Vector Store (Qdrant)
A Qdrant connection is where the AI assistant's knowledge base keeps its embeddings. It holds no tags, appears in no binding, and nothing in the plant depends on it — it exists so the assistant can ground an answer in your own documents and cite them.
Without one, everything else in QUBIQ works exactly as it does now; the assistant simply has no documents to retrieve from.
The vector store belongs to the Knowledge Base module. → Licensing & modules
Configuration
| Field | Notes |
|---|---|
| Connection Name | e.g. Knowledge Vector Store. |
| Host | localhost for a Qdrant you run, or the endpoint from Qdrant Cloud. |
| REST Port | Optional. Blank means 443 when TLS is on, 6333 for a plain local Qdrant. |
| API Key | Optional — leave blank for a local Qdrant with no authentication. |
| Use TLS (HTTPS) | On for a hosted endpoint. |
Only the REST endpoint is configured here; the knowledge service builds the rest of the URL from these fields.
What lands in it
Documents are uploaded per project, from Settings → AI Assistant → Knowledge base: PDF, DOCX, XLSX, TXT, MD or CSV. Each is extracted, split into chunks and embedded, and the assistant retrieves the relevant passages when it needs them — with citations, so an answer can be checked against the source rather than taken on trust.
Embeddings are scoped per project. Two projects on one gateway do not read each other's documents.
Operational notes
- Treat it as a cache, not a system of record. The documents themselves are the record; the vectors can be rebuilt by re-ingesting. Losing the store costs re-ingestion time, not data.
- It has no bearing on the control path. A vector store that is down affects the assistant's ability to cite documents and nothing else.
- Keep it on the plant side of your network boundary if the documents are sensitive. The connection is a URL and a key — pointing it at a cloud endpoint sends your document text there to be embedded.