Buying snapshot
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Decision point | What Vald states |
|---|---|
License | Apache-2.0. Vald is installed in a Kubernetes cluster and scales horizontally on memory and CPU; plan capacity for agents, replicas, and backup storage. |
Installation | Helm: |
Requirements | Kubernetes 1.19~ and AVX2 instructions, which Vald Agent NGT requires. |
Client languages | Golang, Java, Nodejs, and Python, with official client libraries (SDKs). |
Data it handles | Object data such as images, audio, video, text, and binary after conversion to vectors. Named uses include recognition, recommendation, detection, grammar checking, and real-time translation. |
Community help | Vald Slack, architecture and usage documentation, Helm-chart configuration guidance, and a demo repository using sample data. |
Our take
Choose Vald when a team already operates Kubernetes and needs dense-vector search that continues during indexing, with replicas rebalanced after an agent goes down. It is a conditional fit: the Kubernetes deployment, replica, and backup responsibilities make it less suitable when operating that infrastructure is outside the intended workload.
Evaluate Vald
Comparisons
Where it stands out
Indexing without a stop-the-world pause. Vald says its distributed index graph continues to work during indexing, unlike graph indexing that requires locking.
Distributed indexes and replicas. It distributes indexes across agents, stores each index in multiple agents for replicas, and automatically rebalances a replica when an agent goes down.
Horizontal scaling. Vald states that it scales horizontally on memory and CPU for demand, alongside automatic vector indexing and index backup for searches across billions of feature vectors.
Backup options. Automatic index backup can use Object Storage or a Persistent Volume for disaster recovery.
Configuration control. Teams can configure vector dimensions and replica counts; Vald also provides configurable ingress and egress filtering for the gRPC interface.
Product screenshot
Trade-offs to weigh
Infrastructure remains your responsibility. Index replicas use multiple agents, and backups use Object Storage or Persistent Volumes, so capacity planning covers memory, CPU, and storage.
Pin production image versions. Default
values.yamlinstalls nightly images, the latest build from the main branch. Released versions usevX.X.X;stableis the latest long-term supported version.
Public signals
About these signals
Verified factual signals from public sources. They indicate observable activity or interest, not total adoption, product quality, or cost.
16 GitHub commits 90d1.7k GitHub stars0 vulnerabilities across 1 package
See all signals from 4 sources
PyPI · vald-client-python@1.8.1
Alternatives to Vald
The reviewed substitutes for Vald among the vector databases, and what would make each one the better answer.
Direct alternatives
Reviewed substitutes: products bought for the same job, where a team picks one.
- ChromaDB
- ChromaDB and Vald are open-source engines for approximate nearest neighbour search over embeddings. ChromaDB favours a quick start with a lightweight Python API, while Vald is a distributed, cloud-native engine for Kubernetes with automatic indexing, index backup and horizontal scaling to very large vector sets.Applies to: Choosing between these two for the vector databases decision.
- pgvector
- pgvector and Vald are open-source tools for approximate nearest neighbour search over embeddings. pgvector adds vector types and indexes to PostgreSQL, so teams search vectors with SQL in a database they already know, while Vald is a dedicated distributed search engine that runs on Kubernetes and scales out with automatic indexing.Applies to: Choosing between these two for the vector databases decision.
- MongoDB Atlas Vector Search
- Vald and MongoDB Atlas Vector Search answer approximate nearest neighbour queries over embeddings for semantic search. Vald is an open-source distributed engine that teams run on Kubernetes, while MongoDB Atlas Vector Search is a managed feature of MongoDB Atlas that stores vectors with the application's documents.Applies to: Choosing between these two for the vector databases decision.
- Pinecone
- Pinecone and Vald serve approximate nearest neighbour search over large sets of embeddings. Pinecone is a commercial managed vector database used through an API, while Vald is an open-source, distributed engine that teams deploy on their own Kubernetes clusters with automatic indexing and horizontal scaling.Applies to: Choosing between these two for the vector databases decision.
- Zilliz
- Zilliz and Vald run approximate nearest neighbour search across very large vector collections. Zilliz is a commercial managed service for Milvus, so the vendor runs the infrastructure, while Vald is an open-source engine that teams deploy and scale on their own Kubernetes clusters.Applies to: Choosing between these two for the vector databases decision.
Other approaches
A different approach to the same problem. Each substitutes only for the workload named beside it.
- Aerospike
- Multi-model database with vector search capabilities — real-time key-value, document, and vector operations at massive scale with predictable low latency.Applies to: Kubernetes-operated approximate-nearest-neighbor vector search workloads
- Marqo
- Marqo optimises search conversion using click-stream, purchase and event data, creating a personalised experience that knows what your customers are looking for - better than they do.Applies to: For teams that build their own product search on a vector store rather than buying a hosted ecommerce search platform.
- Milvus
- Both are open-source distributed vector search engines for approximate nearest neighbour at scale.Applies to: Self-hosted approximate nearest neighbour search at scale on Kubernetes.