Inspiration
OmniTech provided us with a pointed, real-world problem; how can a group of programmers help streamline a company's workflow? They gave the resources and let us take on this unique opportunity.
What it does
Our IT operations dashboard works by integrating data from Accelo and N-central through a Python FastAPI backend. It will organize support tickets, tasks, device monitoring, AI assistance, system alerts and more. Additionally, OmniCommand contains an AI assistant with access to company documents that can help employees find fast and reliable answers to their questions.
How we built it
Frontend: Our frontend is built using node.js providing an interactive dashboard for IT operations. It integrates ticket and task management, employee assignments, process planning and an AI assistant that communicates with Fast API and Rest API endpoints.
Backend: Our backend is built in python using FastAPI to handle REST API requests, manage application data, and facilitate communication between the frontend and database. Integrated PostgreSQL for persistent data storage and implemented API endpoints to support IT operations, task management, employee assignments, and workflow automation.
RAG System: Our RAG pipeline processes internal documents pulled from the Omni Brain mock API, processing them into a postgresql database using pgvector. Semantic similarity searching was implemented after chunking and embedding each document. From there, an AI model is used to retrieve the relevant data and generate responses to the user.
Challenges we ran into
Mock API Challenges: Due to the nature of Omni Brain API we found it to be incomplete. While we could pull jobs from the API, it did not have the capability of pushing any new jobs. To simulate this we are storing newly created jobs in postgresql.
AI limitations: No AI model is going to give perfect results, and our implementation is no exception. Refining a RAG system is essential for accurate AI response, especially when dealing with queries that have definite responses. To combat this, we focused on developing a thorough chunking and embedding process that allows the AI Assistant to parse only topically-related information.
Accomplishments that we're proud of
sleek design
postgresql integration
RAG implementation
What we learned
team collaboration
RAG systems
database usage and containerization
api calls
What's next for OmniSupport
Our goal is to simplify IT management, improve visibility across multiple systems, and help technicians prioritize problems before they escalate, with Omnicommand we have achieved this and hopefully someday Omnicommand may be used by a real IT team
Built With
- docker
- fastapi
- markdown-it
- openai
- pgvector
- postgresql
- python
- rag
- tiktoken
