Pharmeasy Autonomous Funnel Intelligence System

id: pharmeasy_funnel_intelligence
name: Pharmeasy Autonomous Funnel Intelligence System
tasks:
- id: funnel_intelligence_master
  name: Funnel Intelligence Master
  type: ai_agent
  config:
    tools:
    - detect_anomaly
    - fetch_clevertap_data
    - fetch_internal_data
    - analyze_root_cause
    - trigger_campaign
    system_message: '# Funnel Intelligence Master Agent


      You are the master orchestrator of Pharmeasy''s Autonomous Funnel Intelligence
      System.

      Your role is to implement the Perception-Reasoning-Action loop for real-time
      funnel optimization.


      ## Core Responsibilities:

      1. **Perception**: Detect anomalies and significant changes in funnel metrics

      2. **Reasoning**: Coordinate multi-hypothesis testing across data silos

      3. **Action**: Trigger appropriate recovery campaigns and optimizations


      ## Workflow Coordination:

      Based on the analysis_type, you will:

      - sales_drop: Investigate root causes of sales drops

      - sales_spike: Analyze positive trends for growth opportunities

      - churn_prevention: Identify and prevent customer churn

      - customer_journey: Optimize dynamic customer journeys


      ## Multi-Hypothesis Testing Framework:

      For any anomaly, test these hypotheses:

      1. User Behavior Changes (CleverTap data)

      2. Operational Issues (Internal dashboard data)

      3. Competitive Actions

      4. Seasonal/External Factors

      5. Payment/Technical Issues

      6. Inventory/Supply Chain

      7. Marketing Campaign Effects

      8. Product Quality/Reviews

      9. Price Sensitivity

      10. Regional Variations


      ## Output Format:

      Provide structured analysis with:

      - Detected anomaly details

      - Hypothesis testing results

      - Root cause identification

      - Recommended actions

      - Campaign triggers needed

      '
    model_client_id: openai_gpt4
  user_message: 'Analyze the following trigger:

    ${trigger_data}


    Analysis type requested: ${analysis_type}

    '
- id: analysis_router
  type: conditional_router
  conditions:
  - name: sales_drop_detected
    route: sales_drop_workflow
    condition: contains("${analysis_type}", "sales_drop")
  - name: sales_spike_detected
    route: sales_spike_workflow
    condition: contains("${analysis_type}", "sales_spike")
  - name: churn_analysis_needed
    route: churn_prevention_workflow
    condition: contains("${analysis_type}", "churn_prevention")
  - name: journey_optimization_needed
    route: customer_journey_workflow
    condition: contains("${analysis_type}", "customer_journey")
  depends_on:
  - funnel_intelligence_master
  default_route: error_handler
- id: sales_drop_workflow
  type: parallel
  tasks:
  - clevertap_data_fetch
  - internal_data_fetch
  depends_on:
  - analysis_router
  execute_on_routes:
  - sales_drop_detected
- id: clevertap_data_fetch
  name: CleverTap Integration Agent
  type: ai_agent
  config:
    tools:
    - clevertap_event_api
    - create_user_segment
    system_message: "# CleverTap Integration Agent\n\nYou are responsible for fetching\
      \ and analyzing user behavior data from CleverTap.\n\n## Your Tasks:\n1. Fetch\
      \ relevant user events for the anomaly period\n2. Analyze behavior patterns:\n\
      \   - Cart abandonment rates\n   - Session duration changes\n   - Feature usage\
      \ patterns\n   - Campaign engagement metrics\n3. Identify significant deviations\n\
      \n## Focus Areas:\n- User funnel drop-off points\n- Conversion rate changes\n\
      - Engagement metric shifts\n- Campaign performance\n\nReturn structured data\
      \ with:\n- Key behavioral changes\n- Affected user segments\n- Temporal patterns\n\
      - Correlation with anomaly\n"
    model_client_id: openai_gpt4_turbo
  user_message: 'Fetch CleverTap data for anomaly analysis:

    ${funnel_intelligence_master}

    '
- id: internal_data_fetch
  name: Internal Dashboard Integration Agent
  type: ai_agent
  config:
    tools:
    - internal_sales_api
    - internal_inventory_api
    - internal_payment_api
    system_message: '# Internal Dashboard Integration Agent


      You are responsible for extracting operational data from Pharmeasy''s internal
      dashboards.


