The roadmap provides a high level overview of key areas that will likely span multiple releases.
0.1 was released in early September.
End-to-End one-click deployment.
RESTful API
Support L2 and inner-product searching metric.
Replace the static batch indexing with real time indexing.
Add the fine-grained sort after PQ coarse sort.
Add the numeric field and bitmap filters in the process of searching.
Real-time modified IVFPQ model based on Faiss.
Only memory supported now.
Support docker.
Vearch 0.2 has been released on October 31th 2019.
Numeric index filtering optimization.
Memory and disk supported now.
Video surveillance security scene algorithm plug-in.
Vearch 0.3 released on 20th January 2020.
Support vector search with GPU. Support single online request for GPU, not just batch requests
Python SDK which can be friendly used in local computer or edge device.
Plugin service can be installed in Docker.
Stability testing.
Vearch 3.1.0 released in May 2020.
Support real time HNSW index
Support binary index
Support IVFFLAT index
Vearch 3.2.0 will be released in August 2020.
Clear gamma engine api
Gamma engine supports customerized retrieval models
Support the Grpc service for router
Refactor for the router and partition server
Router grpc client(java, c++, go and python)
More fast and robust retrieval models(q-adc, IVFx+HNSWy+OPQz, NSG, …)
More complex powerful query gramma like SQL or ES(enhance VQL)
Ivf index: change the docid from long to int, try to compress the docids with the algorithm of newpfordelta
Develop a plugin which can be used to index vectors in Spark or Flink clusters with Gamma engine
Vector storage based on the clusters
Support the multiple retrieval models and versions
Support the auto-expired documents
Explore and integrate the best and practical vector compression technology