smart_rag 0.1.0 → 0.2.1

data/docs/USAGE_EXAMPLES.md

@@ -0,0 +1,1008 @@
+# SmartRAG Usage Examples and Best Practices
+
+This guide provides practical examples and best practices for using SmartRAG in your applications.
+
+## Table of Contents
+
+1. [Quick Start](#quick-start)
+2. [Document Management](#document-management)
+3. [Search Operations](#search-operations)
+4. [Research Topic Management](#research-topic-management)
+5. [Tag Management](#tag-management)
+6. [Advanced Usage Patterns](#advanced-usage-patterns)
+7. [Performance Best Practices](#performance-best-practices)
+8. [Error Handling](#error-handling)
+9. [Common Patterns](#common-patterns)
+
+## Quick Start
+
+### Basic Setup
+
+```ruby
+require 'smart_rag'
+require 'logger'
+
+# Initialize configuration
+config = {
+  database: {
+    adapter: 'postgresql',
+    host: ENV['SMARTRAG_DB_HOST'] || 'localhost',
+    database: ENV['SMARTRAG_DB_NAME'] || 'smart_rag_development',
+    user: ENV['SMARTRAG_DB_USER'] || 'smart_rag_user',
+    password: ENV['SMARTRAG_DB_PASSWORD']
+  },
+  llm: {
+    provider: 'openai',
+    api_key: ENV['OPENAI_API_KEY']
+  }
+}
+
+# Create SmartRAG instance
+smart_rag = SmartRAG::SmartRAG.new(config)
+smart_rag.logger = Logger.new(STDOUT)
+smart_rag.logger.level = Logger::INFO
+
+# Test the connection
+puts "✓ SmartRAG initialized successfully"
+stats = smart_rag.statistics
+puts "✓ Database connected: #{stats[:document_count]} documents"
+```
+
+### First Document Addition
+
+```ruby
+# Add your first document
+result = smart_rag.add_document(
+  'https://example.com/ai-article.pdf',
+  title: 'Introduction to AI',
+  generate_embeddings: true,
+  generate_tags: true
+)
+
+puts "✓ Document added: ID #{result[:document_id]}"
+puts "✓ Sections created: #{result[:section_count]}"
+```
+
+### First Search
+
+```ruby
+# Perform your first search
+results = smart_rag.search(
+  'machine learning algorithms',
+  search_type: 'hybrid',
+  limit: 5,
+  include_content: true
+)
+
+puts "\nSearch Results:"
+results[:results].each_with_index do |result, i|
+  puts "#{i + 1}. #{result[:section_title]} (score: #{result[:combined_score].round(3)})"
+  puts "   #{result[:content][0..150]}..."
+end
+```
+
+## Document Management
+
+### Adding Documents from Various Sources
+
+```ruby
+# From a local file
+smart_rag.add_document(
+  '/path/to/document.pdf',
+  title: 'Research Paper 2024',
+  generate_embeddings: true
+)
+
+# From a URL
+smart_rag.add_document(
+  'https://arxiv.org/abs/2301.00001',
+  generate_embeddings: true,
+  generate_tags: true,
+  tags: ['research', 'AI']
+)
+
+# With custom metadata
+smart_rag.add_document(
+  '/path/to/report.docx',
+  title: 'Q3 Financial Report',
+  generate_embeddings: false,  # Skip embeddings for non-technical docs
+  metadata: {
+    department: 'Finance',
+    year: 2024,
+    confidential: true
+  }
+)
+```
+
+### Batch Document Processing
+
+```ruby
+# Process multiple documents efficiently
+documents = [
+  { path: '/docs/paper1.pdf', tags: ['AI'] },
+  { path: '/docs/paper2.pdf', tags: ['ML'] },
+  { path: '/docs/paper3.pdf', tags: ['NLP'] }
+]
+
+# Sequential processing
+documents.each do |doc|
+  begin
+    result = smart_rag.add_document(
+      doc[:path],
+      generate_embeddings: true,
+      tags: doc[:tags]
+    )
+    puts "✓ Processed: #{doc[:path]}"
+  rescue => e
+    puts "✗ Failed: #{doc[:path]} - #{e.message}"
+  end
+end
+
+# Parallel processing for better performance
+require 'concurrent'
+
+pool = Concurrent::FixedThreadPool.new(5)
+documents.each do |doc|
+  pool.post do
+    begin
+      smart_rag.add_document(
+        doc[:path],
+        generate_embeddings: true,
+        tags: doc[:tags]
+      )
+      puts "✓ Processed: #{doc[:path]}"
+    rescue => e
+      puts "✗ Failed: #{doc[:path]} - #{e.message}"
+    end
+  end
+end
+
+pool.shutdown
+pool.wait_for_termination
+```
+
+### Document Management Operations
+
+```ruby
+# List documents with pagination
+docs_page_1 = smart_rag.list_documents(page: 1, per_page: 20)
+docs_page_2 = smart_rag.list_documents(page: 2, per_page: 20)
+
+# Search for specific documents
+ml_docs = smart_rag.list_documents(search: 'machine learning')
+
+# Get document details
+doc = smart_rag.get_document(123)
+puts "Title: #{doc[:title]}"
+puts "Sections: #{doc[:section_count]}"
+puts "Created: #{doc[:created_at]}"
+
+# Remove a document
+result = smart_rag.remove_document(123)
+puts "Deleted sections: #{result[:deleted_sections]}"
+puts "Deleted embeddings: #{result[:deleted_embeddings]}"
+```
+
+## Search Operations
+
+### Hybrid Search (Default)
+
+Hybrid search combines vector and full-text search for optimal results.
