Community Discussions

Cost or control sometimes drives a move from Pinecone to [Qdrant](https://saaskart.co/software/qdrant). How did the migration go and what did you have to re-eng

HNSW parameters and distance metric choice affect recall and speed in [Qdrant](https://saaskart.co/software/qdrant). What settings worked for your embeddings?

For production [Qdrant](https://saaskart.co/software/qdrant), how are you handling snapshots and restores? What is your recovery plan if a node dies?

[Qdrant](https://saaskart.co/software/qdrant) added sparse vectors for hybrid retrieval. How are people combining dense and sparse signals, and did it improve r

[Qdrant](https://saaskart.co/software/qdrant) integrates with both frameworks. Any gotchas wiring it into a RAG pipeline for metadata filtering or hybrid search

At large scale, sharding and replication matter. How is [Qdrant](https://saaskart.co/software/qdrant) holding up for people with very large collections, and wha

[Qdrant](https://saaskart.co/software/qdrant) is easy to self-host with Docker, but Cloud removes ops. Where did teams draw the line based on scale and staffing

Combining metadata filters with vector search is a Qdrant selling point. How does [Qdrant](https://saaskart.co/software/qdrant) perform when filters are highly

Scalar and binary quantization in [Qdrant](https://saaskart.co/software/qdrant) promise big RAM savings. How much did quantization cut your costs and did recall

The big three open-source vector DBs get compared constantly. What made you pick [Qdrant](https://saaskart.co/software/qdrant) over Weaviate or Milvus for your

For large-scale RAG, [Vespa](https://saaskart.co/software/vespa) offers hybrid retrieval and ranking in one place. How does it compare to stitching a vector DB

Managed [Vespa](https://saaskart.co/software/vespa) removes ops but adds cost. Where did the total cost of ownership land for you versus running your own cluste

Beyond search, [Vespa](https://saaskart.co/software/vespa) powers recommendations. How are you combining user features and item embeddings for personalized rank

The ranking and schema language in [Vespa](https://saaskart.co/software/vespa) is powerful but unfamiliar. How long did it take your team to become productive?

Some vector stores struggle with frequent updates. How well does [Vespa](https://saaskart.co/software/vespa) handle high-rate real-time indexing without hurting

At very large scale, few engines keep up. How is [Vespa](https://saaskart.co/software/vespa) performing for people with hundreds of millions to billions of docu

Self-hosting [Vespa](https://saaskart.co/software/vespa) is powerful but complex. What has the operational burden been, and did teams eventually move to Vespa C

[Vespa](https://saaskart.co/software/vespa) can evaluate ONNX models at query time for ranking. How much latency does in-engine inference add, and is it worth i

Combining BM25 and vector signals is a Vespa strength. How are people tuning hybrid ranking in [Vespa](https://saaskart.co/software/vespa) to beat either signal

[Vespa](https://saaskart.co/software/vespa) does far more than vector search. For teams that only need embeddings similarity, is Vespa overkill versus a focused