An AI-powered travel assistant that allows users to upload travel-related PDFs (itineraries, booking confirmations, guides, etc.) and ask natural-language questions based on the document content. The system performs semantic search over the uploaded document and uses a large language model to generate accurate, contextual answers.

• Upload travel-related PDF documents
• Automatic text extraction and chunking
• Semantic search using vector embeddings
• Context-aware Q&A powered by an LLM
• Simple interactive UI
• Cloud-native vector database

High-level flow:

PDF → Text Extraction → Chunking → Embeddings → Qdrant → Retrieval → LLM → Answer

1. PDF Upload The user uploads a PDF file via the Streamlit frontend.

2. Text Extraction & Chunking The backend extracts text from the PDF and splits it into semantically meaningful chunks.

3. Embedding Generation Each text chunk is converted into a vector embedding using Qdrant.

4. Vector Storage (Qdrant) Embeddings and metadata are stored in Qdrant Cloud for efficient similarity search.

5. Semantic Retrieval When a user asks a question, the query is embedded and used to retrieve the most relevant chunks from Qdrant.

6. LLM Answer Generation The retrieved context is passed to a Hugging Face LLM, which generates a grounded answer based on the document content.

• Frontend: Streamlit
• Backend: FastAPI
• Embeddings & Vector Database: Qdrant Cloud
• LLM: Hugging Face

AiTravelAssistant/
├── frontend/
│   └── app.py              # Streamlit UI
├── backend/
│   ├── main.py             # FastAPI entry point
│   ├── Embeddings.py       # convert text to embeddings
│   ├── generator.py        # generate final answer
│   ├── ingest.py           # read and extract from pdf
│   └── retrievers.py       # retrieve relevant info stored in DB previously             
├── requirements.txt
├── .env
└── README.md

uvicorn src.main:app --reload

http://127.0.0.1:8000/docs#/

streamlit run src/app.py

Endpoint

POST /upload

Request

multipart/form-data
file: travel.pdf

Response

{
  "status": "success",
  "chunks_indexed": 42
}

Endpoint

POST /query

Request Body

{
  "question": "What time is my flight to Paris?"
}

Response

{
  "answer": "Your flight to Paris departs at 7:45 PM on June 12th from SFO."
}