+
+```ruby
+# Basic hybrid search
+results = smart_rag.search(
+  'deep learning applications in healthcare',
+  search_type: 'hybrid',
+  limit: 10,
+  alpha: 0.7  # Weight for vector search (0.0 = pure text, 1.0 = pure vector)
+)
+
+# With content and metadata
+results = smart_rag.search(
+  'natural language processing',
+  search_type: 'hybrid',
+  limit: 5,
+  include_content: true,
+  include_metadata: true
+)
+
+# Search with filters
+results = smart_rag.search(
+  'artificial intelligence',
+  search_type: 'hybrid',
+  limit: 10,
+  filters: {
+    document_ids: [1, 2, 3],  # Search only in specific documents
+    tag_ids: [4, 5]           # Filter by tags
+  }
+)
+```
+
+### Vector Search
+
+Useful for semantic similarity searches.
+
+```ruby
+# Pure vector search
+results = smart_rag.vector_search(
+  'neural network architectures',
+  limit: 5
+)
+
+# Vector search with tag boosting
+results = smart_rag.vector_search(
+  'transformer models',
+  limit: 10,
+  tag_boost_weight: 0.1  # Boost results with matching tags
+)
+```
+
+### Full-Text Search
+
+Best for exact keyword matching and boolean queries.
+
+```ruby
+# Basic full-text search
+results = smart_rag.fulltext_search(
+  'convolutional neural networks',
+  limit: 5
+)
+
+# Advanced boolean query
+results = smart_rag.fulltext_search(
+  'artificial AND (intelligence OR learning) AND NOT robotics',
+  limit: 10
+)
+
+# Phrase search
+results = smart_rag.fulltext_search(
+  '"deep reinforcement learning"',
+  limit: 5
+)
+```
+
+### Multi-language Search
+
+SmartRAG automatically detects and handles multiple languages.
+
+```ruby
+# Chinese search
+results = smart_rag.search('人工智能应用', language: 'zh_cn')
+
+# Japanese search
+results = smart_rag.search('機械学習アルゴリズム', language: 'ja')
+
+# Korean search
+results = smart_rag.search('딥러닝 모델', language: 'ko')
+
+# Mixed language search (auto-detect)
+results = smart_rag.search('AI和机器学习的发展', language: 'auto')
+```
+
+## Research Topic Management
+
+### Creating and Organizing Topics
+
+```ruby
+# Create a research topic
+topic = smart_rag.create_topic(
+  'AI in Healthcare',
+  'Applications of artificial intelligence in medical diagnosis and treatment',
+  tags: ['AI', 'healthcare', 'medicine', 'diagnosis'],
+  document_ids: [1, 2, 3]  # Associate existing documents
+)
+
+# Create nested topic structure
+parent_topic = smart_rag.create_topic(
+  'Machine Learning',
+  'Fundamental ML concepts and algorithms'
+)
+
+child_topic = smart_rag.create_topic(
+  'Deep Learning',
+  'Neural network based learning',
+  tags: ['neural_networks', 'deep_learning']
+  # Could link to parent if hierarchical topics are supported
+)
+```
+
+### Managing Topic Content
+
+```ruby
+# Add documents to a topic
+topic_id = 456
+document_id = 123
+
+result = smart_rag.add_document_to_topic(topic_id, document_id)
+puts "Added #{result[:added_sections]} sections to topic"
+
+# Get topic recommendations
+recommendations = smart_rag.get_topic_recommendations(topic_id, limit: 10)
+
+recommendations[:recommendations].each do |rec|
+  puts "Recommended: #{rec[:section_title]}"
+  puts "  Matching tags: #{rec[:matching_tags]}"
+  puts "  Score: #{rec[:relevance_score]}"
+end
+
+# List all topics
+topics = smart_rag.list_topics(page: 1, per_page: 20)
+topics[:topics].each do |topic|
+  puts "#{topic[:title]} (#{topic[:document_count]} documents)"
+end
+```
+
+## Tag Management
+
+### Automatic Tag Generation
+
+```ruby
+# Generate tags for text
+text = """
+Machine learning is a subset of artificial intelligence that enables systems
+to learn and improve from experience without being explicitly programmed.
