# AI use cases for Qdrant (qdrant/qdrant) Repo: [qdrant/qdrant](https://github.com/qdrant/qdrant) | Use cases: 3 | Page: https://dedupedai.com/use-cases/qdrant-qdrant/ Qdrant is a vector similarity search engine and vector database written in Rust. It stores points, which are vectors with a JSON payload, and offers dense, sparse and multivector search, rich payload filtering, hybrid search with fusion, quantization that cuts RAM use by up to 97%, distributed deployment, a Web UI, REST and gRPC APIs, official clients in several languages and Qdrant Edge for small devices. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a local vector database for your RAG app: https://dedupedai.com/use-cases/qdrant-qdrant/run-a-local-vector-database-for-rag/ - Combine keyword and semantic search in one query: https://dedupedai.com/use-cases/qdrant-qdrant/combine-keyword-and-semantic-search/ - Cut vector search memory with quantization: https://dedupedai.com/use-cases/qdrant-qdrant/cut-vector-search-memory-with-quantization/ ## How to run a local vector database for your RAG app with Qdrant Repo: [qdrant/qdrant](https://github.com/qdrant/qdrant) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Start Qdrant in Docker, connect from Python and store your embeddings with payloads. **Who it is for:** A developer prototyping a RAG or semantic search app who wants a real vector store on a laptop. **The problem:** Keeping embeddings in memory or in files does not scale, and filtering by metadata gets messy fast. **The outcome:** A Qdrant server on localhost with a collection your app can search and filter. ### Steps 1. Run `docker run -p 6333:6333 qdrant/qdrant` to start the server 2. Install the Python client and connect with `QdrantClient(url="http://localhost:6333")` 3. Store your embeddings as points with a JSON payload such as source, date or tags 4. Use the Web UI to explore collections and test queries 5. Read the installation and security guides before moving anything to production ### Prompt to paste into your agent ```text Start qdrant/qdrant locally with Docker, create a collection for my , embed them with , and upsert them with payloads for source and date. Then run a search for filtered by and show me the results. ``` ### Caveat Qdrant stores and searches vectors; you still need an embedding model to create them. Risk, insecure-default: The Docker command starts an instance without authentication, open to all network interfaces, so keep it on a private machine or secure it first. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qdrant-qdrant/run-a-local-vector-database-for-rag/ ## How to combine keyword and semantic search in one query with Qdrant Repo: [qdrant/qdrant](https://github.com/qdrant/qdrant) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Use hybrid search with dense and sparse vectors so results match both meaning and exact terms. **Who it is for:** A team whose semantic search misses exact product codes and names that keyword search would catch. **The problem:** Pure semantic search is fuzzy on exact terms, and pure keyword search misses paraphrases. **The outcome:** One query that blends semantic and keyword matches, merged with a fusion strategy such as RRF. ### Steps 1. Run Qdrant locally or on Qdrant Cloud 2. Store a dense vector for meaning and a sparse vector for full-text search on each point 3. Query both vectors in a single request and merge results with Reciprocal Rank Fusion 4. Tune results with Maximal Marginal Relevance or the Relevance Feedback Query if needed 5. Install the Qdrant agent skills so your coding assistant can help with hybrid search and quantization choices ### Prompt to paste into your agent ```text Using qdrant/qdrant, set up a collection with dense and sparse vectors for , run hybrid queries merged with Reciprocal Rank Fusion, and compare the top 5 results against dense-only search for . ``` ### Caveat The README describes hybrid search but leaves the query syntax to the docs and client libraries. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qdrant-qdrant/combine-keyword-and-semantic-search/ ## How to cut vector search memory with quantization with Qdrant Repo: [qdrant/qdrant](https://github.com/qdrant/qdrant) | Difficulty: Advanced | Time: 2 hours | Basis: From the README Turn on built-in quantization and on-disk storage so a big collection fits on a smaller machine. **Who it is for:** A startup whose vector index is outgrowing the RAM of the server it can afford. **The problem:** Millions of full-precision vectors need a lot of memory, and bigger servers cost more every month. **The outcome:** A collection that uses far less RAM, with a speed and precision trade-off you tune yourself. ### Steps 1. Run Qdrant and load your collection 2. Enable vector quantization; the README says it cuts RAM usage by up to 97% 3. Move data to on-disk storage where memory is tight 4. Measure recall on a sample of known queries before and after 5. Scale out later with sharding and replication if one node is not enough ### Prompt to paste into your agent ```text For my qdrant/qdrant collection , enable quantization and on-disk storage, then measure memory use and top 10 recall on before and after. Recommend settings that keep recall above . ``` ### Caveat Quantization trades precision for memory, so always check recall on your own data. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qdrant-qdrant/cut-vector-search-memory-with-quantization/