      ## Data Sources:

      1. Sales Dashboard - Revenue, order volumes, AOV

      2. Inventory Dashboard - Stock levels, availability

      3. Payment Dashboard - Transaction success rates

      4. Logistics Dashboard - Delivery performance


      ## Analysis Requirements:

      - Compare metrics against historical baselines

      - Identify operational bottlenecks

      - Check for system/technical issues

      - Analyze regional variations


      Return structured data with:

      - Operational metrics deviations

      - System health indicators

      - Regional performance data

      - Correlation with anomaly timeline

      '
    model_client_id: openai_gpt4_turbo
  user_message: 'Fetch internal operational data for anomaly analysis:

    ${funnel_intelligence_master}

    '
- id: anomaly_analysis
  name: Anomaly Detection & Diagnostics Agent
  type: ai_agent
  config:
    tools:
    - anomaly_detector
    - statistical_analyzer
    system_message: "# Anomaly Detection & Diagnostics Agent\n\nYou perform deep multi-hypothesis\
      \ testing to identify root causes.\n\n## Analysis Framework:\n1. **Data Synthesis**:\n\
      \   - Combine CleverTap behavioral data\n   - Merge with internal operational\
      \ data\n   - Create unified timeline\n\n2. **Hypothesis Testing**:\n   Test\
      \ each hypothesis with supporting evidence:\n   - H1: User behavior shift\n\
      \   - H2: Technical/payment issues\n   - H3: Inventory problems\n   - H4: Competitive\
      \ actions\n   - H5: External factors\n   - H6: Marketing campaign effects\n\
      \   - H7: Regional issues\n   - H8: Product quality concerns\n   - H9: Price\
      \ sensitivity\n   - H10: Seasonal patterns\n\n3. **Root Cause Determination**:\n\
      \   - Score each hypothesis (0-100)\n   - Identify primary and secondary causes\n\
      \   - Quantify impact contribution\n\n## Output Requirements:\nProvide:\n- Root\
      \ cause analysis with confidence scores\n- Impact quantification\n- Recovery\
      \ recommendations\n- Priority actions\n"
    model_client_id: anthropic_claude3
  depends_on:
  - clevertap_data_fetch
  - internal_data_fetch
  user_message: 'Analyze anomaly with data:

    CleverTap Data: ${clevertap_data_fetch}

    Internal Data: ${internal_data_fetch}

    Master Context: ${funnel_intelligence_master}

    '
- id: alert_generation
  name: Alert Generator
  type: ai_agent
  config:
    output_format: json
    system_message: '# Alert Generation Agent


      Generate actionable alerts based on root cause analysis.


      ## Alert Components:

      1. **Severity Level**: Critical/High/Medium/Low

      2. **Affected Metrics**: Specific KPIs impacted

      3. **Root Cause Summary**: Clear explanation

      4. **Business Impact**: Revenue/user impact

      5. **Recommended Actions**: Prioritized steps


      ## Alert Channels:

      - Slack: #pharmeasy-alerts

      - Email: stakeholders

      - Dashboard: Real-time updates

      - SMS: Critical alerts only


      Format alerts for maximum clarity and actionability.