+It focuses on developing computer programs that can access data and use it
+to learn for themselves.
+"""
+
+tags = smart_rag.generate_tags(text, topic: 'AI Introduction')
+puts "Categories: #{tags[:categories].join(', ')}"
+puts "Content tags: #{tags[:content_tags].join(', ')}"
+
+# Batch generate tags for document sections
+document = smart_rag.get_document(1)
+sections = document[:sections]
+
+tags_by_section = {}
+sections.each do |section|
+  tags = smart_rag.generate_tags(
+    section[:content],
+    topic: document[:title],
+    max_tags: 5
+  )
+  tags_by_section[section[:id]] = tags
+end
+```
+
+### Manual Tag Management
+
+```ruby
+# Create hierarchical tags
+tag_service = SmartRAG::Services::TagService.new
+
+hierarchy = {
+  "Technology" => {
+    "AI" => ["Machine Learning", "Deep Learning", "Neural Networks"],
+    "Programming" => ["Python", "Ruby", "JavaScript"]
+  }
+}
+
+created_tags = tag_service.create_hierarchy(hierarchy)
+
+# Associate tags with content
+tag = SmartRAG::Models::Tag.find_or_create("machine_learning")
+section = SmartRAG::Models::SourceSection[1]
+
+# Add tag to section
+section.add_tag(tag)
+
+# Find content by tag
+ml_sections = tag.sections
+ml_sections.each do |section|
+  puts "#{section.section_title}: #{section.content[0..100]}..."
+end
+
+# Search for tags
+tags = tag_service.search_tags('learn', limit: 10)
+tags.each { |tag| puts "#{tag.name} (#{tag.section_count} sections)" }
+```
+
+## Advanced Usage Patterns
+
+### Context-Aware Search
+
+```ruby
+class ContextualSearch
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+    @search_history = []
+  end
+
+  def search_with_context(query, user_context = {})
+    # Enhance query with context
+    enhanced_query = enhance_query(query, user_context)
+
+    # Perform search
+    results = @smart_rag.search(
+      enhanced_query,
+      search_type: 'hybrid',
+      limit: 10,
+      filters: build_filters(user_context)
+    )
+
+    # Store in history
+    @search_history << { query: query, context: user_context, results: results }
+
+    results
+  end
+
+  private
+
+  def enhance_query(query, context)
+    # Add context-specific terms
+    case context[:domain]
+    when 'healthcare'
+      "#{query} medical health clinical"
+    when 'finance'
+      "#{query} financial economic banking"
+    else
+      query
+    end
+  end
+
+  def build_filters(context)
+    filters = {}
+    filters[:document_ids] = context[:document_ids] if context[:document_ids]
+    filters[:tag_ids] = context[:preferred_tags] if context[:preferred_tags]
+    filters
+  end
+end
+
+# Usage
+contextual_search = ContextualSearch.new(smart_rag)
+results = contextual_search.search_with_context(
+  'risk assessment',
+  user_context: {
+    domain: 'finance',
+    document_ids: [1, 2, 3],
+    preferred_tags: [4, 5]
+  }
+)
+```
+
+### Search Result Processing Pipeline
+
+```ruby
+class SearchPipeline
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+    @processors = []
+  end
+
+  def add_processor(&block)
+    @processors << block
+    self
+  end
+
+  def search(query, options = {})
+    # Initial search
+    results = @smart_rag.search(query, options)
+
+    # Process through pipeline
+    @processors.each do |processor|
+      results = processor.call(results, query, options)
