This project aims to provide a robust and scalable solution for handling complex queries in a distributed environment.
- Name: Manatees
- Member: Luis E. Fernandez de la Vara
- NetID: luisef2
The project focuses on implementing an unstructured search application that utilizes query orchestration. It involves a web client for query input, a query router, and backend services powered by pre-trained models to process and respond to queries.
-
Client: A simple, single-search client that routes queries to the orchestrator. It will be containerized using Docker.
-
Orchestrator: An event-driven NodeJS service endpoint that interfaces with a Python classification service to categorize queries as either imperative or descriptive, routing them accordingly.
-
Imperative Service: Processes imperative queries and returns actionable instructions.
-
Descriptive Service: Handles descriptive queries and provides detailed explanations.
Each component is containerized using Docker, allowing for flexible deployment and management.
This project is significant due to its:
- Lightweight Architecture: A declarative and scalable service architecture using Docker.
- Advanced Query Classification: Efficient routing based on query type.
- Extensibility: Potential applications in various domains like intrusion detection.
- Frontend: JavaScript (NodeJS)
- Text Processing: Python, with potential use of C/C++
- Containerization: Docker
- ML Classifiers: Open-source NLP classifiers
The application aims for high precision with moderate recall capabilities.
The system will be evaluated through manual labeling and classification of queries by volunteers.
The project is estimated to require a minimum of 20 hours, distributed as follows:
- Networking flow and client setup: 5-10 hours
- ML classifier integration: 5-10 hours
- Docker-based deployment: 5-10 hours