Building Enterprise AI Autonomous Agents in Self-Hosted n8n with Qdrant Vector Memory

Overview

n8n is an open-source workflow automation platform that supports AI agent orchestration. By pairing self-hosted n8n with a local Qdrant vector database and an Ollama inference container, engineering teams can build fully sovereign AI agents that process sensitive documents without sending data to third-party APIs.

This guide provides the complete Docker Compose architecture for launching n8n with Qdrant vector memory.

Architecture Blueprint

┌────────────────────────────────────────────────────────┐
│                   SELF-HOSTED DOCKER STACK             │
│                                                        │
│  ┌────────────────────────┐  Trigger / Webhook         │
│  │ n8n Automation Engine  │ ◄───────────────────────── │
│  └───────────┬────────────┘                            │
│              │ Vector Memory Queries (Port 6333)       │
│              ▼                                         │
│  ┌────────────────────────┐                            │
│  │ Qdrant Vector DB      │                            │
│  └───────────┬────────────┘                            │
│              │ Embedding & Chat Completions            │
│              ▼                                         │
│  ┌────────────────────────┐                            │
│  │ Local Ollama Service   │                            │
│  └────────────────────────┘                            │
└────────────────────────────────────────────────────────┘

Docker Compose Manifest

Create /opt/n8n-agent-stack/docker-compose.yml:

version: "3.8"

services:
  qdrant:
    image: qdrant/qdrant:v1.9.2
    container_name: qdrant_db
    restart: unless-stopped
    volumes:
      - qdrant_storage:/qdrant/storage
    ports:
      - "6333:6333"

  n8n:
    image: docker.n8n.io/n8nio/n8n:latest
    container_name: n8n_engine
    restart: unless-stopped
    ports:
      - "5678:5678"
    environment:
      - N8N_BASIC_AUTH_ACTIVE=true
      - N8N_BASIC_AUTH_USER=admin
      - N8N_BASIC_AUTH_PASSWORD=${N8N_PASSWORD}
    volumes:
      - n8n_data:/home/node/.n8n
    depends_on:
      - qdrant

volumes:
  qdrant_storage:
  n8n_data:

Summary Checklist

• Deployed n8n and Qdrant in shared Docker environment.
• Configured Qdrant vector store nodes in n8n AI workflow canvas.
• Connected local Ollama embeddings model (nomic-embed-text).

Self-Hosted Autonomous Agent Architecture

Building sovereign AI agent workflows requires decoupling workflow orchestration (n8n), vector similarity indexing (Qdrant), and LLM text generation (vLLM / Ollama) into isolated container networks.

In this architecture, n8n executes autonomous agent chains by fetching relevant context from Qdrant via gRPC hybrid search, injecting vector payload context into system prompts, and calling local LLM endpoints—ensuring zero data leakage to external cloud APIs.

+-------------------------------------------------------------------------+
|                  Sovereign Autonomous Agent Topology                    |
|  +-------------------------------------------------------------------+  |
|  | n8n Workflow Engine (Agent Core)                                 |  |
|  +-------------------------------------------------------------------+  |
|         || Vector Query (gRPC Port 6334)         || Inference API       |
|         \/                                       \/                     |
|  +-----------------------------------+   +---------------------------+  |
|  | Qdrant Vector Engine              |   | Ollama / vLLM LLM Engine  |  |
|  | (HNSW Index + Sparse Vectors)     |   | (Llama 3.3 / Qwen 2.5)   |  |
|  +-----------------------------------+   +---------------------------+  |
+-------------------------------------------------------------------------+

Step 1: Production Docker Compose Stack (docker-compose.yml)

version: '3.8'

services:
  n8n:
    image: docker.n8n.io/n8nio/n8n:latest
    container_name: n8n-agent-core
    restart: always
    ports:
      - "5678:5678"
    environment:
      - N8N_HOST=n8n.local
      - N8N_PORT=5678
      - N8N_PROTOCOL=http
      - NODE_ENV=production
      - WEBHOOK_URL=http://localhost:5678/
      - EXECUTIONS_DATA_PRUNE=true
      - EXECUTIONS_DATA_MAX_AGE=168
    volumes:
      - n8n_data:/home/node/.n8n

  qdrant:
    image: qdrant/qdrant:latest
    container_name: qdrant-vector-db
    restart: always
    ports:
      - "6333:6333" # HTTP REST
      - "6334:6334" # gRPC
    volumes:
      - qdrant_storage:/qdrant/storage

volumes:
  n8n_data:
  qdrant_storage:

Step 2: Python Script for Vector Collection Initialization

Execute vector schema initialization before connecting n8n workflows:

from qdrant_client import QdrantClient
from qdrant_client.models import VectorParams, Distance, OptimizersConfigDiff

client = QdrantClient(host="localhost", port=6333)

collection_name = "enterprise_knowledge_base"

if not client.collection_exists(collection_name):
    client.create_collection(
        collection_name=collection_name,
        vectors_config=VectorParams(
            size=1536,  # Compatible with bge-large or OpenAI embeddings
            distance=Distance.COSINE
        ),
        optimizers_config=OptimizersConfigDiff(
            default_segment_number=4
        )
    )
    print(f"[+] Qdrant collection '{collection_name}' successfully created.")