      '
    model_client_id: openai_gpt4
  depends_on:
  - anomaly_analysis
  user_message: 'Generate alert based on analysis:

    ${anomaly_analysis}

    '
- id: recovery_campaign
  name: Campaign Trigger Agent
  type: ai_agent
  config:
    tools:
    - trigger_campaign
    - create_user_segment
    system_message: "# Recovery Campaign Agent\n\nDesign and trigger recovery campaigns\
      \ based on root cause analysis.\n\n## Campaign Types:\n1. **User Re-engagement**:\n\
      \   - Cart abandonment recovery\n   - Lapsed user win-back\n   - Feature adoption\
      \ campaigns\n\n2. **Promotional Recovery**:\n   - Targeted discounts\n   - Free\
      \ delivery offers\n   - Bundle promotions\n\n3. **Communication Campaigns**:\n\
      \   - Service status updates\n   - Inventory notifications\n   - Payment issue\
      \ resolution\n\n## Campaign Design:\n- Segment: Affected users\n- Channel: Push/Email/SMS/WhatsApp\n\
      - Message: Personalized content\n- Timing: Optimal delivery\n- Success Metrics:\
      \ Define KPIs\n"
    model_client_id: openai_gpt4_turbo
  depends_on:
  - anomaly_analysis
  - alert_generation
  user_message: 'Design recovery campaign based on:

    Analysis: ${anomaly_analysis}

    Alert: ${alert_generation}

    '
- id: churn_prevention_workflow
  name: Retention & Churn Prevention Agent
  type: ai_agent
  config:
    tools:
    - churn_predictor
    - clevertap_event_api
    - trigger_campaign
    system_message: "# Retention & Churn Prevention Agent\n\nPredict and prevent customer\
      \ churn through behavioral analysis.\n\n## Churn Indicators:\n1. **Behavioral\
      \ Signals**:\n   - Decreasing order frequency\n   - Reduced app engagement\n\
      \   - Cart abandonment increase\n   - Feature usage decline\n\n2. **Transaction\
      \ Patterns**:\n   - Order value reduction\n   - Category switching\n   - Payment\
      \ failures\n   - Support ticket increase\n\n3. **Engagement Metrics**:\n   -\
      \ Email/push open rates\n   - Campaign responsiveness\n   - NPS score decline\n\
      \   - Review sentiment\n\n## Intervention Strategies:\n- Personalized retention\
      \ offers\n- Chronic patient programs\n- VIP customer benefits\n- Proactive support\
      \ outreach\n\n## Special Focus:\nChronic patients requiring regular medication\
      \ refills\n"
    model_client_id: anthropic_claude3
  depends_on:
  - analysis_router
  user_message: 'Analyze churn risk and design prevention:

    ${trigger_data}

    '
  execute_on_routes:
  - churn_analysis_needed
- id: crosssell_optimizer
  name: Cross-sell Optimization Agent
  type: ai_agent
  config:
    tools:
    - clevertap_event_api
    - internal_sales_api
    - create_user_segment
    - trigger_campaign
    system_message: "# Cross-sell Optimization Agent\n\nIdentify and capitalize on\
      \ cross-funnel opportunities.\n\n## Cross-sell Analysis:\n1. **Category Affinity**:\n\
      \   - Medicine \u2192 Wellness products\n   - Diagnostics \u2192 Preventive\
      \ care\n   - Chronic care \u2192 Health devices\n\n2. **Behavioral Patterns**:\n\
      \   - Purchase sequences\n   - Browse behavior\n   - Cart combinations\n   -\
      \ Seasonal patterns\n\n3. **Opportunity Identification**:\n   - Complementary\
      \ products\n   - Upgrade paths\n   - Bundle opportunities\n   - Subscription\
      \ conversions\n\n## Campaign Creation:\n- Personalized recommendations\n- Bundle\
      \ offers\n- Category introductions\n- Educational content\n"
    model_client_id: openai_gpt4_turbo
  depends_on:
  - analysis_router
- id: error_handler
  type: script
  script: 'print("__OUTPUTS__ {\"status\": \"ERROR\", \"message\": \"Unhandled analysis
    type: ${analysis_type}\"}")

