Zvec

Zvec is a free, open-source vector database that runs inside an application process, so indexing and querying do not require a server, daemon, or external infrastructure. It handles dense and sparse vectors, full-text search, and hybrid retrieval with structured filters. Listed index options include Flat, HNSW, IVF, DiskANN, and RaBitQ variants; quantization choices include FP16, INT8, INT4, RaBitQ, and PQ. Write-ahead logging preserves data across process crashes or power failures. Several processes may read a collection at once, but writes are limited to one process at a time. SDKs are listed for Python, Node.js, Go, Rust, Dart, C, and C++. The README lists Linux, macOS, and Windows, with prebuilt SDK binaries also available for Android and iOS. Named integrations include Logseq, MyContext, ReMe, txtai, and CocoIndex. The product family includes Zvec-Grep, a local workspace search CLI and MCP tool for finding code, documentation, and text-based data by exact text, keywords, or meaning. The maker presents Zvec for application developers, from prototyping and local development to embedded, edge, and production systems. It is self-hosted and licensed under Apache-2.0.

Zvec offers broad search, index, and SDK options in a local, open-source package. Its single-writer limit is worth considering for applications with concurrent writes.

Zvec suits application developers who want local vector search for prototyping, development, embedded or edge applications, or production systems. Its SDKs and platform support may fit teams building across the listed languages and operating systems.

Zvec is an open-source vector database that runs inside an application, with no separate service to deploy. It suits developers building local, embedded, or edge search who can route writes through one process. Its broad search and SDK options make it flexible, while exclusive single-process writes narrow its fit for shared workloads.

Overview

Zvec indexes and queries vectors in-process, without a server, daemon, or external infrastructure. That keeps deployment local and removes a service from the application stack, a useful advantage for prototypes and embedded projects. The maker also presents it for production systems, where its write model deserves careful consideration: multiple processes may read a collection at once, but only one process can write. Applications with concurrent writers need a different arrangement or another database.

Write-ahead logging preserves data across process crashes or power failures. Combined with local operation, this makes Zvec suitable when keeping search data alongside an application matters. The Apache-2.0 license also permits use under that open-source license.

Key features

Search and indexing

Zvec supports dense and sparse vectors, full-text search, and hybrid retrieval with structured filters. Its index options include Flat, HNSW, HNSW-RaBitQ, IVF, IVF-RaBitQ, and DiskANN. Quantization choices include FP16, INT8, INT4, RaBitQ, and PQ. This range gives developers room to choose among retrieval and representation approaches; it also means the product is best suited to teams able to make those design choices for their workload.

SDKs and related tools

Official SDKs cover Python, Node.js, Go, Rust, and Dart/Flutter, with C and C++ also linked from the product site. That breadth can accommodate projects using several common application languages. Zvec names integrations with Logseq, MyContext, ReMe, txtai, and CocoIndex. Its product family also includes Zvec-Grep, a local workspace search CLI and MCP tool for people and AI agents. It searches code, documentation, and text-based data by exact text, keywords, or meaning.

Pricing

Zvec costs nothing. The plan is the open-source, in-process vector database under the Apache-2.0 license, and self-hosted deployment keeps the database in the application environment. There is no free trial because the software is offered as a free plan. For developers who want an open-source local database without a separate service, that makes it accessible to try and use; the trade-off is managing deployment yourself and working within the single-process write limit.

Platforms

The README supports Linux, macOS, and Windows, with prebuilt SDK binaries also published for Android and iOS. The product spans application and API use as well as self-hosted deployments. This makes it relevant across desktop, mobile, and server environments, though the single-writer model applies wherever it runs.

Who it's for

Zvec is aimed at application developers. It is a strong fit for prototyping, local development, and embedded or edge applications that benefit from keeping vector search in-process. Production use can also make sense when writes can be serialized through one process. Teams that require multiple processes to write to the same collection should look elsewhere.

Pros and cons

• Pro: Dense, sparse, full-text, and hybrid search with structured filters lets one local database serve several retrieval needs.
• Pro: Multiple index types and quantization choices give developers options for how vectors are indexed and represented.
• Pro: Write-ahead logging protects persistence across crashes and power failures.
• Pro: SDKs span Python, Node.js, Go, Rust, Dart, C, and C++, with mobile binaries for Android and iOS.
• Con: Writes are exclusive to one process, which constrains applications that need concurrent writers.
• Con: Self-hosted, in-process deployment puts operation within the application environment rather than using a separate managed database service.

Alternatives

Compare vector databases if you want to consider more products in this category.

• Chroma is another open-source option, with a free plan and platforms including web and self-hosted deployment; its maker says the same codebase powers its open-source database and Chroma Cloud. Choose it if that combination better fits your deployment needs.
• Actian Zen has no free plan and offers a free trial, with public pricing not stated. Consider it if you want to evaluate a different platform before choosing.
• Pinecone offers a free Starter plan with up to five indexes, 100 namespaces per index, 2 GB storage, 2 million write units per month, and 1 million read units per month. Its API and web platforms suit readers looking for a cloud-oriented option with stated free-tier quotas.
• Qdrant has a free forever tier with a single-node cluster, 0.5 vCPU, 1 GB RAM, and 4 GB disk. Consider it if a free cloud cluster is a better match than an in-process database.
• Vespa has no free plan and offers a trial; its Startup plan prices compute, memory, disk, and GPU memory by the hour. It may suit readers prepared for usage-based charges across those resources.
• Weaviate offers a free plan with one cluster, 100,000 objects, 1 GB memory, 10 GB disk, and one collection. Choose it if those hosted limits fit your needs better than local, in-process deployment.
• Zilliz Cloud offers a free cluster with 5 GB storage, 2.5 million vCUs per month, and up to five collections, plus a pay-as-you-go Dedicated plan. Consider it if a hosted cluster and its defined free allowance suit your project.
• Cloudflare Vectorize has a free Workers plan with monthly queried and stored vector-dimension allowances and account index limits. Choose it if those usage caps and its API and web platforms fit your deployment.

Verdict

Choose Zvec if you are a developer who wants broad search and indexing options in a free, open-source database embedded in an application, and can keep writes within one process. Its local deployment and wide SDK support are compelling for prototypes, edge software, and suitable production systems. Look elsewhere when the same collection needs concurrent writers or you prefer a hosted database service.