Скачать или смотреть Early Walkthrough of a Semantic Search Engine with Qdrant

If you’ve experimented with Milvus or Zilliz Cloud and want to explore how Qdrant works, this video can be a great early reference. I’m in the same boat after working with other vector databases, I wanted to try Qdrant’s approach and see how it handles semantic search with geotagging.

This project was built as part of the Qdrant Hackathon, and the goal was to keep it simple yet meaningful, so others can use it as a starting point for their own experiments.

🔍 What this project achieves:

Converts text into 384-dimensional embeddings using Sentence Transformers

Runs real-time vector similarity search with Qdrant
Adds geotagging support with Haversine distance calculation
Delivers sub-second responses (~250ms) with metadata-rich results

🛠 Tech Stack

Backend: FastAPI + Qdrant + Sentence Transformers
Frontend: Next.js 14 + TypeScript + Tailwind CSS
Daabase: Qdrant (Dockerized)

⚡ Features in the Demo:
Semantic search with cosine similarity
Location-aware filtering
Sample documents with categories, GPS coordinates, tags & timestamps
Vector similarity scoring for ranking

👉 If you’re curious about Qdrant or want a lightweight way to start experimenting with semantic + location-aware search, this walkthrough should help.

🔗 GitHub Repo: https://github.com/Cappybara12/QdrantHack