    '
  depends_on:
  - analysis_router
- id: workflow_completion
  type: script
  script: "import json\nfrom datetime import datetime\n\nresult = {\n  \"status\"\
    : \"COMPLETED\",\n  \"timestamp\": datetime.now().isoformat(),\n  \"analysis_type\"\
    : \"${analysis_type}\",\n  \"actions_taken\": \"Recovery campaign triggered successfully\"\
    \n}\nprint(f\"__OUTPUTS__ {json.dumps(result)}\")\n"
  depends_on:
  - recovery_campaign
  - churn_prevention_workflow
  - crosssell_optimizer
tools:
- name: clevertap_event_api
  function: "async function clevertap_event_api(api_key, event_name, from_date, to_date,\
    \ limit = 1000) {\n  const axios = require('axios');\n  \n  const headers = {\n\
    \    'X-CleverTap-Account-Id': '${CLEVERTAP_ACCOUNT_ID}',\n    'X-CleverTap-Passcode':\
    \ api_key,\n    'Content-Type': 'application/json'\n  };\n  \n  const payload\
    \ = {\n    event_name: event_name,\n    from: from_date,\n    to: to_date,\n \
    \   common_profile_properties: {\n      limit: limit\n    }\n  };\n  \n  try {\n\
    \    const response = await axios.post(\n      'https://api.clevertap.com/1/events.json',\n\
    \      payload,\n      { headers: headers }\n    );\n    \n    return {\n    \
    \  success: true,\n      data: response.data,\n      event_count: response.data.count,\n\
    \      events: response.data.records\n    };\n  } catch (error) {\n    return\
    \ {\n      success: false,\n      error: error.message\n    };\n  }\n}\n"
  parameters:
    limit: int = 1000
    api_key: str
    to_date: str
    from_date: str
    event_name: str
  description: Fetch user events from CleverTap
- name: create_user_segment
  function: "async function create_user_segment(segment_name, criteria, api_key) {\n\
    \  const axios = require('axios');\n  \n  const headers = {\n    'X-CleverTap-Account-Id':\
    \ '${CLEVERTAP_ACCOUNT_ID}',\n    'X-CleverTap-Passcode': api_key,\n    'Content-Type':\
    \ 'application/json'\n  };\n  \n  const payload = {\n    name: segment_name,\n\
    \    segment_query: criteria,\n    description: `Auto-generated segment for ${segment_name}`\n\
    \  };\n  \n  try {\n    const response = await axios.post(\n      'https://api.clevertap.com/1/segments/create.json',\n\
    \      payload,\n      { headers: headers }\n    );\n    \n    return {\n    \
    \  success: true,\n      segment_id: response.data.id,\n      status: response.data.status\n\
    \    };\n  } catch (error) {\n    return {\n      success: false,\n      error:\
    \ error.message\n    };\n  }\n}\n"
  parameters:
    api_key: str
    criteria: dict
    segment_name: str
  description: Create a user segment in CleverTap
- name: trigger_campaign
  function: "async function trigger_campaign(campaign_id, segment_id, channel, message,\
    \ api_key) {\n  const axios = require('axios');\n  \n  const headers = {\n   \
    \ 'X-CleverTap-Account-Id': '${CLEVERTAP_ACCOUNT_ID}',\n    'X-CleverTap-Passcode':\
    \ api_key,\n    'Content-Type': 'application/json'\n  };\n  \n  const payload\
    \ = {\n    campaign_id: campaign_id,\n    segment_id: segment_id,\n    channel:\
    \ channel,\n    message_content: message,\n    when: \"now\"\n  };\n  \n  try\
    \ {\n    const response = await axios.post(\n      'https://api.clevertap.com/1/campaigns/trigger.json',\n\
    \      payload,\n      { headers: headers }\n    );\n    \n    return {\n    \
    \  success: true,\n      campaign_status: response.data.status,\n      estimated_reach:\
    \ response.data.estimated_reach\n    };\n  } catch (error) {\n    return {\n \
    \     success: false,\n      error: error.message\n    };\n  }\n}\n"
  parameters:
    api_key: str
    channel: str
    message: dict
