Short Introduction
This project is an ongoing development of an agentic RAG (Retrieval-Augmented Generation) system for analyzing stock assets. It processes queries by forwarding them to phidata, utilizes gdrant as a vector database for feature storage, and stores raw OHLC data in MongoDB. A FastAPI-based logging service is already set up (along with its own Docker Compose configuration), and the core of the chat system is also functional. Front-end work is currently in progress.
An agentic RAG system for stock asset analysis.
This project aims to provide an automated research and analysis pipeline for stock assets using retrieval-augmented generation (RAG). By sending user queries to phidata, the system orchestrates data retrieval, analysis, and conversational insights. Over time, additional modules and improvements (including a front-end interface) will be integrated.
- RAG Chat System: Employs a retrieval-augmented approach to answer queries related to stock assets.
- GDRant Vector Database: Stores vector embeddings for better data retrieval and semantic search.
- MongoDB for OHLC Data: Tracks raw stock data (Open, High, Low, Close) for analysis.
- FastAPI Logging Service: Records system logs and events, connected to MongoDB.
- Dockerized Components: The logging service comes with its own Docker Compose setup.
A high-level view (actual folder names and structure may vary):
project-root/
│
├─ backend/
│ ├─ app
│ └─ requirements.txt
│
├─ logging-service/
│ ├─ Dockerfile
│ ├─ docker-compose.yml
│ ├─ app
│ └─ ...
│
├─ database/
│ ├─ app/
│ │ └─ ...
│ ├─ dockerfile
│
│
│
├─ frontend/
│ └─ streamlit
│
├─ README.md
└─ ...
- chat-core: Contains the main logic for the chat system and analysis engine.
- logging-service: Implements the FastAPI-based logging system, storing logs in MongoDB.
- data-services: Manages connections to GDRant (for vector embeddings) and MongoDB (for OHLC data).
- frontend: Streamlit Interface.
- Docker and Docker Compose (for the logging service and other containerized parts).
- Python 3.8+ (for the chat-core and other scripts).
- MongoDB (if not using Docker for the logging service or if you want a separate local instance).
- Qdrant (vector DB, can be run locally or via Docker).
-
Clone the repository:
git clone https://github.com/artaasd95/agentic-asset-recommender cd agentic-asset-recommender -
Change to main directory:
cd agentic-asset-recommender -
Run docker compose:
docker-compose up --build
-
Chat Core:
- Start the chat-related services and workers (details depend on your implementation).
- This component will connect to GDRant and MongoDB for data retrieval and storage.
-
Logging Service:
- Refer to Logging Service below for startup instructions.
-
Query Flow:
- The user sends a query (through the chat or another interface).
- The system retrieves relevant data from GDRant or MongoDB.
- The system processes the data and responds using RAG techniques.
- The logging service is powered by FastAPI and uses MongoDB to store logs.
- Docker Compose is provided in the
logging-service/docker-compose.yml. - To run:
cd logging-service docker-compose up -d - This will start the FastAPI logging service and MongoDB (if configured in the compose file).
- Ongoing Development: Core chat functionality and logging system are in place.
- Front-End: A front-end interface is under construction and not yet ready for production.
- Data Services: GDRant integration for vector embeddings is functional, while MongoDB is used to store OHLC data.
Contributions are welcome! Please open an issue to discuss proposed changes or to report bugs. Feel free to submit pull requests for review.