Vector Databases

Vector databases specialized for similarity search, RAG (Retrieval-Augmented Generation) pipelines, and AI-powered applications.

Available Services

Qdrant

Port: 6333 (REST), 6334 (gRPC) | Memory: 512 MB | Maturity: StableHigh-performance vector similarity search engine for building RAG pipelines, semantic search, and AI-powered recommendation systems.Features:

Fast vector search
Filtering and payload
HNSW algorithm
Quantization support
Distributed mode
Rust-based performance

OpenClaw Integration:

Skill: qdrant-memory
Environment: QDRANT_HOST, QDRANT_PORT

Recommends: RedisDocumentation

ChromaDB

Port: 8100 | Memory: 512 MB | Maturity: StableOpen-source AI-native vector database with simple APIs for storing, searching, and filtering vectors.Features:

Easy-to-use API
Multiple embedding models
Metadata filtering
Auto-embedding
Python and JavaScript clients
Lightweight

OpenClaw Integration:

Environment: CHROMADB_HOST, CHROMADB_PORT

Documentation

Milvus

Port: 19530 (API), 9091 (Metrics) | Memory: 2048 MB | Maturity: StableOpen-source vector database built for scalable similarity search and AI applications.Features:

Billion-scale vectors
Hybrid search
Multiple index types
GPU acceleration
Kubernetes-ready
Cloud-native architecture

OpenClaw Integration:

Environment: MILVUS_URI

Documentation

Weaviate

Port: 8082 (REST), 50051 (gRPC) | Memory: 1024 MB | Maturity: StableCloud-native vector database with built-in vectorization modules, hybrid search, and GraphQL API.Features:

GraphQL API
Built-in vectorizers
Hybrid search (vector + keyword)
Multi-tenancy
Replication
Schema-based

OpenClaw Integration:

Environment: WEAVIATE_HOST, WEAVIATE_PORT

Documentation

Usage Examples

RAG Pipeline Stack

npx create-better-openclaw \
  --services qdrant,ollama,open-webui \
  --yes

Research Agent Preset

npx create-better-openclaw --preset researcher --yes

This includes: Qdrant, SearXNG, Browserless, Redis

Knowledge Base Stack

npx create-better-openclaw \
  --services qdrant,postgresql,meilisearch \
  --yes

Vector Database Comparison

Database	Performance	Scalability	API Style	Hybrid Search	Memory
Qdrant	Excellent	Good	REST/gRPC	✅	512 MB
ChromaDB	Good	Moderate	REST	✅	512 MB
Milvus	Excellent	Excellent	REST/gRPC	✅	2048 MB
Weaviate	Excellent	Excellent	GraphQL/REST	✅	1024 MB

RAG Architecture Patterns

Basic RAG

Document → Embedding Model → Vector DB
Query → Embedding Model → Vector Search
Retrieved Context + Query → LLM → Response

Advanced RAG with Reranking

Document → Chunking → Embedding → Vector DB
Query → Vector Search (top 50)
Reranking (top 5)
Context + Query → LLM → Response

Text + Images → Embeddings → Vector DB
Query → Multi-modal Search
Retrieved Content → Multi-modal LLM → Response

Embedding Models

Popular Embedding Models

Model	Dimensions	Use Case	Provider
text-embedding-3-small	1536	General purpose	OpenAI
text-embedding-3-large	3072	High accuracy	OpenAI
all-MiniLM-L6-v2	384	Fast, local	Sentence Transformers
BAAI/bge-large-en	1024	English text	Open source
intfloat/e5-large	1024	Multi-lingual	Open source

Local Embedding with Ollama

# Pull embedding model
docker exec ollama ollama pull mxbai-embed-large

# Use in your application
curl http://localhost:11434/api/embeddings \
  -d '{"model": "mxbai-embed-large", "prompt": "Your text here"}'

Collection Management

Qdrant Collections

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

# Create collection
client.create_collection(
    collection_name="documents",
    vectors_config={"size": 384, "distance": "Cosine"}
)

# Insert vectors
client.upsert(
    collection_name="documents",
    points=[{
        "id": 1,
        "vector": embedding,
        "payload": {"text": "Document content"}
    }]
)

# Search
results = client.search(
    collection_name="documents",
    query_vector=query_embedding,
    limit=5
)

ChromaDB Collections

import chromadb

client = chromadb.HttpClient(host="localhost", port=8100)

# Create collection
collection = client.create_collection(name="documents")

# Add documents (auto-embedding)
collection.add(
    documents=["Document 1", "Document 2"],
    ids=["id1", "id2"],
    metadatas=[{"source": "web"}, {"source": "pdf"}]
)

# Query
results = collection.query(
    query_texts=["search query"],
    n_results=5
)

Optimization Tips

Qdrant Optimization

Index Type: Use HNSW for speed, quantization for memory
Payload: Store minimal metadata for better performance
Filtering: Use indexed payload fields for fast filtering
Batch Operations: Insert vectors in batches
Memory: Allocate sufficient RAM for index

