Qdrant setup

Overview

Qdrant is a vector database used by the FlowX AI Platform to store and search the dense and sparse embeddings that power Knowledge Base indexing and retrieval (RAG). It is consumed by the AI Platform services — primarily the embedder and knowledgebase-rag services.

Bundled with the AI Platform. Qdrant is shipped by FlowX as a subchart of the AI Platform Helm chart — it is deployed and version-managed by that chart, not installed separately. Configure it through the qdrant: block of your AI Platform values. The version is set by the chart and is not pinned in the FlowX component matrix.

For more information about Qdrant, see the Qdrant documentation.

Prerequisites

Infrastructure

Kubernetes cluster with admin access
Persistent storage (block storage) for vector data and snapshots
Network connectivity between Qdrant and the AI Platform services

FlowX Integration

AI Platform chart deployed (Qdrant is a subchart of it)
A Kubernetes secret holding the Qdrant API key
AI services configured with the Qdrant connection variables

Deployment

Qdrant is enabled and configured through the qdrant: block of the AI Platform Helm values. A typical configuration:

qdrant:
  resources:
    requests:
      cpu: 200m
      memory: 1Gi
    limits:
      cpu: 2
      memory: 2Gi
  config:
    cluster:
      enabled: false
  apiKey:
    valueFrom:
      secretKeyRef:
        name: qdrant-generic
        key: apikey

The Qdrant service is exposed inside the cluster as qdrant-headless starting with 5.9.2 (ai-platform-qdrant-headless on 5.9.0/5.9.1), serving gRPC on port 6334 and the REST/health API on port 6333. If you reference the Qdrant host explicitly in any service configuration, update it when upgrading to 5.9.2 — see the 5.9.2 deployment guidelines.

API key secret

Qdrant authenticates clients with an API key, supplied through a Kubernetes secret. Create a secret named qdrant-generic with an apikey entry:

apiVersion: v1
kind: Secret
metadata:
  name: qdrant-generic
type: Opaque
data:
  apikey: <base64-encoded-api-key>

Generate a secure API key with openssl rand -base64 32 and base64-encode it for the secret.

The API key in the qdrant-generic secret must match the QDRANT_CONNECTION_API_KEY configured on the AI Platform services that connect to Qdrant.

Storage

Provision persistent volumes for Qdrant. Recommended baseline sizes:

Volume	Size	Purpose
Qdrant data	30Gi	Vector embeddings
Qdrant snapshots	30Gi	Backup snapshots

Size the data volume to your Knowledge Base footprint. Vector storage grows with the number and dimensionality of indexed documents.

Connecting the AI Platform services

The AI Platform services read the following variables to reach Qdrant. The connection values are set by default in the chart; supply the API key from the qdrant-generic secret.

Variable	Description	Default Value
`QDRANT_CONNECTION_GRPC_ENDPOINT`	Qdrant gRPC endpoint	—
`QDRANT_CONNECTION_API_KEY`	Qdrant API key	—
`QDRANT__CLUSTER__ENABLED`	Enable Qdrant cluster mode	`true`

The knowledgebases_design collection is created automatically by the embedder service on first use — no manual provisioning step is required.

Search tuning (knowledgebase-rag)

These limits control the fanout of dense and sparse Qdrant queries inside the knowledgebase-rag service. Both are read at request time, so changing them does not require a redeploy beyond the env-var rollout.

Variable	Description	Default Value
`QDRANT_PREFETCH_LIMIT`	Maximum points returned by each dense/sparse prefetch stage in a hybrid search before fusion. Lower values reduce Qdrant memory pressure on large KBs.	`100`
`QDRANT_FUSION_LIMIT`	Maximum points returned after RRF fusion (hybrid search) or by a single-stage dense/keyword search. Caps the reranker’s input set.	`80`

Defaults are tuned for typical Knowledge Bases. Raise them only if recall is insufficient for very large or long-tailed KBs; lower them if you see Qdrant memory pressure or slow queries under filter-heavy workloads.

Planner seeding

Variable	Description	Default Value
`PLANNER_SEED_ENABLED`	When `true`, the `planner` service seeds its Qdrant benchmark collection on startup. Leave off unless seeding is required.	`false`

Verification

Check Qdrant pods

Ensure the Qdrant pods are running:

kubectl get pods -l app.kubernetes.io/name=qdrant

Test the health endpoint

Verify Qdrant is responding on the REST port:

kubectl port-forward svc/qdrant-headless 6333:6333   # 5.9.0/5.9.1: svc/ai-platform-qdrant-headless
# From another terminal
curl http://localhost:6333/healthz

Check AI Platform integration

Review the embedder logs for a successful Qdrant connection:

kubectl logs -l app=embedder | grep -i qdrant   # 5.9.0/5.9.1: -l app=ai-platform-embedder

Common connection failures: a missing QDRANT_CONNECTION_API_KEY, a key that does not match the qdrant-generic secret, or the Qdrant cluster not being fully initialized.

AI Platform setup

Full AI Platform deployment, including the services that consume Qdrant

Knowledge Base RAG

How retrieval-augmented generation uses the vector store

Microservices

AI Platform

Plugins

Observability

Access management

Qdrant setup

Overview

Prerequisites

Infrastructure

FlowX Integration

Deployment

API key secret

Storage

Connecting the AI Platform services

Search tuning (knowledgebase-rag)

Planner seeding

Verification

AI Platform setup

Knowledge Base RAG

​Overview

​Prerequisites

Infrastructure

FlowX Integration

​Deployment

​API key secret

​Storage

​Connecting the AI Platform services

​Search tuning (knowledgebase-rag)

​Planner seeding

​Verification

​Related resources

AI Platform setup

Knowledge Base RAG

Overview

Prerequisites

Deployment

API key secret

Storage

Connecting the AI Platform services

Search tuning (knowledgebase-rag)

Planner seeding

Verification

Related resources