AI Analysis

ActiveRabbit's AI engine analyzes your errors and performance issues to provide actionable insights and fix suggestions. This guide explains how to get the most out of AI-powered analysis.

How It Works

When an error is captured, our AI analyzes:

• Stack trace - Identifies the root cause location
• Error message - Extracts key information
• Code context - Reviews surrounding code
• Historical data - Compares with similar past errors
• Rails patterns - Matches against known issues

Fix Suggestions

For common error patterns, ActiveRabbit suggests specific code fixes:

# Example: NoMethodError on nil
# Your code:
user.profile.avatar_url

# AI Analysis:
# "Null reference error - profile may be nil"
#
# Suggested fix:
user&.profile&.avatar_url
# or
user.profile&.avatar_url if user.profile.present?

# With explanation:
# "Use safe navigation operator (&.) to handle cases where 
#  profile is nil. Consider adding a presence check if the 
#  nil case should be handled differently."

Error Categories

The AI automatically categorizes errors:

Root Cause Analysis

# When you have a chain of errors, AI traces to root cause:
#
# Symptom: ActionView::Template::Error in orders/show
#   └─ Caused by: NoMethodError in OrderPresenter
#       └─ Root cause: nil product in line_item
#
# AI identifies: "Data integrity issue - LineItem has 
# product_id referencing deleted Product"
#
# Suggested fixes:
# 1. Add foreign key constraint with ON DELETE behavior
# 2. Add validation: validates :product, presence: true
# 3. Handle gracefully: line_item.product || NullProduct.new

Smart Grouping

AI groups errors by root cause, not just stack trace:

# These errors are grouped together despite different stack traces:
#
# Error 1: NoMethodError: undefined method 'name' for nil
#   app/views/users/show.html.erb:15
#
# Error 2: NoMethodError: undefined method 'email' for nil  
#   app/views/users/profile.html.erb:8
#
# AI groups as: "User object nil in views"
# Common cause: Current user session expired or invalid user_id

Performance Insights

AI analyzes performance data to identify issues:

# AI Performance Analysis Example:
#
# Detected: Response time increased 340% over 2 weeks
# 
# Analysis:
# - P95 latency: 89ms → 392ms
# - Primary contributor: Database queries (+280ms)
# - Specific query: SELECT * FROM orders WHERE user_id = ?
# 
# Root cause: Missing index after migration
# 
# Suggested fix:
# rails g migration AddIndexToOrdersUserId
#
# add_index :orders, :user_id
#
# Estimated improvement: ~85% reduction in query time

Enabling AI Features

ActiveRabbit.configure do |config|
  config.api_key = ENV['ACTIVERABBIT_API_KEY']
  
  # Enable AI analysis (enabled by default)
  config.ai_analysis = true
  
  # Upload source code for better analysis
  config.include_source_context = true
  
  # Number of source lines to include
  config.source_context_lines = 10
end

Code Context

For best AI analysis, enable source code context:

# During deployment, upload source maps
activerabbit releases new \
  --release v1.2.3 \
  --commit $(git rev-parse HEAD)

# Upload source code (for AI analysis)
activerabbit releases upload-source \
  --release v1.2.3 \
  --path ./app

# This enables:
# - Accurate source location in stack traces
# - AI analysis of actual code, not just error messages
# - Suggested fixes with proper context

AI Insights Dashboard

The dashboard shows AI-generated insights including:

• Trend Analysis - "Error rate increased 50% after deploy at 3:45 PM"
• Impact Assessment - "This error affects 12% of checkout attempts"
• Priority Score - Based on frequency, user impact, and fixability
• Similar Issues - "3 other errors share the same root cause"

Custom AI Rules

# Add custom context to improve AI analysis
ActiveRabbit.configure do |config|
  config.before_send do |event|
    # Add feature flags context
    event.extra[:feature_flags] = current_feature_flags
    
    # Add deployment info
    event.extra[:deployment] = {
      version: ENV['APP_VERSION'],
      deployed_at: ENV['DEPLOYED_AT'],
      deployer: ENV['DEPLOYED_BY']
    }
    
    event
  end
end

# AI uses this context to identify:
# - Errors specific to feature flag states
# - Regression introduced in specific deployments
# - Environment-specific issues

Privacy & Security

AI analysis respects your privacy settings:

• Filtered parameters are never sent to AI
• Source code is processed in isolated environments
• No data is used to train external models
• Self-hosted customers can run AI locally