qdrant-sparse

Create sparse and hybrid collections in Qdrant using Python.

3|Updated Dec 22, 2025

Search Tags:#retrieval #qdrant #idf #vector-index #upsert #hybrid-search #sparse-vector

Ready to run

npx skills add https://github.com/JoaquinCampo/Skills --skill qdrant-sparse

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: qdrant-sparse
Source: https://github.com/JoaquinCampo/Skills/tree/main/qdrant-sparse
Command: npx skills add https://github.com/JoaquinCampo/Skills --skill qdrant-sparse

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill provides a structured approach to using Qdrant's sparse vector features for lexical retrieval, enabling efficient storage, indexing, and querying of sparse embeddings to improve relevance and latency.

Core Features & Use Cases

Sparse-only collections with IDF-weighted sparse vectors
Hybrid collections combining dense vectors with sparse vectors
Multiple sparse vector fields and batch upserts with payloads
Sparse search with optional payload filtering and score thresholds
Hybrid search using prefetch and fusion strategies (RRF and DBSF)
Performance tuning guidance for production workloads

Quick Start

Create a sparse-only collection named 'docs', upsert a batch of points with sparse vectors, then perform a sparse search using the 'text' field.

Frequently Asked Questions about qdrant-sparse

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I perform sparse vector search in Qdrant using Python?▼

Sparse vector search in Qdrant uses the qdrant-client SDK to create collections with IDF-weighted sparse vectors, upsert points, and query using payload filtering and score thresholds for lexical retrieval.

What is the difference between sparse-only and hybrid search collections in Qdrant?▼

Sparse-only collections store IDF-weighted sparse vectors for lexical retrieval, while hybrid collections combine dense and sparse vectors, enabling search with prefetch and fusion strategies like RRF and DBSF.

How do I upsert sparse vectors with payloads in Qdrant?▼

Upserting sparse vectors in Qdrant involves formatting sparse embeddings and using the qdrant-client SDK to batch upsert points with multiple sparse vector fields and associated payloads.

Does Qdrant sparse vector search support SPLADE and BM42 use cases?▼

Yes, Qdrant sparse vector search supports miniCOIL, SPLADE, and BM42 use cases, enforcing the IDF modifier for sparse vectors to improve retrieval relevance and latency.

How can I tune hybrid search performance for production workloads in Qdrant?▼

Hybrid search performance tuning in Qdrant involves configuring prefetch parameters and selecting appropriate fusion strategies, such as RRF or DBSF, to optimize retrieval latency and relevance.