Weaviate Local Setup Skill

Run Weaviate locally using Docker for development, testing, and avoiding network restrictions in Claude Desktop/Web.

Why Use Local Weaviate?

Benefits:

• No network restrictions in Claude Desktop/Web
• Free (no cloud costs)
• Full control over data and configuration
• Fast development/testing cycles
• Works offline
• No API key management for cloud instances

Best for:

• Development and testing
• Learning Weaviate
• Working in Claude Desktop with network restrictions
• Privacy-sensitive projects

Prerequisites

Required:

• Docker Desktop installed and running
• Available ports: 8080 (Weaviate), 8081 (optional for Weaviate Console)
• Python 3.8+ installed

Optional (for specific vectorizers):

• OpenAI API key (for text2vec-openai)
• Cohere API key (for text2vec-cohere)
• Anthropic API key (for generative-anthropic)

Python Environment Setup

IMPORTANT: Do this FIRST before using any Weaviate skills!

Claude will create a virtual environment and install dependencies to avoid conflicts with your system Python.

Step 1: Create Virtual Environment

# Navigate to the weaviate-claude-skills directory
cd ~/Documents/weaviate-claude-skills

# Create virtual environment
python3 -m venv .venv

# Activate it
source .venv/bin/activate  # macOS/Linux
# OR
.venv\Scripts\activate     # Windows

Step 2: Install Dependencies

# Install required packages
pip install weaviate-client python-dotenv

# Optional: Install additional packages for specific vectorizers
pip install openai  # If using OpenAI vectorizer
pip install cohere  # If using Cohere vectorizer

Or install everything at once:

pip install -r requirements.txt

Step 3: Verify Installation

import subprocess
import sys

# Check if dependencies are installed
try:
    import weaviate
    from dotenv import load_dotenv
    print("✅ All required packages are installed!")
except ImportError as e:
    print(f"❌ Missing package: {e}")
    print("Installing dependencies...")
    subprocess.check_call([sys.executable, "-m", "pip", "install",
                          "weaviate-client", "python-dotenv"])
    print("✅ Dependencies installed successfully!")

When Using Claude

Claude will check and ensure dependencies are installed before running any Weaviate code.

If you see errors about missing packages, Claude will:

1. Check if the virtual environment exists
2. Create it if needed
3. Install required dependencies
4. Proceed with your request

Pro Tip: Keep the virtual environment activated throughout your Claude session for best results.

Quick Start

Option 1: Basic Setup (No API Keys Required)

Use Weaviate's built-in vectorizer (no external API needed):

# Start Weaviate with transformers (runs locally, no API key)
docker run -d \
  --name weaviate \
  -p 8080:8080 \
  -p 50051:50051 \
  -e QUERY_DEFAULTS_LIMIT=25 \
  -e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true \
  -e PERSISTENCE_DATA_PATH='/var/lib/weaviate' \
  -e DEFAULT_VECTORIZER_MODULE='text2vec-transformers' \
  -e ENABLE_MODULES='text2vec-transformers' \
  -e TRANSFORMERS_INFERENCE_API='http://t2v-transformers:8080' \
  -e CLUSTER_HOSTNAME='node1' \
  semitechnologies/weaviate:1.28.1

# Start the transformers module
docker run -d \
  --name t2v-transformers \
  -e ENABLE_CUDA=0 \
  semitechnologies/transformers-inference:sentence-transformers-multi-qa-MiniLM-L6-cos-v1

Connection:

WEAVIATE_URL=localhost:8080
WEAVIATE_API_KEY=  # Leave empty for local

Option 2: With OpenAI Vectorizer

Use OpenAI embeddings (requires OpenAI API key):

docker run -d \
  --name weaviate \
  -p 8080:8080 \
  -p 50051:50051 \
  -e QUERY_DEFAULTS_LIMIT=25 \
  -e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true \
  -e PERSISTENCE_DATA_PATH='/var/lib/weaviate' \
  -e DEFAULT_VECTORIZER_MODULE='text2vec-openai' \
  -e ENABLE_MODULES='text2vec-openai,generative-openai' \
  -e CLUSTER_HOSTNAME='node1' \
  semitechnologies/weaviate:1.28.1

Connection (.env):

