Multi-Modal AI Agent Assistant (GPT-4o & MongoDB) with Gmail Tool Use

## 🧠 AI Agent Processing, MongoDB Vector Store, and Gmail Access

This is the core intelligence hub of your memory assistant:

    The AI Agent (a Langchain Agent node) receives the processed text input (from voice transcriptions, image descriptions, extracted document content, or direct text messages).

    Persistent Memory (MongoDB Recall): The AI Agent is primarily configured to utilize a MongoDB Atlas Vector Store Tool (which you'll connect as a custom tool, enabling RAG).

        When the agent needs to retrieve information, it generates an embedding of the user's query.

        It then queries your MongoDB Atlas cluster (which is configured with an Atlas Vector Search index) to find the most semantically relevant memory chunks.

        These retrieved chunks serve as context for GPT-4o to formulate a grounded and relevant response. This ensures the assistant "remembers" information you've previously provided.

    Language Model: The OpenAI Chat Model (specifically GPT-4o in this setup) acts as the brain for the AI Agent. It synthesizes answers based on its own knowledge, the conversation history (via MongoDB Chat Memory for chat history, separate from vector store), and crucially, the retrieved context from MongoDB Atlas Vector Search.

    Gmail Tools: The AI Agent is also given access to Gmail Tools (Send a message in Gmail and Search for a messages in Gmail). This means the AI can perform actions like:

        Sending emails on your behalf if instructed.

        Searching your past emails to retrieve information, acting as an external data source beyond the dedicated memory store.

    Finally, the Respond node sends the AI Agent's generated answer back to the user via Telegram.