DataRobot Vector Databases

Overview

DataRobot Vector Databases provide FAISS-based internal vector stores and managed connections to external vector databases. They are integrated into the DataRobot AI platform to support retrieval-augmented generation (RAG) and other embedding-based AI workloads.

The feature focuses on creating and configuring vector databases, managing versions, and registering and deploying these databases as part of production AI systems.

Features

Vector database types

• Internal FAISS-based vector databases managed directly inside DataRobot.
• External vector database connections, including:
  • Pinecone
  • Elasticsearch
  • Milvus

Creation and configuration

• Create new internal vector databases within the DataRobot platform.
• Configure connection details for supported external vector database providers.
• Associate vector databases with AI projects and deployments.

Data sources and ingestion

• Add internal data sources (data already within DataRobot) to internal vector databases.
• Add and manage external data sources as inputs to vector databases.
• Configure ingestion and update behavior for connected databases (where supported by the underlying system).

Versioning and lifecycle management

• Versioning for internal vector databases, allowing tracking and promotion of specific versions.
• Versioning for connected / external databases, capturing versions of configurations or snapshots as supported.
• Manage multiple versions for experimentation, testing, and rollback.

Registration and deployment

• Register vector databases as reusable assets within the DataRobot platform.
• Deploy vector databases into production environments as part of AI applications.
• Use deployed vector databases as a retrieval layer for models and generative pipelines.

RAG and AI use cases

• Power retrieval-augmented generation (RAG) workflows by providing similarity search over embedded documents.
• Support other embedding-based AI use cases, such as semantic search and recommendation, through the same internal/external vector database abstraction.

Integrations

• Internal: FAISS-based vector store integrated natively with DataRobot.
• External vector databases:
  • Pinecone
  • Elasticsearch
  • Milvus

Pricing

Pricing information is not provided in the available content. Refer to DataRobot’s official pricing or sales materials for details on plans and costs related to vector databases.