Vector Database

Vector Database runs a managed Qdrant database on TIR. It stores the vectors your machine learning models produce and finds the ones most similar to a query. Use it for similarity search, recommendations, RAG pipelines and natural language applications. Each database is a cluster of nodes with a dashboard, monitoring, snapshots, a REST API and a gRPC API. TIR runs Qdrant for you.

Where to find it

In TIR, open your project. In the side menu, under Storage, select Vector Database. The page, Manage Vector Database, lists the databases in the project.

What you can do

• Create a vector database with a default collection, a plan, and 3 to 10 nodes.
• Connect to a vector database: find the endpoint and API keys, open the Qdrant dashboard, and copy sample commands.
• Scale up a vector database by adding nodes, then use the new nodes.
• Resize the disk of every node.
• Monitor a vector database with hardware and service graphs.
• Stop or start a vector database.
• Take, restore and delete a snapshot of the whole database, or back up one collection with Qdrant snapshots.
• Delete a vector database when you are finished with it.
• Use the database as the vector store for LangChain or LlamaIndex.

Through the TIR API you can also list, create, show, update and delete databases, resize their disks, and list, take, restore and delete snapshots.

Concepts

• Vector database concepts: points, collections, shards and replicas, and how many vectors a plan holds.
• Vector Database billing: what is charged for compute, disk and snapshots.

Where to start

Start with Get started with Vector Database. For help with a message or a step, Troubleshooting Vector Database has the fixes, and the Vector Database FAQ has short answers. To feed a database from your own documents, see Build a RAG pipeline with the GenAI API.