Vector DB
Vector DB
Built-in vector storage for embeddings, search, and RAG — uses your existing API key.
OpenAdapter ships a multi-tenant vector store you can use directly with your gateway API key. The gateway proxies /v1/vectors/* calls through to the upstream vector service after auto-provisioning a tenant for your account on the first call.
What you get
- Per-user tenancy — a vector tenant is created automatically the first time you hit any
/v1/vectors/*endpoint. You never see the upstream key; the gateway substitutes it for you. - Collections scoped to your account
- Upsert / search for points with vectors + payloads
- Embeddings at
/v1/embeddings, OpenAI-compatible, so you can stay end-to-end in OpenAdapter - Single bill — vector calls draw from the same quota as chat. Each vector API call costs 0.05 of one request.
Quick map
| Endpoint | Purpose |
|---|---|
GET /v1/vectors/collections | List your collections |
POST /v1/vectors/collections | Create a collection |
GET /v1/vectors/collections/{name} | Collection metadata |
DELETE /v1/vectors/collections/{name} | Drop a collection |
POST /v1/vectors/collections/{name}/points | Upsert vectors |
POST /v1/vectors/collections/{name}/search | Vector similarity search |
POST /v1/embeddings | Generate embeddings (OpenAI-compatible) |
Topics
- Collections — create, list, delete
- Upsert & embeddings — ingest data
- Search — k-nearest queries
- RAG pattern — embed + search + chat in three calls