WEAVIATE_URL=localhost:8080
WEAVIATE_API_KEY=  # Leave empty
OPENAI_API_KEY=your-openai-key-here

Option 3: Docker Compose (Recommended for Production)

See

docker-compose.yml

# Start Weaviate
docker-compose up -d

# Check status
docker-compose ps

# View logs
docker-compose logs -f weaviate

# Stop Weaviate
docker-compose down

# Stop and remove data
docker-compose down -v

Docker Commands Reference

Container Management

# Start Weaviate
docker start weaviate

# Stop Weaviate
docker stop weaviate

# Restart Weaviate
docker restart weaviate

# Check if running
docker ps | grep weaviate

# View logs
docker logs weaviate

# Follow logs in real-time
docker logs -f weaviate

# Remove container (keeps data)
docker rm weaviate

# Remove container and data volume
docker rm -v weaviate

Health Checks

# Check if Weaviate is ready
curl http://localhost:8080/v1/.well-known/ready

# Check Weaviate metadata
curl http://localhost:8080/v1/meta

# Expected response:
# {"hostname":"http://[::]:8080","modules":{...},"version":"1.28.1"}

Data Persistence

Weaviate data is stored in Docker volumes. To persist data across container restarts:

# Create a named volume
docker volume create weaviate-data

# Run with named volume
docker run -d \
  --name weaviate \
  -p 8080:8080 \
  -v weaviate-data:/var/lib/weaviate \
  -e PERSISTENCE_DATA_PATH='/var/lib/weaviate' \
  -e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true \
  semitechnologies/weaviate:1.28.1

# List volumes
docker volume ls

# Inspect volume
docker volume inspect weaviate-data

# Backup data (export volume to tar)
docker run --rm -v weaviate-data:/data -v $(pwd):/backup \
  ubuntu tar czf /backup/weaviate-backup.tar.gz /data

# Restore data (import tar to volume)
docker run --rm -v weaviate-data:/data -v $(pwd):/backup \
  ubuntu tar xzf /backup/weaviate-backup.tar.gz -C /

Python Connection Code

Once Weaviate is running locally, connect using:

import weaviate
import os
from dotenv import load_dotenv

load_dotenv()

# Connect to local Weaviate (no authentication)
client = weaviate.connect_to_local(
    host="localhost",
    port=8080,
    grpc_port=50051
)

try:
    print("✅ Connected to local Weaviate!")

    # Check if ready
    if client.is_ready():
        print("🟢 Weaviate is ready")

        # Get metadata
        meta = client.get_meta()
        print(f"📦 Version: {meta['version']}")

except Exception as e:
    print(f"❌ Error: {e}")

finally:
    client.close()

Alternative connection with API key header (if enabled):

client = weaviate.connect_to_local(
    host="localhost",
    port=8080,
    grpc_port=50051,
    headers={
        "X-OpenAI-Api-Key": os.getenv("OPENAI_API_KEY")
    }
)

Environment Variables for Local Setup

Update your

.env

# Local Weaviate Connection
WEAVIATE_URL=localhost:8080
WEAVIATE_API_KEY=  # Leave empty for local instances

# Vectorizer API Keys (only needed if using these vectorizers)
OPENAI_API_KEY=your-openai-api-key
COHERE_API_KEY=your-cohere-api-key
ANTHROPIC_API_KEY=your-anthropic-api-key
HUGGINGFACE_API_KEY=your-huggingface-api-key

Available Vectorizer Modules

Text Vectorizers

text2vec-transformers

text2vec-openai

text2vec-cohere

text2vec-huggingface

text2vec-palm

Multi-Modal Vectorizers

multi2vec-clip

multi2vec-bind

img2vec-neural

Generative Modules (for RAG)

generative-openai

generative-cohere

generative-anthropic

generative-palm

Docker Image Tags

Stable versions:

• semitechnologies/weaviate:1.28.1

  (latest stable)

• semitechnologies/weaviate:1.27.0

• semitechnologies/weaviate:1.26.0

Preview/Beta:

• semitechnologies/weaviate:preview

Module images:

• semitechnologies/transformers-inference:sentence-transformers-multi-qa-MiniLM-L6-cos-v1