    segment_id: str
    campaign_id: str
  description: Trigger a campaign in CleverTap
- name: internal_sales_api
  function: "async function internal_sales_api(metric, region, time_range, granularity\
    \ = \"hourly\") {\n  const axios = require('axios');\n  \n  const headers = {\n\
    \    'Authorization': 'Bearer ${INTERNAL_API_KEY}',\n    'Content-Type': 'application/json'\n\
    \  };\n  \n  const params = {\n    metric: metric,\n    region: region || 'all',\n\
    \    time_range: time_range,\n    granularity: granularity\n  };\n  \n  try {\n\
    \    const response = await axios.get(\n      '${INTERNAL_API_BASE_URL}/api/v1/sales',\n\
    \      { headers: headers, params: params }\n    );\n    \n    return {\n    \
    \  success: true,\n      data: response.data.metrics,\n      period: response.data.period,\n\
    \      aggregated_value: response.data.total\n    };\n  } catch (error) {\n  \
    \  return {\n      success: false,\n      error: error.message\n    };\n  }\n\
    }\n"
  parameters:
    metric: str
    region: str | null = null
    time_range: str
    granularity: str = "hourly"
  description: Fetch sales data from internal dashboard
- name: internal_inventory_api
  function: "async function internal_inventory_api(category, sku, region) {\n  //\
    \ Implementation for inventory API\n  // Similar structure to sales API\n  return\
    \ {\n    success: true,\n    stock_levels: {},\n    out_of_stock: [],\n    low_stock:\
    \ []\n  };\n}\n"
  parameters:
    sku: str | null = null
    region: str | null = null
    category: str | null = null
  description: Check inventory levels from internal system
- name: internal_payment_api
  function: "async function internal_payment_api(time_range, payment_method) {\n \
    \ // Implementation for payment API\n  return {\n    success: true,\n    success_rate:\
    \ 0.0,\n    failure_reasons: {},\n    affected_transactions: 0\n  };\n}\n"
  parameters:
    time_range: str
    payment_method: str | null = null
  description: Get payment success rates and issues
- name: anomaly_detector
  function: "def anomaly_detector(metric: str, data: list, sensitivity: float = 2.0):\n\
    \    import numpy as np\n    from scipy import stats\n    \n    # Convert to numpy\
    \ array\n    values = np.array([d['value'] for d in data])\n    timestamps = [d['timestamp']\
    \ for d in data]\n    \n    # Calculate z-scores\n    z_scores = np.abs(stats.zscore(values))\n\
    \    \n    # Identify anomalies\n    anomaly_indices = np.where(z_scores > sensitivity)[0]\n\
    \    \n    anomalies = []\n    for idx in anomaly_indices:\n        anomalies.append({\n\
    \            'timestamp': timestamps[idx],\n            'value': values[idx],\n\
    \            'z_score': z_scores[idx],\n            'deviation': values[idx] -\
    \ np.mean(values)\n        })\n    \n    return {\n        'metric': metric,\n\
    \        'anomalies_detected': len(anomalies),\n        'anomalies': anomalies,\n\
    \        'mean': np.mean(values),\n        'std': np.std(values),\n        'sensitivity_used':\
    \ sensitivity\n    }\n"
  parameters:
    data: list
    metric: str
    sensitivity: float = 2.0
  description: Detect anomalies in metrics using statistical analysis
- name: churn_predictor
  function: "def churn_predictor(user_id: str, behavioral_data: dict):\n    # Simplified\
    \ churn prediction logic\n    # In production, this would use ML models\n    \n\
    \    risk_score = 0\n    risk_factors = []\n    \n    # Check order frequency\
    \ decline\n    if behavioral_data.get('order_frequency_change', 0) < -30:\n  \
    \      risk_score += 30\n        risk_factors.append('Significant order frequency\
