RAG: Personal Portfolio

Overview:

RAG stands for Retrieval Augumented Generation.

• Retrieve only the useful data from the huge chunk of data.
• Use the Retrieve chunk and modify the data as per the use fo the question asked.
• Very exonomix option for integrating AI in any of our existed product or solutions.

This is a Personal Portfolio project which answers only for the Data ingested during the time of Ingestion. I have used this porject as the of my Personal Portfolio.

RAG Demo Video

Motivation:

The Motivaiton of making this project is that the personal portfolio are very similar type that everything is website based project. So, **RAG: Portfolio * is a use of Agentic AI at very base level and made a Chatbot which only talk about my details and facts I will provide in data points.

✨ You can use this project with some minor tweeks and can check the commit history so you will get the idea the data i have build this on.

Tech Stack:

• Language: Python 3.11
• Library/Framework: FastAPI, LangGraph
• Vector Database: Qdrant
• LLM: openai-gpt-oss-20b (OpenAI)
• Embedding Model: gemini-embeddings-v2 (Google), all-mpnet-base-v2 (Sentence Transformers)
• Observability: Pydantic Logfire
• Guardrails: NeMo Guardrails (Input, Output, Topic, JailBreak)

File/Folder Structure:

rag
|
|- .venv/               # virtual environment
|- app/                 # main application
|- DATA/                # raw data in any format or file type
|- processed_data/      # Local Copy of processed and chunked data stored in vector DB
|- .env                 # API keys and tokens
|- .gitignore           
|- README.md
|- requirements.txt     # libraries and packages used during the project

Data Flow Diagram (DFDs)

Fig: 1.1- User Flow

Fig: 1.2- Ingestion Pipeline

Fig 1.3- Agentic Decision Flow

Fig: 1.4- Guardrails

Local Setup:

1. Clone the Repository:

git clone https://github.com/hariom2809/rag.git

2. Get your API keys and tokens from the Gemini, Groq and Qdrant Cloud

# Grow APIs 
GROQ_API_KEY=your_groq_api_key
GROQ_FALLBACK_API_KEY=your_groq_fallback_api_key
GROQ_MODEL=openai/gpt-oss-20b

# Qdrant APIs
QDRANT_API_KEY=your_qdrant_api_key
QDRANT_CLUSTER_ENDPOINT=your_cluster_endpoint

# Gemini APIs
GEMINI_API_KEY=your_api_key

Look inot the .env.example there you will have all environment variable

3. Make Virtual Environment
   • For this project I am using the UV Python package manager you can do the same with cPython

uv venv --python 3.11

For this we are using the runtime of 3.11 . Can Verify from the runtime

4. Activate the Virtual Environment

   • Windows

   source .venv/Scripts/activate

   • Linux/MacOS

   source .venv/bin/activate

5. INstall the required dependencies

uv pip install -r requirements.txt

6. Now collect all of your documents and Data to be ingested in db and place it at the DATA folder under a subfolder by any name. Suppose we gave ti name raw_data -> DATA/raw_data/....

7. Go the Config file in the app folder app/config.py and change the name of your Qdrant collecion as of your choice

8. Run the Ingestion Process

python -m app.ingestion.processor DATA/{your_folder_name} {your_destination_folder_name}

The Data for the Local instance will got save at the processed_data directory

9. You are all set now run the server

uvicorn app.main:app --reload --port 8000

Now you can test your application on Engpoint

http://localhosta;8000/docs

Author

Hariom Gupta Email: hariomgupta2809@gmail.com Linkedin: linkedin/hariom2809