Whatever message this page gives is out now! Go check it out!
Qdrant vector store configuration options
Last update:
May 18, 2026
This document describes settings when Qdrant is selected as the vector store provider (where embedding vectors and payloads are stored and searched). Qdrant does not produce embeddings: configure the embedding model separately (for example under AI Services > Embedding models, or via Source / Provider in the Embedding model section).
Cross-check field names and defaults against your product build and Qdrant documentation.
Global behavior notes
- Dimension must match embeddings: The dimension must exactly match the output size of the embedding model used for this workflow. If they differ, collection creation, upserts, or search will fail or behave incorrectly.
- Secrets: The API key (when used) grants access to your Qdrant Cloud cluster. Store configurations securely and restrict who can view or export them.
- Local vs cloud: Qdrant Cloud usually requires an API key. A local Qdrant instance without authentication may leave the API key empty.
- Timeouts: Connection-related values are in milliseconds unless your UI states otherwise.
Basic configuration
Field | Description |
|---|---|
Provider | Qdrant. The selected vector store provider for this configuration. |
URL | Required. HTTP or HTTPS endpoint for your Qdrant instance (for example http://localhost:6333 for a local default, or the URL shown in Qdrant Cloud). Must be reachable from the application server. |
API key | Optional for local deployments without auth. Required for typical Qdrant Cloud clusters. Create and rotate keys in the Qdrant Cloud console; treat any saved key as a secret. |
Collection name | Optional with a product default (often default). Logical collection for vectors and payload. If the collection does not exist, the product may create it using the dimension and distance metric you specify—confirm auto-create behavior in your release notes. |
Dimension | Required (or defaulted). Vector length. Must match your embedding model (for example 1536 for OpenAI text-embedding-3-small at default size, 384 for common MiniLM-style models). |
Distance metric | How similarity is measured between vectors: for example Cosine, Euclidean (L2), Manhattan, or Dot (inner product). Choose a metric aligned with your embedding model and normalization; cosine is common for normalized embeddings. |
Top K | Maximum number of hits to return from a similarity search. Must be a positive integer when set. |
Min score | Optional minimum similarity score; results below the threshold are dropped. Often interpreted between 0 and 1 depending on client and metric; leave empty to return all top-K matches without filtering by score. |
---
Connection settings
Field | Description |
|---|---|
Call timeout (ms) | Maximum time to wait for a single client call to finish. Default is often 60000 ms (60 seconds). |
Connection timeout (ms) | Maximum time to wait while establishing a connection. Default is often 20000 ms (20 seconds). |
Keep alive | When enabled, keeps HTTP connections warm (for example HTTP/2 keep-alive), which can reduce latency on steady workloads. Default is often on. |
Keep alive time (ms) | Interval between keep-alive signals when the connection is idle. Default is often 30000 ms (30 seconds). |
Keep alive timeout (ms) | How long to wait for a keep-alive acknowledgment. Default is often 5000 ms (5 seconds). |
Idle timeout (ms) | How long a connection may stay idle before the client closes it. Default is often 600000 ms (10 minutes / 600 seconds). |
---
Retry settings
Field | Description |
|---|---|
Max retries | How many times to retry after a failed request. Default is often 3. |
Initial backoff (ms) | Delay before the first retry. Default is often 200 ms. |
Max backoff (ms) | Upper cap on delay between retries. Default is often 1000 ms (1 second). |
Backoff multiplier | Factor applied between retry attempts (exponential backoff). Default is often 2. |
---
Embedding model (related UI)
Field | Description |
|---|---|
Source | None — skip embedding configuration only when valid for this flow. Pre-configured — use a profile from AI Services > Embedding models. Custom — configure embedding inline. |
Provider | After you choose a source, select the embedding provider (OpenAI, Mistral, Ollama, and so on). That drives vector generation; Qdrant settings define storage, metric, and search behavior. |
---
We're glad. Tell us how this page helped.
We're sorry. Can you tell us what didn't work for you?
Share additional feedback to help us improve.
0/255
|
