Tag: tag

Oracle AI Database 26ai in-database ONNX embedding model compared to pgvector external API embedding architecture — data sovereignty and latency comparison

Oracle AI Database 26ai Vector Search vs PostgreSQL pgvector — A DBA’s Perspective

Posted on by Sanjeeva Kumar

We have run vector search workloads on both Oracle AI Database 26ai and PostgreSQL with pgvector and we are here to show you the architectural breakdown of Oracle AI Database 26ai vector search vs pgvector. The technical capabilities of both are real — pgvector works, and PostgreSQL shops have built production vector search on it. Read more about Oracle AI Database 26ai Vector Search vs PostgreSQL pgvector — A DBA’s Perspective[…]

Oracle AI Database 26ai hybrid search combining VECTOR_DISTANCE semantic similarity with SQL WHERE filters in a single execution plan — no separate round trip required

Hybrid Search in Oracle AI Database 26ai — Combining SQL and Semantic Queries

Posted on by Sanjeeva Kumar

We had a support tool prototype running vector search against an Oracle error knowledge base. Pure semantic search — query comes in, top five similar errors come back. Worked well in testing. In production, the first complaint arrived within a week: “It keeps returning memory errors when I search for connection problems — they are Read more about Hybrid Search in Oracle AI Database 26ai — Combining SQL and Semantic Queries[…]

Oracle AI Database 26ai vector index comparison showing HNSW graph layers and IVF partition clusters for semantic similarity search

Vector Indexes in Oracle AI Database 26ai — How HNSW and IVF Work Under the Hood

Posted on by Sanjeeva Kumar

In Post -2 of this series, We had a vector search query running clean in development — ten milliseconds, HNSW index in the plan, results looking right. The moment the same query hit a table with eight million rows in staging, response time jumped to forty seconds. The execution plan had quietly switched to TABLE Read more about Vector Indexes in Oracle AI Database 26ai — How HNSW and IVF Work Under the Hood[…]

Oracle Vector Search: Your First Hands-On Walkthrough

Posted on by Sanjeeva Kumar

In Post 1 we established the architecture — Oracle 23ai puts vector search inside the database engine, not in a separate system. We titled that post using “23ai” deliberately — because that is the name most production DBAs are still searching for, still referencing in upgrade discussions. Since then, Oracle announced Oracle AI Database 26ai Read more about Oracle Vector Search: Your First Hands-On Walkthrough[…]