A custom GPT assistant that enhanced your personal context

The purpose of this tool is to enhance LLM's understanding on personal customize content and learn from materials by feeding in course resources such as pdf, txt and markdown files.

This AI agent utilizes the semantic reasoning capabilities of LLaMA3, and integrates LangChain along with a vector database to enhance contextual understanding across diverse learning materials.

LINK TO HOMEPAGE

Homepage

LINK TO BACKEND REPO

• Langchain Framework
• Chroma Vector Database

Features supported

1. upload context file
2. dedicate summary for the context
3. chat within the pure context file

why still needs specialised AI agent when LLM like chatgpt and deekseek already exist?

1. Customisation for Specific Use Cases: LLM are general-purpose and may not be optimised for specific workflows
2. Efficiency & Cost Savings: specialised AI agents only focus on relevant inputs
3. Better Context Management: this is the most crucial point, in a general purpose LLM, since it remembers you previous questions, it will use those context as part of its computation

How to use

To install or libraries required, run:

pip install -r requirement.txt

Then install Ollama deepseek model locally

Ollama pull deepseek-r1:1.5b

Or using the GROQ api:

Fill in your api key in .env

Finally unigpt-backend by using:

fastapi run api.py

and open unigpt-frontend run:

npm run dev

Deployment services

• backend: fastapi -> Google Cloud
• frontend: Next.js -> Vercel

Tech stack

• next.js and fast api
• typing-extensions
• langchain
• kaleido
• tiktoken
• python-multipart
• cohere
• openai
• Ollama
• transformers
• huggingface-hub
• chromadb
• sentence-transformers
• PyPDF2