESPONSE


10:03:24 STEP_FINISH


10:03:25 SUCCESS

Example: VPN Restoration AI flow

Section titled “Example: VPN Restoration AI flow”
Employee Request
        ↓
REQUEST
        ↓
Verify Employee Status
        ↓
TOOL_EXECUTION_START
        ↓
TOOL_EXECUTION_END
        ↓
APPROVAL
        ↓
PAUSED
        ↓
RESUME
        ↓
Create IT Ticket
        ↓
RESPONSE

Why AI flow Logs Matter

Section titled “Why AI flow Logs Matter”

AI flow Logs provide:

  • Execution visibility
  • Approval tracking
  • Operational debugging
  • Compliance auditing
  • Production observability

This allows organizations to operate AI AI flows with confidence.

Section titled “Related Observability Features”

AI flow Logs work alongside:

  • Agent Logs
  • RAG Logs
  • Execution Traces
  • Token Usage
  • Cost Analytics

Together they provide complete visibility into production AI systems.

Next Steps

Section titled “Next Steps”

Continue with:

  • RAG Logs
  • Execution Traces
  • Token Usage
  • Cost Analytics
  • Debugging