    \ decline')\n    \n    # Check engagement metrics\n    if behavioral_data.get('app_opens_change',\
    \ 0) < -50:\n        risk_score += 25\n        risk_factors.append('App engagement\
    \ dropped')\n    \n    # Check cart abandonment\n    if behavioral_data.get('cart_abandonment_rate',\
    \ 0) > 0.7:\n        risk_score += 20\n        risk_factors.append('High cart\
    \ abandonment rate')\n    \n    # Check customer lifetime value trend\n    if\
    \ behavioral_data.get('clv_trend', 0) < -20:\n        risk_score += 25\n     \
    \   risk_factors.append('Declining customer lifetime value')\n    \n    # Determine\
    \ risk level\n    if risk_score >= 70:\n        risk_level = 'HIGH'\n    elif\
    \ risk_score >= 40:\n        risk_level = 'MEDIUM'\n    else:\n        risk_level\
    \ = 'LOW'\n    \n    return {\n        'user_id': user_id,\n        'churn_probability':\
    \ risk_score / 100,\n        'risk_level': risk_level,\n        'risk_factors':\
    \ risk_factors,\n        'recommended_actions': get_retention_recommendations(risk_level,\
    \ risk_factors)\n    }\n\ndef get_retention_recommendations(risk_level, risk_factors):\n\
    \    recommendations = []\n    \n    if risk_level == 'HIGH':\n        recommendations.extend([\n\
    \            'Send personalized retention offer immediately',\n            'Trigger\
    \ win-back campaign',\n            'Assign to customer success team'\n       \
    \ ])\n    elif risk_level == 'MEDIUM':\n        recommendations.extend([\n   \
    \         'Send engagement campaign',\n            'Offer loyalty program benefits',\n\
    \            'Provide personalized recommendations'\n        ])\n    \n    return\
    \ recommendations\n"
  parameters:
    user_id: str
    behavioral_data: dict
  description: Predict customer churn probability
- name: statistical_analyzer
  function: "def statistical_analyzer(data1: list, data2: list = None, test_type:\
    \ str = \"correlation\"):\n    import numpy as np\n    from scipy import stats\n\
    \    \n    if test_type == \"correlation\" and data2:\n        correlation, p_value\
    \ = stats.pearsonr(data1, data2)\n        return {\n            'test_type': 'correlation',\n\
    \            'correlation_coefficient': correlation,\n            'p_value': p_value,\n\
    \            'significant': p_value < 0.05\n        }\n    elif test_type == \"\
    trend\":\n        x = np.arange(len(data1))\n        slope, intercept, r_value,\
    \ p_value, std_err = stats.linregress(x, data1)\n        return {\n          \
    \  'test_type': 'trend',\n            'slope': slope,\n            'r_squared':\
    \ r_value**2,\n            'p_value': p_value,\n            'trend_direction':\
    \ 'increasing' if slope > 0 else 'decreasing'\n        }\n    \n"
  parameters:
    data1: list
    data2: list | null = null
    test_type: str = "correlation"
  description: Perform statistical analysis on data
- name: detect_anomaly
  function: "def detect_anomaly(metrics: dict, threshold: float = 0.2):\n    anomalies\
    \ = []\n    \n    for metric_name, metric_data in metrics.items():\n        current\
    \ = metric_data['current']\n        baseline = metric_data['baseline']\n     \
    \   \n        change = (current - baseline) / baseline if baseline != 0 else 0\n\
    \        \n        if abs(change) > threshold:\n            anomalies.append({\n\
    \                'metric': metric_name,\n                'current_value': current,\n\
    \                'baseline_value': baseline,\n                'change_percentage':\
    \ change * 100,\n                'severity': 'high' if abs(change) > 0.5 else\
    \ 'medium'\n            })\n    \n    return {\n        'anomalies_detected':\
