Feature: Add Azure CosmosDB NoSQL tools for memory management, vector search, and semantic caching

[aayush3011]

🚀 Summary

This PR introduces a comprehensive suite of Azure CosmosDB NoSQL tools for CrewAI, enabling advanced memory management, vector search capabilities, and intelligent semantic caching for multi-agent applications.

✨ New Features

🔍 Vector Search Tool

• Semantic similarity search using vector embeddings
• Hybrid search combining vector and full-text search with RRF (Reciprocal Rank Fusion)
• Configurable similarity thresholds and result limits
• Cross-partition querying for comprehensive search

⚡ Semantic Cache Tool

• Intelligent LLM response caching using semantic similarity
• Cost optimization by reducing redundant API calls
• TTL support for automatic cache expiration
• OpenAI/Azure OpenAI integration for embedding generation

🧠 Memory Store Tool

• CRUD operations for agent memory management (conversation history, short/long-term memory)
• Hierarchical partition key support for multi-agent isolation
• Flexible memory types with generic content storage
• Batch operations for efficient data management

🧪 Testing

• 55+ comprehensive test cases covering all functionality
• Extensive mocking for Azure CosmosDB and OpenAI APIs
• Edge case coverage including error handling and validation
• 100% test coverage for critical paths

📚 Documentation

• Detailed README files for each tool with usage examples
• Azure CosmosDB documentation links for relevant features
• End-to-end CrewAI integration examples
• Configuration guides and best practices

🔗 Azure CosmosDB Features Utilized

• Vector Search - Native vector similarity search capabilities
• Full-Text Search - Enhanced text search with ranking
• Hierarchical Partition Keys - Multi-level data isolation
• TTL (Time-to-Live) - Automatic data expiration
• Indexing Policies - Optimized query performance
• Batch Operations - Efficient bulk data operations

💡 Use Cases

Vector Search

# Find similar documents
search_tool.run(
    operation="vector_search",
    query_text="machine learning algorithms",
    top_k=5,
    similarity_threshold=0.8
)

Semantic Cache

# Cache LLM responses intelligently
cache_tool.run(
    operation="search",
    prompt="What is machine learning?"
)

Memory Store

# Store agent conversation history
memory_tool.run(
    operation="store_memory",
    agent_id="assistant_agent",
    memory_type="conversation",
    content={"user": "Hello", "assistant": "Hi there!"}
)

🔧 Configuration

All tools support flexible configuration including:

• Authentication - Azure Key or Token Credential
• Database/Container settings - Custom names and properties
• Performance tuning - Throughput, indexing, and caching policies
• OpenAI integration - Both OpenAI and Azure OpenAI support

🚦 Breaking Changes

None - This is a new feature addition.

📋 Dependencies Added

• azure-cosmos - Azure CosmosDB SDK
• azure-core - Azure authentication
• openai - OpenAI/Azure OpenAI integration

[aayush3011]

@lucasgomide @thiagomoretto @lorenzejay Can I please get a review on this.

[aayush3011]

@lucasgomide @thiagomoretto @lorenzejay Can I please get a review on this.

[aayush3011]

@lucasgomide @thiagomoretto @lorenzejay Can I please get a review on this.