VOGSY Al architecture
This image is a schematic diagram illustrating the technical architecture of the VOGSY AI agent. It demonstrates how VOGSY utilizes secure, cutting-edge Google Cloud technologies to create a multi-agent AI system that is both trusted and transparent.
The diagram shows the flow of data from the user's request through VOGSY's secure processing environment, the iterative interaction with Google's powerful AI models and VOGSY's own APIs, and the final response back to the user.
Here is a summary of the process flow and a description of each component:
The process flow (how it works)
The request: A user asks a question or gives an instruction in the VOGSY application (frontend). This is sent as a "Long-running, asynchronous request" to the cloud.
Secure entry & orchestration: The request enters VOGSY's secure environment hosted on Google Cloud Run. The FastAPI framework receives it and passes it to the central controller: the Agent Development Kit (ADK) framework.
Context & routing: The ADK immediately logs the request for transparency and checks its internal Memory and Session databases to understand the context. Crucially, it uses multi-agent orchestration & routing to decide which specialized internal agent is best suited to handle the task.
The "think-act-observe" loop: This is the core reasoning engine. The process enters an iterative loop:
Think: The ADK sends the context and the user's goal to the Gemini AI model (hosted in Google Vertex AI) to determine the next best step.
Act: Based on Gemini's instructions, the ADK may need to fetch data or perform an action. It does this by securely calling VOGSY APIs.
Observe: The results from the API calls are fed back into the ADK.
This loop continues until the AI has enough information to answer the user's request completely.
Transparency updates: Because complex tasks take time, the ADK continuously sends "Asynchronous 'thinking' process & step updates" back to the frontend, keeping the user informed of what the AI is doing.
Final response: Once the iterative loop is complete and a solution is found, the final response is delivered to the user in the frontend application.
Component descriptions
Here is a breakdown of the specific components shown in the architecture:
VOGSY application (frontend): The user interface where interaction with the AI agent takes place. It handles long-running requests and displays real-time progress updates.
Google Cloud Run (Python container): This is the secure, isolated execution environment for VOGSY's proprietary code. It ensures that data processing happens within a protected boundary, leveraging Google's robust security infrastructure.
Agent Development Kit (ADK) framework: The "brain" and central orchestrator of the system designed by VOGSY. It manages the conversation flow, short-term memory, and the critical "multi-agent orchestration," breaking down complex queries and routing them to specialized sub-agents for accurate resolution.
MemoryService & SessionService: Temporary databases located securely inside the container that maintain conversation history and user session data, allowing the AI to understand context over multiple interactions.
Request logging: A dedicated component that records activities for auditing, ensuring transparency and accountability in the AI's actions.
Google Vertex AI (Gemini AI model): Google's enterprise AI platform hosting the powerful Gemini large language model. VOGSY uses Gemini as the reasoning engine to understand natural language, plan complicated tasks, and generate human-like responses.
VOGSY APIs: The secure gateways to your actual business data within VOGSY. The AI agent uses these APIs to perform concrete actions (like searching for projects, reading financial data, or updating records) based on the plan it formulated.