Getting started - QWED Documentation

Learn how to set up and run QWED in your development environment.

Don’t call your LLM directly. QWED runs as a backend server with your LLM API keys configured in environment variables. See Common pitfalls for details.

Architecture overview

QWED uses a backend server model:

Your application
  ↓ (SDK calls)
QWED backend server (you run this)
  ├─ Your LLM API key (from .env)
  ├─ LLM calls (OpenAI/Anthropic/etc)
  ├─ Formal verifiers (SymPy, Z3)
  └─ Returns verified result

Step 1: Clone and install

# Clone the repository
git clone https://github.com/QWED-AI/qwed-verification.git
cd qwed-verification

# Install Python dependencies
pip install -r requirements.txt

Step 2: Configure LLM provider

QWED is model agnostic — select any LLM based on your needs:

Create environment file

# Copy example configuration
cp .env.example .env

Select your LLM provider

Edit .env and add your API key for one provider:

OpenAI
Anthropic Claude
Azure OpenAI
Google Gemini

# .env
ACTIVE_PROVIDER=openai
OPENAI_API_KEY=sk-proj-...

Get an API key: https://platform.openai.com/api-keys

# .env
ACTIVE_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-api03-...

Get an API key: https://console.anthropic.com

# .env
ACTIVE_PROVIDER=azure_openai
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_KEY=...
AZURE_OPENAI_DEPLOYMENT=gpt-4
AZURE_OPENAI_API_VERSION=2024-02-01

Get Credentials: https://portal.azure.com → Azure OpenAI Service

# .env
ACTIVE_PROVIDER=gemini
GOOGLE_API_KEY=AIzaSy...

Get an API key: https://makersuite.google.com/app/apikey

For AWS Bedrock and more options, see LLM configuration.

Step 3: Run the QWED backend server

# Start the backend server
python -m qwed_api

# You should see:
# INFO: Uvicorn running on http://0.0.0.0:8000

Keep this terminal running. The server must stay active.

Step 4: Install the SDK

In a new terminal, install the QWED SDK in your project:

Python
TypeScript
Go

pip install qwed

npm install @qwed-ai/sdk

go get github.com/qwed-ai/qwed-go

Step 5: Use the SDK

Connect your application to the local QWED backend:

Python
TypeScript
Go

from qwed import QWEDClient

# Connect to local backend
client = QWEDClient(
    api_key="qwed_local",  # Local auth key
    base_url="http://localhost:8000"
)

# Verify a mathematical claim
result = client.verify("Is 2+2 equal to 4?")

print(result.verified)  # True
print(result.evidence)  # {"calculated": 4, "claimed": 4}

import { QWEDClient } from '@qwed-ai/sdk';

// Connect to local backend
const client = new QWEDClient({
  apiKey: 'qwed_local',
  baseUrl: 'http://localhost:8000'
});

// Verify a mathematical claim
const result = await client.verify('Is 2+2 equal to 4?');

console.log(result.verified);  // true
console.log(result.evidence);  // {calculated: 4, claimed: 4}

package main

import (
    "context"
    "fmt"
    "github.com/qwed-ai/qwed-go"
)

func main() {
    client := qwed.NewClient("qwed_local", "http://localhost:8000")
    
    result, err := client.Verify(context.Background(), "Is 2+2 equal to 4?")
    if err != nil {
        panic(err)
    }
    
    fmt.Println(result.Verified)  // true
}

Verify installation

Run this test to ensure everything is working:

Python
TypeScript

from qwed import QWEDClient

client = QWEDClient(
    api_key="qwed_local",
    base_url="http://localhost:8000"
)

# Test 1: Math verification
assert client.verify("2+2=4").verified == True
print("✅ Math verification works!")

# Test 2: Error detection
assert client.verify("2+2=5").verified == False
print("✅ Error detection works!")

# Test 3: Code security
result = client.verify_code("eval(user_input)", language="python")
assert result.blocked == True
print("✅ Security detection works!")

print("\n🎉 QWED is working correctly!")

import { QWEDClient } from '@qwed-ai/sdk';

const client = new QWEDClient({
  apiKey: 'qwed_local',
  baseUrl: 'http://localhost:8000'
});

async function testIntegration() {
  // Test 1: Math verification
  const result1 = await client.verify('2+2=4');
  console.assert(result1.verified === true);
  console.log('✅ Math verification works!');

  // Test 2: Error detection  
  const result2 = await client.verify('2+2=5');
  console.assert(result2.verified === false);
  console.log('✅ Error detection works!');

  // Test 3: Code security
  const result3 = await client.verifyCode('eval(user_input)', { language: 'python' });
  console.assert(result3.blocked === true);
  console.log('✅ Security detection works!');

  console.log('\n🎉 QWED is working correctly!');
}

testIntegration();

Next steps

Once your integration is working:

Common pitfalls — avoid integration mistakes
Testing your integration — validate your setup
Production deployment — deploy with confidence

Troubleshooting

Backend won’t start

Problem:

Error: ANTHROPIC_API_KEY environment variable not set

Solution: Check your .env file has the correct API key for your selected provider.

SDK can’t connect

Problem:

Connection refused to http://localhost:8000

Solution:

Make sure backend is running (python -m qwed_api)
Check backend terminal for errors
Verify port 8000 is not blocked

Wrong results

Problem: Verification results are incorrect or unexpected. Solution:

Check which LLM provider you configured
Verify your API key is valid
See Testing guide for detailed diagnostics

​Architecture overview

​Step 1: Clone and install

​Step 2: Configure LLM provider

​Create environment file

​Select your LLM provider

​Step 3: Run the QWED backend server

​Step 4: Install the SDK

​Step 5: Use the SDK

​Verify installation

​Next steps

​Troubleshooting

​Backend won’t start

​SDK can’t connect

​Wrong results

​Need help?

Architecture overview

Step 1: Clone and install

Step 2: Configure LLM provider

Create environment file

Select your LLM provider

Step 3: Run the QWED backend server

Step 4: Install the SDK

Step 5: Use the SDK

Verify installation

Next steps

Troubleshooting

Backend won’t start

SDK can’t connect

Wrong results

Need help?