• semitechnologies/transformers-inference:sentence-transformers-all-MiniLM-L6-v2

• semitechnologies/multi2vec-clip:sentence-transformers-clip-ViT-B-32

Common Issues & Troubleshooting

Port Already in Use

# Check what's using port 8080
lsof -i :8080

# Kill the process (if needed)
kill -9 <PID>

# Or run Weaviate on a different port
docker run -d -p 8081:8080 --name weaviate ...

Container Won't Start

# Check logs
docker logs weaviate

# Remove and recreate
docker rm -f weaviate
docker run ...

Cannot Connect from Python

# Make sure you're using the correct connection method for local
client = weaviate.connect_to_local()  # NOT connect_to_weaviate_cloud()

# Check Weaviate is actually running
# curl http://localhost:8080/v1/.well-known/ready

Data Not Persisting

Make sure you're using a volume:

docker run -v weaviate-data:/var/lib/weaviate ...

Module Not Available

Enable the module in the

ENABLE_MODULES

-e ENABLE_MODULES='text2vec-openai,generative-openai'

Performance Tuning

Memory Limits

# Set memory limits
docker run -d \
  --name weaviate \
  --memory="4g" \
  --memory-swap="4g" \
  ...

CPU Limits

# Limit CPUs
docker run -d \
  --name weaviate \
  --cpus="2.0" \
  ...

Query Defaults

# Increase default query limit
-e QUERY_DEFAULTS_LIMIT=100

# Set maximum query results
-e QUERY_MAXIMUM_RESULTS=10000

Weaviate Console (Optional UI)

Run the Weaviate Console for a web UI:

docker run -d \
  --name weaviate-console \
  -p 8081:8080 \
  semitechnologies/weaviate-console:latest

# Access at: http://localhost:8081
# Enter Weaviate URL: http://localhost:8080

Migration: Cloud to Local

Export from Cloud

import weaviate
import json

# Connect to cloud
cloud_client = weaviate.connect_to_weaviate_cloud(
    cluster_url=os.getenv("WEAVIATE_URL"),
    auth_credentials=weaviate.auth.Auth.api_key(os.getenv("WEAVIATE_API_KEY"))
)

collection = cloud_client.collections.get("YourCollection")

# Export all objects
objects = []
for item in collection.iterator():
    objects.append({
        "properties": item.properties,
        "vector": item.vector
    })

# Save to file
with open("export.json", "w") as f:
    json.dump(objects, f)

cloud_client.close()

Import to Local

# Connect to local
local_client = weaviate.connect_to_local()

# Create collection (same schema as cloud)
# ... create collection code ...

# Import data
collection = local_client.collections.get("YourCollection")

with collection.batch.dynamic() as batch:
    for obj in objects:
        batch.add_object(properties=obj["properties"], vector=obj.get("vector"))

local_client.close()

Best Practices

1. Use Docker Compose for complex setups with multiple modules
2. Always use volumes for data persistence
3. Monitor logs during development:

   docker logs -f weaviate

4. Backup regularly using volume exports
5. Use transformers module for development (no API costs)
6. Switch to OpenAI for production (better quality)
7. Set memory limits to prevent OOM crashes
8. Test connections before importing large datasets

Integration with Other Skills

This skill works perfectly with:

• weaviate-connection

  - Just use

  localhost:8080

  as the URL

• weaviate-collection-manager

  - Create collections on local instance

• weaviate-data-ingestion

  - Upload data locally (faster, no network limits)

• weaviate-query-agent

  - Query local data (faster responses)

Example Workflow

# 1. Start Weaviate locally
docker run -d --name weaviate -p 8080:8080 \
  -e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true \
  -e PERSISTENCE_DATA_PATH='/var/lib/weaviate' \
  semitechnologies/weaviate:1.28.1

# 2. Verify it's running
curl http://localhost:8080/v1/.well-known/ready

# 3. Update .env
# WEAVIATE_URL=localhost:8080
# WEAVIATE_API_KEY=

# 4. Use other skills normally
# Claude: "Connect to my local Weaviate instance"
# Claude: "Create a collection called Documents"
# Claude: "Upload these 100 PDFs"
# Claude: "Search for information about X"

Resources

• Weaviate Docker Installation
• Weaviate Modules
• Docker Documentation
• Weaviate Python Client

Built for the Weaviate Skills Collection

Questions? Check the main README or Weaviate documentation.