Integrating W&B Inference with Claude Code: A step-by-step guide

import openai

client = openai.OpenAI(
    base_url='https://api.inference.wandb.ai/v1',
    api_key="your-wandb-api-key-here"",  # Replace with your actual key
    project="your-team/your-project",   # Replace with your team/project
)


try:
    response = client.chat.completions.create(
        model="Qwen/Qwen3-Coder-480B-A35B-Instruct",
        messages=[
            {"role": "system", "content": "You are a helpful coding assistant."},
            {"role": "user", "content": "Write a simple Python function to calculate factorial."}
        ],
        max_tokens=1000,
        temperature=0.7
    )
    
    print("✅ W&B Inference connection successful!")
    print("Response:", response.choices[0].message.content)
    print("Model used:", response.model)
    print("Tokens used:", response.usage.total_tokens if response.usage else "N/A")


except Exception as e:
    print("❌ Connection failed:", str(e))

pip install openai
python test_wb_inference.py

# W&B Inference configuration
export OPENAI_API_KEY="your-wandb-api-key-here""
export OPENAI_BASE_URL="https://api.inference.wandb.ai/v1"
export OPENAI_PROJECT="your-team/your-project"

# Optional: Disable non-essential traffic for better performance
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

echo "✅ Claude Code configured for W&B Inference"
echo "Base URL: $OPENAI_BASE_URL"
echo "Project: $OPENAI_PROJECT"
echo ""
echo "You can now run Claude Code commands like:"
echo "  claude 'Write a Python function to sort a list'"
echo "  claude --help"

chmod +x setup_claude_wb.sh

# Load the configuration
source setup_claude_wb.sh

# Test with a simple command
claude "Write a simple 'Hello, World!' program in Python"

# Test with a more complex coding task
claude "Create a Python class for a basic calculator with add, subtract, multiply, and divide methods"

# For heavy coding work - use the most capable model
export CLAUDE_DEFAULT_MODEL="Qwen/Qwen3-Coder-480B-A35B-Instruct"

# For quick tasks - use the faster model  
export CLAUDE_DEFAULT_MODEL="Qwen/Qwen2.5-Coder-32B-Instruct"

# Add to your ~/.bashrc or ~/.zshrc
echo 'export OPENAI_API_KEY="your-wandb-api-key-here"' >> ~/.bashrc
echo 'export OPENAI_BASE_URL="https://api.inference.wandb.ai/v1"' >> ~/.bashrc  
echo 'export OPENAI_PROJECT="your-team/your-project"' >> ~/.bashrc
echo 'export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1' >> ~/.bashrc

# Reload your shell
source ~/.bashrc

# project-a-config.sh
export OPENAI_PROJECT="mycompany/project-a"
export CLAUDE_PROJECT_CONTEXT="Working on project A - focus on React/TypeScript"

# project-b-config.sh  
export OPENAI_PROJECT="mycompany/project-b"
export CLAUDE_PROJECT_CONTEXT="Working on project B - focus on Python/Django"

# In VS Code terminal, source your configuration
source setup_claude_wb.sh

# Now use Claude Code within VS Code
claude "Add error handling to this function" --file src/utils.py
claude "Write unit tests for this class" --file src/models.py

# Generate commit messages
git diff --cached | claude "Write a concise commit message for these changes"

# Code review assistance
git diff main..feature-branch | claude "Review this code for potential issues and suggest improvements"

# Documentation generation
claude "Generate README documentation for this project" --file package.json

// VS Code settings.json for extensions like "Claude" or "AI Coder"
{
  "aiCoder.openai.baseUrl": "https://api.inference.wandb.ai/v1",
  "aiCoder.openai.apiKey": "your-wandb-api-key",
  "aiCoder.openai.model": "Qwen/Qwen3-Coder-480B-A35B-Instruct"
}

# usage_tracker.py - Monitor your Claude Code usage
import openai
import weave

# Initialize Weave for automatic logging
weave.init("your-team/your-project")

client = openai.OpenAI(
    base_url='https://api.inference.wandb.ai/v1',
    api_key="your-wandb-api-key",
    project="your-team/your-project",
)

# Your Claude Code requests will now be automatically logged to W&B

# Verify your API key
curl -H "Authorization: Bearer your-api-key" \
     -H "OpenAI-Project: your-team/your-project" \
     https://api.inference.wandb.ai/v1/models

# Should return a list of available models

# List available models
curl -H "Authorization: Bearer your-api-key" \
     -H "OpenAI-Project: your-team/your-project" \
     https://api.inference.wandb.ai/v1/models | jq '.data[].id'

# Verify environment variables are set
echo "API Key: ${OPENAI_API_KEY:0:10}..."
echo "Base URL: $OPENAI_BASE_URL"
echo "Project: $OPENAI_PROJECT"

# Test with verbose output
claude --debug "Simple test query"

# Add delays between requests if hitting rate limits
export CLAUDE_REQUEST_DELAY=1  # 1 second delay between requests

#!/bin/bash
# daily_dev_setup.sh

# Load W&B configuration
source setup_claude_wb.sh

# Start development session
echo "🚀 Starting development session with W&B Inference"

# Quick status check
claude "What's the current date and time?" > /dev/null 2>&1
if [ $? -eq 0 ]; then
    echo "✅ Claude Code + W&B Inference ready"
else
    echo "❌ Configuration issue - check your setup"
    exit 1
fi

# Optional: Start your IDE
# code .

#!/bin/bash
# code_review.sh

# Review staged changes
if git diff --cached --quiet; then
    echo "No staged changes to review"
    exit 1
fi

echo "🔍 Reviewing staged changes..."
git diff --cached | claude "Review this code for:
1. Potential bugs or issues
2. Code quality and best practices  
3. Security concerns
4. Performance implications
5. Suggestions for improvement

Be constructive and specific in your feedback."

# Save current settings
echo "Current ANTHROPIC_API_KEY: ${ANTHROPIC_API_KEY:0:10}..."
echo "Current ANTHROPIC_BASE_URL: $ANTHROPIC_BASE_URL"

# Test W&B setup alongside existing configuration
export OPENAI_API_KEY="your-wandb-key"
export OPENAI_BASE_URL="https://api.inference.wandb.ai/v1"

# Test a simple query
claude "Hello, this is a test of W&B Inference"

# If successful, update your permanent configuration