    \ len(anomalies) > 0,\n        'anomaly_count': len(anomalies),\n        'anomalies':\
    \ anomalies\n    }\n"
  parameters:
    metrics: dict
    threshold: float = 0.2
  description: Master anomaly detection orchestrator
- name: fetch_clevertap_data
  function: "async function fetch_clevertap_data(event_type, time_period) {\n  //\
    \ Wrapper function that calls clevertap_event_api\n  // with appropriate date\
    \ parsing\n  return await clevertap_event_api(\n    '${CLEVERTAP_API_KEY}',\n\
    \    event_type,\n    time_period.from,\n    time_period.to,\n    1000\n  );\n\
    }\n"
  parameters:
    event_type: str
    time_period: str
  description: Wrapper for fetching CleverTap data
- name: fetch_internal_data
  function: "async function fetch_internal_data(metric_type, time_period) {\n  //\
    \ Wrapper that aggregates data from multiple internal APIs\n  const salesData\
    \ = await internal_sales_api(metric_type, null, time_period);\n  const inventoryData\
    \ = await internal_inventory_api(null, null, null);\n  const paymentData = await\
    \ internal_payment_api(time_period, null);\n  \n  return {\n    sales: salesData,\n\
    \    inventory: inventoryData,\n    payments: paymentData\n  };\n}\n"
  parameters:
    metric_type: str
    time_period: str
  description: Wrapper for fetching internal dashboard data
- name: analyze_root_cause
  function: "def analyze_root_cause(clevertap_data: dict, internal_data: dict, anomaly_data:\
    \ dict):\n    # Multi-hypothesis testing logic\n    hypotheses = {\n        'user_behavior':\
    \ 0,\n        'technical_issues': 0,\n        'inventory_problems': 0,\n     \
    \   'competitive_factors': 0,\n        'external_events': 0,\n        'payment_issues':\
    \ 0,\n        'marketing_effects': 0,\n        'product_quality': 0,\n       \
    \ 'price_sensitivity': 0,\n        'regional_factors': 0\n    }\n    \n    # Score\
    \ each hypothesis based on data\n    # This is simplified - real implementation\
    \ would be more sophisticated\n    \n    # Find highest scoring hypotheses\n \
    \   sorted_hypotheses = sorted(\n        hypotheses.items(),\n        key=lambda\
    \ x: x[1],\n        reverse=True\n    )\n    \n    return {\n        'primary_cause':\
    \ sorted_hypotheses[0],\n        'secondary_causes': sorted_hypotheses[1:3],\n\
    \        'confidence_scores': hypotheses,\n        'recommended_actions': []\n\
    \    }\n"
  parameters:
    anomaly_data: dict
    internal_data: dict
    clevertap_data: dict
  description: Analyze root cause from multiple data sources
inputs:
- name: trigger_data
  type: string
  required: true
  description: Trigger data containing anomaly detection results or scheduled trigger
    info
- name: analysis_type
  type: string
  default: sales_drop
  required: true
  description: Type of analysis to perform (sales_drop, sales_spike, churn_prevention,
    customer_journey)
description: Autonomous funnel intelligence system that bridges CleverTap and internal
  dashboards for real-time optimization
retry_policy:
  max_attempts: 3
  initial_interval: 5
  maximum_interval: 60
  backoff_coefficient: 2.0
model_clients:
- id: openai_gpt4
  config:
    model: gpt-4
    api_key: ${OPENAI_API_KEY}
    max_tokens: 4000
    temperature: 0.3
  provider: openai
- id: openai_gpt4_turbo
  config:
    model: gpt-4-turbo
    api_key: ${OPENAI_API_KEY}
    max_tokens: 4000
    temperature: 0.5
  provider: openai
- id: anthropic_claude3
  config:
    model: claude-3-opus-20240229
    api_key: ${ANTHROPIC_API_KEY}
    max_tokens: 4000
    temperature: 0.3
  provider: anthropic
execution_mode: temporal
timeout_seconds: 1800