How vsaravind01/smart-wealth works
Main Components
• Frontend: This is the user interface that displays real-time market insights, investment recommendations, and tailored planning. It contains typical assets (HTML, CSS, JavaScript files) and component-based widgets (chat area, mutual fund and stock widgets) as seen in the /frontend directory.
• Backend API: Under the backend/api directory, the API layer handles routes and endpoints (agent, mutual_fund, services, stock) that the frontend calls.
• Core Processing Modules: Several directories under /backend/core, such as:
• finance_agents_network, which implements the reinforcement learning agents (e.g., investor_agent, market_analyzer_agent, etc.)
• market_forecaster, which provides real-time market forecasting services.
• Document Loader: Responsible for data ingestion from different document sources (news, expert documents, NSE, websites, etc.), as observed in the backend/document_loader folder.
• Data & Models: The backend/models directory houses various document representations and utilities for processing data.
• Cluster & Analytics: The backend/clusters directory contains clustering scripts and notebooks (for example, clustering_v1.ipynb) that drive data insights or segmentation.
• Data Persistence & Vector Stores: The project leverages a MongoDB-based store (backend/mongo_store) and has vector store integrations (e.g., azure_cosmos_db, bob_web_db) for efficient retrieval and semantic search.
Architectural Patterns and Key Design Principles
• Layered Architecture: Emphasize how the design clearly segments presentation (frontend), application logic (API and core services), and data/storage concerns.
• Modular Design: The division into smaller folders like document loader, models, and clusters suggests a design that promotes separation of concerns.
• Microservice-ish Separation in the backend: Although contained in one repository, each core function (API, agents, data ingestion, storage) has its own directory and role.
• Use of Advanced Retrieval and ML Components: Note that the design integrates advanced retrieval strategies (RAG Fusion, semantic chunking, semantic caching) and reinforcement learning agents for investment strategies.
• Use clear labels: Each component (API Gateway, Document Loader, Financial Agents, Clustering Module, etc.) should be clearly labeled.
Final Assembly
• From the API gateway, branch into multiple modules:
• Directly to the finance agents network (reinforcement learning agents).
• To market forecaster services.
• To document loaders (which link into models).
• Interconnections from these modules to the MongoDB/vector store.
• Additionally, illustrate the clustering module as supporting analytics.