Zilliz logo
Zilliz logo
user icon

Deepset.ai’s Semantic Search Framework

Haystack is an end-to-end framework that enables users to build powerful and production-ready pipelines for different search use cases. Using state-of-the-art natural language processing (NLP) models, Haystack users can create neural question answering (QA) or semantic document search applications. This helps create unique search experiences and improve the relevance of QA systems.

Haystack integrates Milvus, an open-source vector database built by Zilliz, to power its core vector processing module.

Objective

Build a reliable, scalable semantic search pipeline that can be easily deployed for document retrieval based on meaning instead of keywords.

Challenges

  • High efficiency while storing, indexing, and performing similarity calculation on massive vector datasets.
  • Scalability in a distributed environment.
  • Production-ready robustness and reliability.

Why Verctor Database

  • Database optimized specifically for vectors.
  • Integrates multiple approximate nearest neighbors (ANN) libraries.
  • Supports dynamic data management.
  • Runs as a separate service.

Results

  • Milvus provides great performance on vector storage, indexing, processing, and querying.
  • Milvus filters the entire document store and retrieves the candidate documents.

"Milvus makes a perfect complement to a DPRRetriever or an EmbeddingRetriever. In our experimental runs, we have seen great performance from it."

- Branden Chan, ML Engineer at deepset.ai

Read the full story