If you already have a vector database populated — from your own ingestion pipeline, another product, or years of accumulated data — connect it directly instead of re-uploading everything into Orvoq. Supported: Pinecone, Qdrant, Weaviate, Chroma, Milvus, OpenSearch, and Azure AI Search.
This is a read-only connection. Orvoq never writes to your vector database, chunks anything, or writes embeddings into it — it only queries what's already there.
Setting it up
From Knowledge → Add source → Vector Database, choose your provider and enter its endpoint URL, API key, and index (or collection) name.
Map the fields your database actually uses — which field holds the content, which holds the title, and (for OpenSearch, Milvus, and Azure AI Search) which field holds the vector itself.
Set the embedding provider and model your data was indexed with — a search query has to be embedded the same way your existing vectors were, or the search won't return anything meaningful.
Click Test connection to confirm before saving.
How it's queried
At query time, Orvoq embeds the search query using this connection's own embedding settings — not Orvoq's platform default — and queries your endpoint live. If your credentials expire or the endpoint becomes unreachable, the connection shows a “needs attention” status rather than silently returning nothing.