ApertureDB 0.20.0 compatibility
Python SDK 1.0.1, 0.5.4, and 0.5.3 support this release. SDK 1.0.1 supports both TCP/TLS and HTTP and can stay installed through a server upgrade to 1.0.0. For 0.5.x applications, 0.5.4 fixes helper and schema issues present in 0.5.3.
ApertureDB 0.20.0 accepts direct database upgrades from 0.19.14, 0.19.15 and 0.19.16. Complete startup on 0.20.0 before upgrading to 1.0.0. Release notes
The external LlamaIndex ApertureDBVectorStore adapter uses this server’s descriptor-set API.
Descriptor sets
On server 0.20.0, this request creates a 128-dimensional descriptor set with exact Flat search and the L2 distance metric:
[{"AddDescriptorSet": {
"name": "Faces",
"dimensions": 128,
"engines": {"flat": {}},
"metric": "L2"
}}]
On server 1.0.0, the corresponding vector definition is:
[{"CreateIndex": {
"target": "descriptor",
"class": "Faces",
"property": "_blob",
"kind": "flat",
"params": {
"dimensions": 128,
"metric": "L2"
}
}}]
Add vectors with AddDescriptor.set: "Faces" on server 0.20.0 and
AddDescriptor.class: "Faces" on server 1.0.0. Native searches select the class
with with_class and an explicit compatible engine.
Supported descriptor sets become classes and vector indexes during a 0.20.0 database upgrade. Customer connections to descriptor-set nodes prevent that conversion; see the release upgrade notes. Keep a pre-upgrade backup and migrate application queries before upgrading.
The complete historical parameter contract is retained in the 0.20.0 source reference.