Qdrant
Connect your own Qdrant vector store for vector pipelines you build on the canvas.
The Qdrant integration connects Compass to a Qdrant
vector store you run. It's for building your own vector pipelines on the canvas —
for example, Embeddings feeding an upsert, or
an embedded query feeding a search.
You may not need this
Datasets already give you managed search over your documents — the vector store behind them is built into Compass, and you never connect it yourself. Reach for a Qdrant connection when you want to manage collections and vectors directly.
Connecting
Under Studio → Connections, click Qdrant, then New connection, and
enter the Url of your Qdrant server (for example http://qdrant:6333). If
your server requires a key, choose API Key under Auth type; otherwise
leave it on None.
Compass doesn't call Qdrant when you save, so a wrong address shows up the first time a workflow uses the connection.
Actions
| Action | Kind | What it does |
|---|---|---|
| search | Read | Find the points in a collection nearest a vector, up to limit (default 5), optionally filtered with where. |
| list_collections | Read | List the collections. |
| get_collection | Read | Describe a collection: its vector dimension and point count. |
| upsert | Write | Insert or replace points in a collection. |
| delete | Write | Delete points matching a filter. A filter is required. |
| create_collection | Write | Create a collection with a name, dimension, and distance (cosine, dot, or euclid). Creating one that already exists is a no-op. |
| drop_collection | Write | Delete a collection and everything in it. |
Qdrant is available as an Integration node in workflows only — it can't be given to an agent as a tool. To let an agent search your documents, give it a dataset instead.