![\"\"](\"https:\/\/www.marktechpost.com\/wp-content\/uploads\/2026\/02\/Screenshot-2026-02-10-at-7.16.14-AM-1.png\")
Why embedded vector search matters for RAG<\/strong>?<\/h3>\nRAG and semantic search pipelines need more than a bare index. They need vectors, scalar fields, full CRUD, and safe persistence. Local knowledge bases change as files, notes, and project states change.<\/p>\n
Index libraries such as Faiss provide approximate nearest neighbor search but do not handle scalar storage, crash recovery, or hybrid queries. You end up building your own storage and consistency layer. Embedded extensions such as DuckDB-VSS add vector search to DuckDB but expose fewer index and quantization options and weaker resource control for edge scenarios. Service based systems such as Milvus or managed vector clouds require network calls and separate deployment, which is often overkill for on-device tools.<\/p>\n
Zvec claims to fit in specifically for these local scenarios. It gives you a vector-native engine with persistence, resource governance, and RAG oriented features, packaged as a lightweight library.<\/p>\n
Core architecture: in-process and vector-native<\/strong><\/h3>\nZvec is implemented as an embedded library. You install it with pip install zvec<\/code> and open collections directly in your Python process. There is no external server or RPC layer. You define schemas, insert documents, and run queries through the Python API.<\/p>\nThe engine is built on Proxima, Alibaba Group\u2019s high performance, production grade, battle tested vector search engine. Zvec wraps Proxima with a simpler API and embedded runtime. The project is released under the Apache 2.0 license.<\/p>\n
Current support covers Python 3.10 to 3.12 on Linux x86_64, Linux ARM64, and macOS ARM64.<\/p>\n
The design goals are explicit:<\/strong><\/p>\n\n- Embedded execution in process<\/li>\n
- Vector native indexing and storage<\/li>\n
- Production ready persistence and crash safety<\/li>\n<\/ul>\n
This makes it suitable for edge devices, desktop applications, and zero-ops deployments.<\/p>\n
Developer workflow: from install to semantic search<\/strong><\/h3>\nThe quickstart documentation shows a short path from install to query.<\/p>\n
\n- Install the package:
pip install zvec<\/code><\/li>\n- Define a
CollectionSchema<\/code> with one or more vector fields and optional scalar fields.<\/li>\n- Call
create_and_open<\/code> to create or open the collection on disk.<\/li>\n- Insert
Doc<\/code> objects that contain an ID, vectors, and scalar attributes.<\/li>\n- Build an index and run a
VectorQuery<\/code> to retrieve nearest neighbors.<\/li>\n<\/ol>\n\n\n\n\n<\/div>\n<\/div>\n<\/div>\n<\/div>\nCopy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\npip install zvec<\/code><\/pre>\n<\/div>\n<\/div>\nExample:<\/strong><\/p>\n\n\n\n\n<\/div>\n<\/div>\n<\/div>\n<\/div>\nCopy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\nimport zvec\n\n# Define collection schema\nschema = zvec.CollectionSchema(\n name=\"example\",\n vectors=zvec.VectorSchema(\"embedding\", zvec.DataType.VECTOR_FP32, 4),\n)\n\n# Create collection\ncollection = zvec.create_and_open(path=\".\/zvec_example\", schema=schema,)\n\n# Insert documents\ncollection.insert([\n zvec.Doc(id=\"doc_1\", vectors={\"embedding\": [0.1, 0.2, 0.3, 0.4]}),\n zvec.Doc(id=\"doc_2\", vectors={\"embedding\": [0.2, 0.3, 0.4, 0.1]}),\n])\n\n# Search by vector similarity\nresults = collection.query(\n zvec.VectorQuery(\"embedding\", vector=[0.4, 0.3, 0.3, 0.1]),\n topk=10\n)\n\n# Results: list of {'id': str, 'score': float, ...}, sorted by relevance \nprint(results)<\/code><\/pre>\n<\/div>\n<\/div>\nResults come back as dictionaries that include IDs and similarity scores. This is enough to build a local semantic search or RAG retrieval layer on top of any embedding model.<\/p>\n
Performance: VectorDBBench and 8,000+ QPS<\/strong><\/h3>\nZvec is optimized for high throughput and low latency on CPUs. It uses multithreading, cache friendly memory layouts, SIMD instructions, and CPU prefetching.<\/p>\n
In VectorDBBench<\/a> on the Cohere 10M dataset, with comparable hardware and matched recall, Zvec reports more than 8,000 QPS. This is more than 2\u00d7 the previous leaderboard #1, ZillizCloud, while also substantially reducing index build time in the same setup.<\/p>\n\n![\"\"](\"https:\/\/www.marktechpost.com\/wp-content\/uploads\/2026\/02\/Screenshot-2026-02-10-at-7.00.44-AM-1.png\")
