Add MongoDB Text-to-MQL Agent Tutorial to Gen-AI Showcase

[MMBazel]

Add MongoDB Text-to-MQL Agent Tutorial to Gen-AI Showcase

This PR adds a comprehensive tutorial showcasing how to build natural language agents using MongoDB Atlas and LangChain's Text-to-MQL tools. The guide walks developers through two agent architectures:

• ReAct-style agent – ideal for fast prototyping, demos, and dynamic tool usage.
• LangGraph-style agent – designed for production with deterministic, auditable flows.

Key Features

• Use the langchain_mongodb.agent_toolkit to generate, validate, and execute secure MongoDB aggregation pipelines from plain English.
• Simplified prompt design using MONGODB_AGENT_SYSTEM_PROMPT.
• Detailed comparison of ReAct vs. graph-style agents (speed, accuracy, tradeoffs).
• Patterns for extending agents with memory, hybrid search (Vector + Full-Text + MQL), and observability.

Audience

This tutorial is intended for developers and teams building:

• GenAI agents over structured data
• Internal tools and analytics bots
• Production-grade agentic systems with audit/control requirements

[ajosh0504]

Nice work! I like the comparison between an easy (React) and a more customized (LangGraph) agent. Few comments:

• I found the wording "graph-based agent" a bit confusing. I took it to mean it was using graph-based retrieval or GraphRAG. Might want to specifically say "custom LangGraph agent" or similar.
• The long outputs at every step make the notebook a bit hard to read. Consider clearing the outputs of some cells
• In the LangGraph agent, curious why there is a conditional edge after generate_query? Shouldn't check_query, run_query, etc., always be run if the agent decides to generate a query?
• Instead of saving the output to a local file, you could persist it in MongoDB? Could also use the MongoDB LangGraph checkpointer to show how to use these persisted outputs as memory for the agent.