Qdrant¶

Type: qdrant

Connects to a Qdrant vector database via its REST API. Daimon calls a configurable embeddings endpoint to generate vectors before upserting or querying.

Best for: production semantic search with a dedicated vector database.

Docker quickstart¶

docker run -d -p 6333:6333 qdrant/qdrant

Configuration¶

# Declare the embedder first
- name: embedder
  type: embedding/openai
  metadata:
    base_url: http://localhost:11434/v1   # Ollama
    model: nomic-embed-text
    dimensions: "768"

- name: qdrant-docs
  type: qdrant
  metadata:
    base_url: http://localhost:6333
    collection: daimon
    embedder: embedder            # reference the component above
    create_if_missing: "true"
    # api_key: ...                # Qdrant Cloud only

Metadata keys¶

Key	Default	Description
`base_url`	`http://localhost:6333`	Qdrant server URL
`collection`	`daimon`	Collection name
`embedder`	—	Name of a declared `embedding/openai` component
`create_if_missing`	`"false"`	Auto-create collection on startup
`dimensions`	`"1536"`	Vector dimensions (must match the embedder)
`api_key`	—	Qdrant Cloud API key

Notes¶

If embedder is not set, the store falls back to a deterministic hash vector — useful for configuration smoke tests but not semantically meaningful.
Scores are Qdrant cosine similarity values (0–1, higher is more similar) when the collection is configured with cosine distance.
The collection's distance metric is set at creation time. If the collection already exists, create_if_missing has no effect on its metric.