https:\/\/zvec.org\/en\/blog\/introduction\/<\/figcaption><\/figure>\n<\/div>\nThese metrics show that an embedded library can reach cloud level performance for high volume similarity search, as long as the workload resembles the benchmark conditions.<\/p>\n
RAG capabilities: CRUD, hybrid search, fusion, reranking<\/strong><\/h3>\nThe feature set is tuned for RAG and agentic retrieval.<\/p>\n
Zvec supports:<\/strong><\/p>\n\n- Full CRUD on documents so the local knowledge base can change over time.<\/li>\n
- Schema evolution to adjust index strategies and fields.<\/li>\n
- Multi vector retrieval for queries that combine several embedding channels.<\/li>\n
- A built in reranker that supports weighted fusion and Reciprocal Rank Fusion.<\/li>\n
- Scalar vector hybrid search that pushes scalar filters into the index execution path, with optional inverted indexes for scalar attributes.<\/li>\n<\/ul>\n
This allows you to build on device assistants that mix semantic retrieval, filters such as user, time, or type, and multiple embedding models, all within one embedded engine.<\/p>\n
Key Takeaways<\/strong><\/h3>\n\n- Zvec is an embedded, in-process vector database positioned as the \u2018SQLite of vector database\u2019 for on-device and edge RAG workloads.<\/li>\n
- It is built on Proxima, Alibaba\u2019s high performance, production grade, battle tested vector search engine, and is released under Apache 2.0 with Python support on Linux x86_64, Linux ARM64, and macOS ARM64.<\/li>\n
- Zvec delivers >8,000 QPS on VectorDBBench with the Cohere 10M dataset, achieving more than 2\u00d7 the previous leaderboard #1 (ZillizCloud) while also reducing index build time.<\/li>\n
- The engine provides explicit resource governance via 64 MB streaming writes, optional mmap mode, experimental
memory_limit_mb<\/code>, and configurable concurrency<\/code>, optimize_threads<\/code>, and query_threads<\/code> for CPU control.<\/li>\n- Zvec is RAG ready with full CRUD, schema evolution, multi vector retrieval, built in reranking (weighted fusion and RRF), and scalar vector hybrid search with optional inverted indexes, plus an ecosystem roadmap targeting LangChain, LlamaIndex, DuckDB, PostgreSQL, and real device deployments.<\/li>\n<\/ul>\n
\nCheck out the\u00a0Technical details<\/a> <\/strong>and Repo<\/a>.<\/strong>\u00a0Also,\u00a0feel free to follow us on\u00a0Twitter<\/mark><\/a><\/strong>\u00a0and don\u2019t forget to join our\u00a0100k+ ML SubReddit<\/a><\/strong>\u00a0and Subscribe to\u00a0our Newsletter<\/a><\/strong>. Wait! are you on telegram?\u00a0now you can join us on telegram as well.<\/a><\/strong><\/p>\nThe post Alibaba Open-Sources Zvec: An Embedded Vector Database Bringing SQLite-like Simplicity and High-Performance On-Device RAG to Edge Applications<\/a> appeared first on MarkTechPost<\/a>.<\/p>","protected":false},"excerpt":{"rendered":"Alibaba Tongyi Lab research te…<\/p>\n","protected":false},"author":1,"featured_media":397,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-396","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/396","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=396"}],"version-history":[{"count":0,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/396\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/media\/397"}],"wp:attachment":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=396"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=396"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=396"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Zvec is implemented as an embedded library. You install it with The engine is built on Proxima, Alibaba Group\u2019s high performance, production grade, battle tested vector search engine. Zvec wraps Proxima with a simpler API and embedded runtime. The project is released under the Apache 2.0 license.<\/p>\n Current support covers Python 3.10 to 3.12 on Linux x86_64, Linux ARM64, and macOS ARM64.<\/p>\n The design goals are explicit:<\/strong><\/p>\n This makes it suitable for edge devices, desktop applications, and zero-ops deployments.