Tutorial: Running inference with Qwen3 235B A22B Thinking-2507 using W&B Inference

pip install openai wandb weave

# Set environment variables in your notebook
%env WANDB_API_KEY=your-wandb-api-key-here

export WANDB_API_KEY="your-wandb-api-key-here"

import os
# Set environment variables programmatically
os.environ["WANDB_API_KEY] = "your-wandb-api-key-here"

import os
import openai
import weave

PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

resp = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are a science educator who explains complex topics clearly."},
        {"role": "user", "content": "Explain quantum computing fundamentals, advantages, applications, and limitations. Use simple analogies for someone without a physics background."}
    ],
    temperature=0.7,
    max_tokens=1000,
)

print(resp.choices[0].message.content)


import os
import openai
import weave

PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }    
)

resp = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are a creative writing assistant specializing in science fiction."},
        {"role": "user", "content": "Write a compelling 800-word time travel story with well-developed characters, plot twists, and philosophical themes about causality and free will."}
    ],
    temperature=0.8,
    top_p=0.9,
    max_tokens=2000,
)
print(resp.choices[0].message.content)


import os
import sys
import openai
import weave


PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

stream = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are a space exploration historian and science communicator."},
        {"role": "user", "content": "Create a comprehensive narrative about humanity's space exploration journey from early dreams to current interplanetary plans. Include key moments, discoveries, heroes, innovations, and philosophical implications."}
    ],
    stream=True,
    temperature=0.7,
)

sys.stdout.write("Response: ")
for chunk in stream:
    delta = chunk.choices[0].delta
    if delta and delta.content:
        sys.stdout.write(delta.content)
        sys.stdout.flush()
print()


import io
import requests
import openai
import weave
from pypdf import PdfReader

PROJECT = "wandb_inference"
weave.init(PROJECT)

PDF_URL = "https://docs.aws.amazon.com/pdfs/bedrock-agentcore/latest/devguide/bedrock-agentcore-dg.pdf"
QUESTION = "How does AgentCore's memory architecture work?"

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

r = requests.get(PDF_URL, timeout=60)
r.raise_for_status()

reader = PdfReader(io.BytesIO(r.content))
pages = reader.pages[:100]
text = "\n\n".join(page.extract_text() or "" for page in pages)

doc_snippet = text

prompt = (
    f"Based on this AWS Bedrock AgentCore documentation, answer: {QUESTION}\n\n"
    f"Documentation:\n{doc_snippet}\n\n"
    "Provide analysis with direct quotes."
)

resp = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are an AWS Bedrock AgentCore expert. Analyze the documentation and answer questions based only on the provided text."},
        {"role": "user", "content": prompt}
    ],
    temperature=0.2,
    max_tokens=1500,
)

print(resp.choices[0].message.content)


import os
import openai
import weave

PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

code_snippet = """
// JavaScript function with Chinese identifiers and mixed language comments
function 计算总价(商品列表, 折扣率) {
    let 总价 = 0;  // Total price accumulator
    for (const 商品 of 商品列表) {
        总价 += 商品.价格 * 商品.数量  // Calculate item total: price * quantity
    }
    const 折扣金额 = 总价 * 折扣率  // Calculate discount amount
    return 总价 - 折扣金额  // Return final price after discount
}

# Python function with Chinese docstring and mixed naming conventions
def validate_用户输入(user_data):
    '''
    验证用户输入数据的完整性和有效性
    Validates user input data for completeness and validity
    
    This function checks if all required fields are present in the user data
    and raises appropriate errors if validation fails.
    '''
    required_fields = ['name', 'email', '年龄']  # Mix of English and Chinese field names
    for field in required_fields:
        if field not in user_data:
            raise ValueError(f"Missing required field: {field}")
    return True
"""

task = (
    "Analyze this code: "
    "1) Explain each function in English and Chinese "
    "2) Assess code quality issues "
    "3) Recommend improvements "
    "4) Provide refactored version with consistent naming"
)

resp = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are a software architect with expertise in English and Chinese."},
        {"role": "user", "content": f"{task}\n\nCode:\n{code_snippet}"}
    ],
    temperature=0.2,
    max_tokens=1200,
)

print(resp.choices[0].message.content)


import openai
import weave
import os


PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

resp = client.chat.completions.create(
    model="Qwen/Qwen3-235B-A22B-Thinking-2507",
    messages=[
        {"role": "system", "content": "You are a SaaS business strategist with expertise in growth optimization and churn reduction."},
        {"role": "user", "content": """
        Analyze our SaaS metrics and provide a strategic plan:
        
        Metrics:
        - Monthly churn: 15%
        - CAC: $150
        - CLV: $800  
        - MRR per customer: $50
        - Active customers: 2,000
        - Monthly marketing spend: $60,000
        
        Provide: 1) Metric assessment vs benchmarks, 2) Critical issues, 3) Improvement strategy, 4) Implementation timeline, 5) Success metrics.
        """}
    ],
    temperature=0.7,
    max_tokens=1000,
)

print(resp.choices[0].message.content)


import os
import openai
import weave

PROJECT = "wandb_inference"
weave.init(PROJECT)

client = openai.OpenAI(
    base_url="https://api.inference.wandb.ai/v1",
    api_key=os.getenv("WANDB_API_KEY"),
    project=PROJECT,
    default_headers={
        "OpenAI-Project": "wandb_fc/quickstart_playground"  # replace with your actual team/project
    }
)

@weave.op()
def analyze_customer_feedback(feedback_text, sentiment_threshold=0.5):
    """
    Analyze comprehensive customer feedback and return detailed sentiment analysis with actionable insights.
    This function performs deep sentiment analysis while tracking execution via weave.op decorator
    since the OpenAI client doesn't have built-in weave integration.
    """
    resp = client.chat.completions.create(
        model="Qwen/Qwen3-235B-A22B-Thinking-2507",
        messages=[
            {"role": "system", "content": "You are a customer experience analyst. Provide sentiment scores (-1 to +1), identify themes, and suggest improvements."},
            {"role": "user", "content": f"Analyze this customer feedback: {feedback_text}\n\nThreshold: {sentiment_threshold}\n\nProvide: 1) Sentiment score, 2) Key themes, 3) Specific concerns, 4) Recommended actions."}
        ],
        temperature=0.1,
        max_tokens=500,
    )
    return resp.choices[0].message.content

if __name__ == "__main__":
    out = analyze_customer_feedback(
        "I've used this software for three months. Core functionality works, but recent interface changes complicate my workflow. Features are buried in submenus, layout feels counterintuitive, and it loads slower since the update.",
        sentiment_threshold=0.3,
    )
    print(out)