+    end
+
+    results
+  end
+end
+
+# Create pipeline with processors
+pipeline = SearchPipeline.new(smart_rag)
+
+# Add relevance scoring
+pipeline.add_processor do |results, query, options|
+  results[:results].each do |result|
+    result[:relevance_score] = calculate_relevance(result, query)
+  end
+  results
+end
+
+# Add result filtering
+pipeline.add_processor do |results, query, options|
+  min_score = options[:min_score] || 0.5
+  results[:results].select! { |r| r[:relevance_score] >= min_score }
+  results[:metadata][:filtered_count] = results[:results].length
+  results
+end
+
+# Use pipeline
+results = pipeline.search(
+  'neural networks',
+  min_score: 0.7,
+  limit: 20
+)
+```
+
+### Caching Search Results
+
+```ruby
+require 'redis'
+
+class CachedSmartRAG
+  def initialize(smart_rag, redis_client)
+    @smart_rag = smart_rag
+    @redis = redis_client
+    @cache_ttl = 3600  # 1 hour
+  end
+
+  def search(query, options = {})
+    # Create cache key
+    cache_key = create_cache_key(query, options)
+
+    # Try to get from cache
+    cached = @redis.get(cache_key)
+    if cached
+      puts "Cache hit for: #{query}"
+      return JSON.parse(cached, symbolize_names: true)
+    end
+
+    # Perform search
+    results = @smart_rag.search(query, options)
+
+    # Store in cache
+    @redis.setex(cache_key, @cache_ttl, results.to_json)
+
+    puts "Cache miss for: #{query}"
+    results
+  end
+
+  private
+
+  def create_cache_key(query, options)
+    key_parts = [query, options.sort].flatten.join(':')
+    "search:#{Digest::MD5.hexdigest(key_parts)}"
+  end
+end
+
+# Usage
+redis = Redis.new
+$cached_rag = CachedSmartRAG.new(smart_rag, redis)
+
+# First search - cache miss
+results1 = $cached_rag.search('deep learning', limit: 10)
+
+# Second search - cache hit
+results2 = $cached_rag.search('deep learning', limit: 10)
+```
+
+### Building a Q&A System
+
+```ruby
+class QA_system
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def answer(question, options = {})
+    # Search for relevant information
+    search_results = @smart_rag.search(
+      question,
+      search_type: 'hybrid',
+      limit: options[:context_limit] || 5,
+      include_content: true
+    )
+
+    # Generate answer based on search results (requires LLM integration)
+    answer = generate_answer(question, search_results[:results])
+
+    {
+      question: question,
+      answer: answer,
+      sources: extract_sources(search_results[:results]),
+      confidence: calculate_confidence(search_results[:results])
+    }
+  end
+
+  private
+
+  def generate_answer(question, results)
+    return "I don't have enough information to answer this question." if results.empty?
+
+    # Combine relevant content
+    context = results.map { |r| r[:content] }.join("\n\n---\n\n")
+
+    # Here you would call an LLM API to generate the answer
+    # This is a simplified version
+    "Based on the available information: #{context[0..500]}..."
+  end
+
+  def extract_sources(results)
+    results.map do |result|
+      {
+        section_id: result[:section_id],
+        title: result[:section_title],
+        score: result[:combined_score]
+      }
+    end
+  end
+
+  def calculate_confidence(results)
+    return 0.0 if results.empty?
