Using a Skills-compatible agent like Claude Code? The Greenflash agent skill handles this automatically. Run /greenflash-onboard-events instead of copying this prompt.

Add Business Event Tracking to Our Greenflash Integration

You are a coding agent tasked with extending our existing Greenflash Python SDK integration to track business events. We already have message logging working. Your goal is to add event tracking so that Greenflash can connect AI interactions to real business outcomes like conversions, upgrades, and churn.

High-Level Objective

Extend the current implementation to:

Track business events that represent meaningful user milestones
Link events to conversations when the outcome relates to an AI interaction
Assign influence and value to quantify the business impact
Use the correct Python SDK syntax

Reference: The full documentation for the Events API is available at: https://docs.greenflash.ai/features/events

Why Track Events?

Events close the loop between AI interactions and business outcomes. While Greenflash already analyzes conversation quality and sentiment, events provide the ground truth:

Attribute Success: Link specific AI responses directly to business outcomes like upgrades or purchases
Validate Quality: Confirm heuristic signals (like sentiment) with tangible outcome data
Deepen Insights: Train Greenflash to identify conversation patterns that drive real-world results

When to Send Events

Capture moments that matter. Send an event whenever a user reaches a milestone that represents value creation or loss.

Positive Indicators (Success Signals)

signup_completed
trial_started
task_success
upgrade_purchased
meeting_booked
lead_converted

Negative Indicators (Friction/Churn Signals)

cancellation_requested
error_encountered
workflow_abandoned
refund_requested

Contextual Indicators (Usage Context)

feature_viewed
step_completed
usage_logged

Core Event Fields

Field	Type	Required	Description
`event_type`	string	Yes	The name of the event (e.g., `upgrade`, `cancellation_requested`)
`product_id`	uuid	Yes	Your Greenflash Product ID
`conversation_id`	uuid	No	Link to the conversation that influenced this outcome
`influence`	enum	No	`positive`, `negative`, or `neutral` (defaults to `neutral`)
`value`	string	Yes	The measurable value (e.g., `"149.00"`)
`value_type`	string	No	The type of value (e.g., `currency`, `count`)
`properties`	dict	No	Additional context as key-value pairs
`insert_id`	string	No	Unique ID for deduplication

Basic Event Tracking

Sync Example

# Track a successful upgrade after a support conversation
client.events.create(
    event_type="upgrade",
    product_id=product_id,
    conversation_id=conversation_id,  # Link to the AI interaction
    influence="positive",
    value="149.00",
    value_type="currency",
    properties={
        "plan": "pro",
        "billing_cycle": "annual",
    },
)

Async (Fire-and-Forget) Example

# Fire-and-forget so we don't block the main flow
asyncio.create_task(client.events.create(
    event_type="upgrade",
    product_id=product_id,
    conversation_id=conversation_id,
    influence="positive",
    value="149.00",
    value_type="currency",
    properties={
        "plan": "pro",
        "billing_cycle": "annual",
    },
))

Event Examples by Use Case

Customer Support: Post-Chat Upgrade

# User upgraded to Pro after a helpful support chat
client.events.create(
    event_type="upgrade",
    product_id=product_id,
    conversation_id=support_conversation_id,
    influence="positive",
    value="149.00",
    value_type="currency",
    properties={
        "plan": "pro",
        "billing_cycle": "annual",
        "previous_plan": "free",
    },
)

Customer Support: Cancellation Request

# User requested cancellation during a support chat
client.events.create(
    event_type="cancellation_requested",
    product_id=product_id,
    conversation_id=support_conversation_id,
    influence="negative",
    value="49.00",
    value_type="currency",
    properties={
        "reason": "too_expensive",
        "plan": "starter",
        "months_subscribed": 3,
    },
)

Sales: Meeting Booked

# AI-assisted outreach led to a booked meeting
client.events.create(
    event_type="meeting_booked",
    product_id=product_id,
    conversation_id=outreach_conversation_id,
    influence="positive",
    value="50000",
    value_type="currency",
    properties={
        "pipeline_stage": "qualification",
        "engagement_touches": 5,
        "days_since_first_touch": 7,
    },
)

Workflow: Task Completed Successfully

# AI-powered workflow completed a task
client.events.create(
    event_type="task_success",
    product_id=product_id,
    conversation_id=workflow_session_id,
    influence="positive",
    value="1500",
    value_type="numeric",
    properties={
        "task_type": "data_extraction",
        "records_processed": 1500,
        "time_saved_minutes": 45,
    },
)

Workflow: Error Encountered

# AI workflow encountered an error
client.events.create(
    event_type="error_encountered",
    product_id=product_id,
    conversation_id=workflow_session_id,
    influence="negative",
    value="1",
    value_type="numeric",
    properties={
        "error_type": "validation_failure",
        "error_message": "Invalid date format in row 42",
        "records_affected": 1,
    },
)

Linking Events to Conversations

The conversation_id field is crucial for connecting business outcomes to AI interactions. Always include it when:

The event happened during an AI session
The event happened shortly after an AI interaction
There’s a clear causal relationship between the conversation and the outcome

