Releases·Tool Release
Databricks Lakebase Search hits GA with Lakebase_vector and Lakebase_text extensions
Databricks announced general availability of Lakebase Search, adding two extensions: Lakebase_vector for ANN search over pgvector and Lakebase_text for BM25 ranking over tsvector. The features target agent memory and agentic RAG, letting teams turn OLTP Postgres data into vector-searchable stores without a separate vector system.
1 engagement1 sourceWed, Sep 30, 2026, 01:42 PM
AI Overview
- Databricks announces GA of Lakebase_vector (ANN over pgvector) and Lakebase_text (BM25 over tsvector) for agent memory and agentic RAG.
- Lakebase integrates into agent frameworks like LangGraph and LangChain with one line of code, on or off Databricks.
- The extensions vectorize existing OLTP datasets, removing the need for a separate vector system.
Related
Social·Community
Users seek advice on optimizing CPU-only RAG stack latency
A developer detailed a CPU-only RAG stack using Postgres, pgvector, and small ONNX models, reporting high latency (~35s per grounded answer). The community discussion focuses on optimization strategies for CPU-bound retrieval and generation.
A developer detailed a CPU-only RAG stack using Postgres, pgvector, and small ONNX models, reporting high latency (~35s per grounded answer). The community discussion focuses on optimization strategies for CPU-bound retrieval and generation.
21 engagement1 sourceredditSun, Aug 16, 2026, 11:21 AM
News·Company News
Fast Retrieval Was Only Half the Data Architecture We Needed
Zilliz CTO James Luan published three posts and companion blog articles introducing Vector Lakebase, a direction for extending Milvus beyond real-time vector serving into a lake-native architecture. It combines 'One Data' (open lake storage as the shared system of record for content, vectors, metadata, and lineage) with 'One Index' (versioned, lake-level index assets reused across hot, on-demand, and offline compute). Luan argues cheaper vector serving alone doesn't remove coordination costs like re-embedding, backfill sync, and cold-index access.
Zilliz CTO James Luan published three posts and companion blog articles introducing Vector Lakebase, a direction for extending Milvus beyond real-time vector serving into a lake-native architecture. It combines 'One Data' (open lake storage as the shared system of record for content, vectors, metadata, and lineage) with 'One Index' (versioned, lake-level index assets reused across hot, on-demand, and offline compute). Luan argues cheaper vector serving alone doesn't remove coordination costs like re-embedding, backfill sync, and cold-index access.
Updated Wed, Sep 23, 2026, 06:58 PM — Follow-up post argues cheaper vector serving still leaves coordination work (cold-index access, batch/real-time sync) un
0206 engagement3 sourcesredditWed, Sep 16, 2026, 10:07 PM
News·Company News
AWS promotes vector search as retrieval layer for agentic AI
AWS is marketing its vector search solutions as the retrieval layer for agentic AI, emphasizing that agents need fast, relevant access to enterprise data where it already lives. The post highlights how vector search grounds agents in real data across databases, object stores, and unstructured sources.
AWS is marketing its vector search solutions as the retrieval layer for agentic AI, emphasizing that agents need fast, relevant access to enterprise data where it already lives. The post highlights how vector search grounds agents in real data across databases, object stores, and unstructured sources.
0 engagement1 sourcerssThu, Aug 20, 2026, 04:06 PM
Releases·Tool Release
Amazon Bedrock Managed Knowledge Base simplifies enterprise RAG for agents
Amazon Bedrock Managed Knowledge Base is a new service that abstracts away the complexity of building enterprise-grade retrieval-augmented generation (RAG) pipelines for AI agents. It handles connectors, parsers, vector stores, knowledge graphs, and retrieval logic, enabling teams to ground agents on enterprise data without stitching together custom infrastructure.
Amazon Bedrock Managed Knowledge Base is a new service that abstracts away the complexity of building enterprise-grade retrieval-augmented generation (RAG) pipelines for AI agents. It handles connectors, parsers, vector stores, knowledge graphs, and retrieval logic, enabling teams to ground agents on enterprise data without stitching together custom infrastructure.
0 engagement2 sourcesrssThu, Jul 16, 2026, 09:29 PM
News·Company News
Databricks renames Genie Spaces to Genie Agents, adds Supervisor Agent and Knowledge Assistant
Databricks has renamed several AI agent products: Genie Spaces are now Genie Agents, Multi-Agent Supervisor is now Supervisor Agent (GA in February), and Knowledge Assistant went GA in late January. A new front door called Genie One sits above these for business users. The changes, effective June 2026, have caused confusion as many blog posts still use old names.
Databricks has renamed several AI agent products: Genie Spaces are now Genie Agents, Multi-Agent Supervisor is now Supervisor Agent (GA in February), and Knowledge Assistant went GA in late January. A new front door called Genie One sits above these for business users. The changes, effective June 2026, have caused confusion as many blog posts still use old names.
4 engagement1 sourceredditThu, Jul 23, 2026, 03:09 PM