+
+    # Simple confidence based on top result score
+    [results.first[:combined_score], 1.0].min
+  end
+end
+
+# Usage
+qa = QA_system.new(smart_rag)
+response = qa.answer(
+  'What are the applications of transformers in NLP?',
+  context_limit: 3
+)
+
+puts "Answer: #{response[:answer]}"
+puts "Confidence: #{(response[:confidence] * 100).round(1)}%"
+puts "Sources:"
+response[:sources].each do |source|
+  puts "  - #{source[:title]} (ID: #{source[:section_id]})"
+end
+```
+
+## Performance Best Practices
+
+### 1. Batch Operations
+
+```ruby
+# Instead of individual operations
+bad_practice = documents.map do |doc|
+  smart_rag.add_document(doc[:path], generate_embeddings: true)
+end
+
+# Use batch processing
+good_practice = documents.each_slice(10) do |batch|
+  # Process batch in parallel
+  batch.map do |doc|
+    Concurrent::Promises.future do
+      smart_rag.add_document(doc[:path], generate_embeddings: true)
+    end
+  end.map(&:value)
+end
+```
+
+### 2. Connection Pooling
+
+```ruby
+# Configure database connection pool
+config = {
+  database: {
+    adapter: 'postgresql',
+    host: 'localhost',
+    database: 'smart_rag',
+    user: 'user',
+    password: 'pass',
+    pool: 25,              # Increase pool size
+    timeout: 5000,         # Connection timeout
+    max_connections: 100   # Maximum connections
+  }
+}
+```
+
+Equivalent YAML key:
+
+```yaml
+adapter: postgresql
+```
+
+### 3. Efficient Searching
+
+```ruby
+# Use appropriate search types for queries
+# For exact matches
+text_results = smart_rag.fulltext_search('error code 404', limit: 5)
+
+# For conceptual similarity
+vector_results = smart_rag.vector_search('debugging techniques', limit: 5)
+
+# For general queries
+hybrid_results = smart_rag.search('how to fix bugs', search_type: 'hybrid')
+
+# Adjust alpha based on use case
+# - Technical/keyword-heavy: lower alpha (0.3-0.5)
+# - Conceptual/exploratory: higher alpha (0.7-0.9)
+```
+
+### 4. Caching Strategies
+
+```ruby
+# Cache embeddings for repeated content
+class EmbeddingCache
+  def initialize
+    @cache = {}
+  end
+
+  def get_embedding(text)
+    hash = Digest::MD5.hexdigest(text)
+    @cache[hash] ||= generate_embedding(text)
+  end
+end
+
+# Cache search results
+class SearchCache
+  def initialize(redis, ttl: 3600)
+    @redis = redis
+    @ttl = ttl
+  end
+
+  def fetch(query, options = {}, &block)
+    key = cache_key(query, options)
+
+    if result = @redis.get(key)
+      JSON.parse(result, symbolize_names: true)
+    else
+      result = block.call
+      @redis.setex(key, @ttl, result.to_json)
+      result
+    end
+  end
+end
+```
+
+### 5. Database Optimization
+
+```sql
+-- Create optimized indexes
+CREATE INDEX CONCURRENTLY idx_section_fts_content
+  ON section_fts USING gin(to_tsvector('english', content));
+
+CREATE INDEX CONCURRENTLY idx_embeddings_vector
+  ON embeddings USING ivfflat (vector vector_cosine_ops)
+  WITH (lists = 100);
+
+-- Monitor and optimize slow queries
+EXPLAIN ANALYZE
+SELECT * FROM hybrid_search('machine learning', 10);
+```
+
+## Error Handling
+
+### Comprehensive Error Handling
+
+```ruby
+begin
+  result = smart_rag.add_document(
+    '/path/to/document.pdf',
+    generate_embeddings: true
+  )
+rescue SmartRAG::Errors::ArgumentError => e
+  puts "Invalid arguments: #{e.message}"
+rescue SmartRAG::Errors::DatabaseError => e
+  puts "Database error: #{e.message}"
+  # Attempt to reconnect or use fallback
+rescue SmartRAG::Errors::EmbeddingGenerationError => e
+  puts "Embedding generation failed: #{e.message}"
+  # Retry or skip embeddings
+rescue SmartRAG::Errors::DocumentProcessingError => e
+  puts "Document processing failed: #{e.message}"
+  # Log and continue with next document
+rescue => e
+  puts "Unexpected error: #{e.message}"
+  # Log for investigation
+end
+```
+
+### Retry Logic
+
+```ruby
+require 'retriable'
+
+class RetryableSmartRAG
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def add_document(path, options = {})
+    Retriable.retriable(
+      on: [SmartRAG::Errors::EmbeddingGenerationError],
+      tries: 3,
+      base_interval: 1,
+      multiplier: 2
+    ) do
+      @smart_rag.add_document(path, options)
+    end
+  end
+
+  def search(query, options = {})
+    Retriable.retriable(
+      on: [SmartRAG::Errors::DatabaseError],
+      tries: 3,
+      base_interval: 0.5
+    ) do
+      @smart_rag.search(query, options)
+    end
+  end
+end
+```
+
+## Common Patterns
+
+### Pattern 1: Document Processing Pipeline
+
+```ruby
+class DocumentPipeline
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def process(files, options = {})
+    results = []
+
+    files.each do |file|
+      begin
+        # Step 1: Add document
+        doc_result = @smart_rag.add_document(
+          file,