ChromaDB Optimization

Embedding Function: Choose appropriate embedding model
Distance Metric: Use cosine similarity for most cases
Persistence: Enable persistence for production
Batch Size: Process documents in batches
Metadata: Keep metadata small and indexed

Milvus Optimization

Index Selection: Choose IVF_FLAT, IVF_SQ8, or HNSW
Segmentation: Configure segment size appropriately
Resource Groups: Allocate resources per workload
GPU Acceleration: Use GPU for large-scale search
Sharding: Distribute data across shards

Use Cases

Semantic Search

# Search documents by meaning, not keywords
# Example: "python web framework" finds Flask, Django, FastAPI

Question Answering

# Retrieve relevant context from knowledge base
# Pass context to LLM for accurate answers

Recommendation Systems

# Find similar products, articles, or content
# Based on embedding similarity

Document Chat

# Upload documents → Chunk → Embed → Store
# Chat interface retrieves relevant chunks for LLM

Image Search

# Store image embeddings (CLIP, etc.)
# Search by text or image similarity

Integration Examples

Qdrant + Ollama + Open WebUI

npx create-better-openclaw \
  --services qdrant,ollama,open-webui,redis \
  --yes

ChromaDB + Dify

npx create-better-openclaw \
  --services chromadb,dify,postgresql,redis \
  --yes

Milvus + LiteLLM + Flowise

npx create-better-openclaw \
  --services milvus,litellm,flowise \
  --yes

Monitoring and Maintenance

Health Checks

# Qdrant
curl http://localhost:6333/healthz

# ChromaDB
curl http://localhost:8100/api/v1/heartbeat

# Milvus
curl http://localhost:9091/healthz

# Weaviate
curl http://localhost:8082/v1/.well-known/ready

Metrics

# Qdrant metrics
curl http://localhost:6333/metrics

# Milvus metrics (Prometheus format)
curl http://localhost:9091/metrics

Backups

# Qdrant snapshot
curl -X POST http://localhost:6333/collections/documents/snapshots

# Copy data volumes
docker cp qdrant:/qdrant/storage ./qdrant-backup

Performance Benchmarks

Query Latency (approximate)

Database	1K vectors	100K vectors	1M vectors
Qdrant	<1ms	1-5ms	5-20ms
ChromaDB	<1ms	5-10ms	20-50ms
Milvus	<1ms	1-5ms	5-15ms
Weaviate	<1ms	5-10ms	10-30ms

Throughput (queries/sec)

Database	Single Node	Distributed
Qdrant	1000+	10000+
ChromaDB	500+	N/A
Milvus	2000+	20000+
Weaviate	1000+	10000+

Note: Performance varies based on vector dimensions, index type, and hardware.

Service Catalog

Skill Packs

​Vector Databases

​Available Services

Qdrant

ChromaDB

Milvus

Weaviate

​Usage Examples

​RAG Pipeline Stack

​Research Agent Preset

​Knowledge Base Stack

​Vector Database Comparison

​RAG Architecture Patterns

​Basic RAG

​Advanced RAG with Reranking

​Multi-Modal RAG

​Embedding Models

​Popular Embedding Models

​Local Embedding with Ollama

​Collection Management

​Qdrant Collections

​ChromaDB Collections

​Optimization Tips

​Qdrant Optimization

​ChromaDB Optimization

​Milvus Optimization

​Use Cases

​Semantic Search

​Question Answering

​Recommendation Systems

​Document Chat

​Image Search

​Integration Examples

​Qdrant + Ollama + Open WebUI

​ChromaDB + Dify

​Milvus + LiteLLM + Flowise

​Monitoring and Maintenance

​Health Checks

​Metrics

​Backups

​Performance Benchmarks

​Query Latency (approximate)

​Throughput (queries/sec)

Build docs developers (and LLMs) love

Vector Databases

Available Services

Usage Examples

RAG Pipeline Stack

Research Agent Preset

Knowledge Base Stack

Vector Database Comparison

RAG Architecture Patterns

Basic RAG

Advanced RAG with Reranking

Multi-Modal RAG

Embedding Models

Popular Embedding Models

Local Embedding with Ollama

Collection Management

Qdrant Collections

ChromaDB Collections

Optimization Tips

Qdrant Optimization

ChromaDB Optimization

Milvus Optimization

Use Cases

Semantic Search

Question Answering

Recommendation Systems

Document Chat

Image Search

Integration Examples

Qdrant + Ollama + Open WebUI

ChromaDB + Dify

Milvus + LiteLLM + Flowise

Monitoring and Maintenance

Health Checks

Metrics

Backups

Performance Benchmarks

Query Latency (approximate)

Throughput (queries/sec)