# Store the conversation ID from your message logging
conversation_id = response.conversation_id

# Later, when a business event occurs...
client.events.create(
    event_type="upgrade",
    product_id=product_id,
    conversation_id=conversation_id,  # Links this outcome to the AI interaction
    influence="positive",
    value="149.00",
    value_type="currency",
)

Sampling for High-Volume Events

For high-volume applications, use sampling to control ingestion:

# Sample 10% of page view events
client.events.create(
    event_type="feature_viewed",
    product_id=product_id,
    sample_rate=0.1,  # 10% sampling
    value="dashboard",
    value_type="text",
    properties={
        "feature": "dashboard",
    },
)

# Always capture critical events
client.events.create(
    event_type="upgrade",
    product_id=product_id,
    force_sample=True,  # Never drop this event
    influence="positive",
    value="149.00",
    value_type="currency",
)

Idempotency for Reliable Event Tracking

Use \`insert_id\` to prevent duplicate events when retrying failed requests: \`\`\`python import uuid

Generate a unique ID for this event (store it if you need to retry)

event_insert_id = str(uuid.uuid4()) client.events.create( event_type=“upgrade”, product_id=product_id, conversation_id=conversation_id, influence=“positive”, value=“149.00”, value_type=“currency”, insert_id=event_insert_id, # Same ID = same event (no duplicates) ) \`\`\`

Best Practice: For critical events like purchases, generate the \`insert_id\` before attempting the request and persist it. If the request fails, retry with the same \`insert_id\` to guarantee exactly-once delivery.

Async vs Sync Considerations

When adding event tracking, match our existing async/sync pattern:

Sync apps: Call client.events.create(...) directly
Async apps: Use AsyncGreenflash and wrap calls in asyncio.create_task() for fire-and-forget

Important: In Python, calling an async function without await returns a coroutine that never executes. Always use asyncio.create_task() for fire-and-forget patterns in async code.

Integration Checklist

Your code should:

Identify key business milestones in our application (upgrades, cancellations, task completions, etc.)
Add event tracking at each milestone using client.events.create(...)
Link events to conversations by passing conversation_id when relevant
Set appropriate influence (positive, negative, neutral)
Include value and value_type for revenue-impacting events
Add useful properties for deeper analysis
Use fire-and-forget (asyncio.create_task()) in async code to avoid blocking

What You Must Deliver

Event tracking at key business milestones
Proper conversation linking where applicable
Correct influence assignment (positive/negative/neutral)
Value tracking for revenue-related events
No disruption to existing message logging behavior

This will enable Greenflash to connect AI interactions to real business outcomes, validating conversation quality with tangible results.

Integrating Greenflash

Logging Agentic Workflows

Logging System Prompts

Logging Business Events

Business Events (Python)

Add Business Event Tracking to Our Greenflash Integration

High-Level Objective

Why Track Events?

When to Send Events

Positive Indicators (Success Signals)

Negative Indicators (Friction/Churn Signals)

Contextual Indicators (Usage Context)

Core Event Fields

Basic Event Tracking

Sync Example

Async (Fire-and-Forget) Example

Event Examples by Use Case

Customer Support: Post-Chat Upgrade

Customer Support: Cancellation Request

Sales: Meeting Booked

Workflow: Task Completed Successfully

Workflow: Error Encountered

Linking Events to Conversations

Sampling for High-Volume Events

Idempotency for Reliable Event Tracking

Generate a unique ID for this event (store it if you need to retry)

Async vs Sync Considerations

Integration Checklist

What You Must Deliver

​Add Business Event Tracking to Our Greenflash Integration

​High-Level Objective

​Why Track Events?

​When to Send Events

​Positive Indicators (Success Signals)

​Negative Indicators (Friction/Churn Signals)

​Contextual Indicators (Usage Context)

​Core Event Fields

​Basic Event Tracking

​Sync Example

​Async (Fire-and-Forget) Example

​Event Examples by Use Case

​Customer Support: Post-Chat Upgrade

​Customer Support: Cancellation Request

​Sales: Meeting Booked

​Workflow: Task Completed Successfully

​Workflow: Error Encountered

​Linking Events to Conversations

​Sampling for High-Volume Events

​Idempotency for Reliable Event Tracking

​Generate a unique ID for this event (store it if you need to retry)

​Async vs Sync Considerations

​Integration Checklist

​What You Must Deliver

Add Business Event Tracking to Our Greenflash Integration

High-Level Objective

Why Track Events?

When to Send Events

Positive Indicators (Success Signals)

Negative Indicators (Friction/Churn Signals)

Contextual Indicators (Usage Context)

Core Event Fields

Basic Event Tracking

Sync Example

Async (Fire-and-Forget) Example

Event Examples by Use Case

Customer Support: Post-Chat Upgrade

Customer Support: Cancellation Request

Sales: Meeting Booked

Workflow: Task Completed Successfully

Workflow: Error Encountered

Linking Events to Conversations

Sampling for High-Volume Events

Idempotency for Reliable Event Tracking

Generate a unique ID for this event (store it if you need to retry)

Async vs Sync Considerations

Integration Checklist

What You Must Deliver