+          generate_embeddings: false  # Delay embedding generation
+        )
+
+        # Step 2: Generate tags
+        tags = @smart_rag.generate_tags(
+          extract_text(file),
+          topic: options[:topic]
+        )
+
+        # Step 3: Apply tags
+        if doc_result[:document_id]
+          document = SmartRAG::Models::SourceDocument[doc_result[:document_id]]
+          sections = document.sections
+
+          sections.each do |section|
+            tag_objects = tags[:content_tags].map do |tag_name|
+              SmartRAG::Models::Tag.find_or_create(tag_name)
+            end
+
+            section.add_tag(*tag_objects)
+          end
+
+          # Step 4: Generate embeddings (batch)
+          document.sections.each do |section|
+            embedding = @smart_rag.generate_embedding(section.content)
+            store_embedding(section.id, embedding)
+          end
+        end
+
+        results << { success: true, file: file, document_id: doc_result[:document_id] }
+      rescue => e
+        results << { success: false, file: file, error: e.message }
+      end
+    end
+
+    results
+  end
+end
+```
+
+### Pattern 2: Incremental Indexing
+
+```ruby
+class IncrementalIndexer
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def index_new_documents(source_dir, last_check = nil)
+    # Find new or modified documents
+    pattern = File.join(source_dir, '**/*.pdf')
+    documents = Dir.glob(pattern)
+
+    if last_check
+      documents.select! { |doc| File.mtime(doc) > last_check }
+    end
+
+    # Process in batches
+    documents.each_slice(10) do |batch|
+      batch_results = process_batch(batch)
+      log_results(batch_results)
+    end
+  end
+
+  private
+
+  def process_batch(files)
+    files.map do |file|
+      begin
+        result = @smart_rag.add_document(file, generate_embeddings: true)
+        { file: file, success: true, document_id: result[:document_id] }
+      rescue => e
+        { file: file, success: false, error: e.message }
+      end
+    end
+  end
+end
+```
+
+### Pattern 3: Search Analytics
+
+```ruby
+class SearchAnalytics
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def analyze_search_patterns(days = 30)
+    logs = smart_rag.search_logs(limit: 1000)
+
+    analytics = {
+      total_searches: logs.length,
+      avg_execution_time: logs.sum { |l| l[:execution_time_ms] } / logs.length,
+      popular_queries: popular_queries(logs),
+      failed_searches: logs.count { |l| l[:results_count] == 0 },
+      trend_analysis: trend_analysis(logs, days)
+    }
+  end
+
+  def identify_content_gaps(logs)
+    no_result_queries = logs.select { |l| l[:results_count] == 0 }
+
+    # Group similar queries
+    clusters = cluster_queries(no_result_queries)
+
+    # Identify topics needing more content
+    clusters.map do |cluster|
+      {
+        topic: cluster[:topic],
+        query_count: cluster[:queries].length,
+        sample_queries: cluster[:queries].first(3)
+      }
+    end
+  end
+end
+```
+
+### Pattern 4: Multi-tenant Applications
+
+```ruby
+class MultiTenantSmartRAG
+  def initialize(smart_rag)
+    @smart_rag = smart_rag
+  end
+
+  def add_document(tenant_id, path, options = {})
+    # Add tenant isolation
+    options[:metadata] ||= {}
+    options[:metadata][:tenant_id] = tenant_id
+
+    @smart_rag.add_document(path, options)
+  end
+
+  def search(tenant_id, query, options = {})
+    # Filter by tenant
+    options[:filters] ||= {}
+    options[:filters][:metadata] = { tenant_id: tenant_id }
+
+    @smart_rag.search(query, options)
+  end
+
+  def get_statistics(tenant_id)
+    # Get tenant-specific stats
+    @smart_rag.statistics(tenant_id: tenant_id)
+  end
+end
+```
+
+## Summary
+
+This guide has covered:
+
+1. **Quick Start** - Basic setup and first operations
+2. **Document Management** - Adding, organizing, and managing documents
+3. **Search Operations** - Various search types and advanced filtering
+4. **Research Topics** - Organizing content into thematic collections
+5. **Tag Management** - Automatic and manual tagging strategies
+6. **Advanced Patterns** - Production-ready implementations
+7. **Performance** - Best practices for optimal performance
+8. **Error Handling** - Robust error management strategies
+9. **Common Patterns** - Reusable solutions for typical scenarios
+
+For more information, see:
+- [API Documentation](API_DOCUMENTATION.md) - Complete API reference
+- [Performance Guide](PERFORMANCE_GUIDE.md) - Performance optimization details
+- [Migration Guide](MIGRATION_GUIDE.md) - Version upgrade instructions
+
+## Support
+
+- GitHub Issues: https://github.com/your-org/smart_rag/issues
+- Documentation Issues: Report any errors or inconsistencies in examples
+- Community Forum: Share your usage patterns and learn from others