<\/p>\n The quickstart documentation shows a short path from install to query.<\/p>\n Example:<\/strong><\/p>\n Results come back as dictionaries that include IDs and similarity scores. This is enough to build a local semantic search or RAG retrieval layer on top of any embedding model.<\/p>\n Zvec is optimized for high throughput and low latency on CPUs. It uses multithreading, cache friendly memory layouts, SIMD instructions, and CPU prefetching.<\/p>\n In VectorDBBench<\/a> on the Cohere 10M dataset, with comparable hardware and matched recall, Zvec reports more than 8,000 QPS. This is more than 2\u00d7 the previous leaderboard #1, ZillizCloud, while also substantially reducing index build time in the same setup.<\/p>\n These metrics show that an embedded library can reach cloud level performance for high volume similarity search, as long as the workload resembles the benchmark conditions.<\/p>\n The feature set is tuned for RAG and agentic retrieval.<\/p>\n Zvec supports:<\/strong><\/p>\n This allows you to build on device assistants that mix semantic retrieval, filters such as user, time, or type, and multiple embedding models, all within one embedded engine.<\/p>\n Check out the\u00a0Technical details<\/a> <\/strong>and Repo<\/a>.<\/strong>\u00a0Also,\u00a0feel free to follow us on\u00a0Twitter<\/mark><\/a><\/strong>\u00a0and don\u2019t forget to join our\u00a0100k+ ML SubReddit<\/a><\/strong>\u00a0and Subscribe to\u00a0our Newsletter<\/a><\/strong>. Wait! are you on telegram?\u00a0now you can join us on telegram as well.<\/a><\/strong><\/p>\n The post Alibaba Open-Sources Zvec: An Embedded Vector Database Bringing SQLite-like Simplicity and High-Performance On-Device RAG to Edge Applications<\/a> appeared first on MarkTechPost<\/a>.<\/p>","protected":false},"excerpt":{"rendered":" Alibaba Tongyi Lab research te…<\/p>\n","protected":false},"author":1,"featured_media":397,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-396","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/396","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=396"}],"version-history":[{"count":0,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/posts\/396\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=\/wp\/v2\/media\/397"}],"wp:attachment":[{"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=396"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=396"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/connectword.dpdns.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=396"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}pip install zvec<\/code> and open collections directly in your Python process. There is no external server or RPC layer. You define schemas, insert documents, and run queries through the Python API.<\/p>\n\n
Developer workflow: from install to semantic search<\/strong><\/h3>\n
\n
pip install zvec<\/code><\/li>\nCollectionSchema<\/code> with one or more vector fields and optional scalar fields.<\/li>\ncreate_and_open<\/code> to create or open the collection on disk.<\/li>\nDoc<\/code> objects that contain an ID, vectors, and scalar attributes.<\/li>\nVectorQuery<\/code> to retrieve nearest neighbors.<\/li>\n<\/ol>\npip install zvec<\/code><\/pre>\n<\/div>\n<\/div>\nimport zvec\n\n# Define collection schema\nschema = zvec.CollectionSchema(\n name=\"example\",\n vectors=zvec.VectorSchema(\"embedding\", zvec.DataType.VECTOR_FP32, 4),\n)\n\n# Create collection\ncollection = zvec.create_and_open(path=\".\/zvec_example\", schema=schema,)\n\n# Insert documents\ncollection.insert([\n zvec.Doc(id=\"doc_1\", vectors={\"embedding\": [0.1, 0.2, 0.3, 0.4]}),\n zvec.Doc(id=\"doc_2\", vectors={\"embedding\": [0.2, 0.3, 0.4, 0.1]}),\n])\n\n# Search by vector similarity\nresults = collection.query(\n zvec.VectorQuery(\"embedding\", vector=[0.4, 0.3, 0.3, 0.1]),\n topk=10\n)\n\n# Results: list of {'id': str, 'score': float, ...}, sorted by relevance \nprint(results)<\/code><\/pre>\n<\/div>\n<\/div>\nPerformance: VectorDBBench and 8,000+ QPS<\/strong><\/h3>\n
RAG capabilities: CRUD, hybrid search, fusion, reranking<\/strong><\/h3>\n
\n
Key Takeaways<\/strong><\/h3>\n
\n
memory_limit_mb<\/code>, and configurable concurrency<\/code>, optimize_threads<\/code>, and query_threads<\/code> for CPU control.<\/li>\n
\n