aiwand 0.4.9

A simple AI toolkit for text processing using OpenAI and Gemini APIs

AIWand 🪄

The simplest way to unify OpenAI and Gemini APIs - Drop-in replacement for your existing AI code with automatic provider switching and structured output handling.

🎯 Simple Migration - One Line Change

Before - Direct API calls with provider-specific code:

# OpenAI specific code
content = openai_client.chat.completions.create(
    model="gpt-4o",
    messages=messages,
    temperature=0.8,
    top_p=0.9,
    response_format={"type": "json_object"}
)
result = json.loads(content.choices[0].message.content)  # Manual parsing

# OR Gemini specific code
content = gemini_client.chat.completions.create(
    model="gemini-2.0-flash",
    messages=messages,
    temperature=0.8,
    top_p=0.9,
    response_format=SomeSchema
)
result = content.parsed  # Different response handling

After - Unified AIWand code that works with both:

import aiwand

# Same code works with OpenAI, Gemini, and their structured outputs!
content = aiwand.make_ai_request(
    model="gpt-4o",          # or "gemini-2.0-flash" 
    messages=messages,
    temperature=0.8,
    top_p=0.9,
    response_format=CarouselContent  # Pydantic model - automatic parsing!
)
# 'content' is already your parsed Pydantic object - no post-processing needed! ✨

✨ Why AIWand?

• 🔄 Drop-in Replacement - Minimal code changes, maximum benefits
• 🧠 Smart Provider Detection - Automatically uses OpenAI or Gemini based on model name
• 🏗️ Structured Output Magic - Handles Pydantic models automatically for both providers
• ⚡ No Post-Processing - Get parsed objects directly, skip manual JSON handling
• 🎯 Unified API - Same code works across different AI providers
• 🔑 Zero Configuration - Works with just environment variables
• 📱 High-Level Functions - Built-in summarization, chat, text generation, and classification

🚀 Quick Start

Installation

pip install aiwand

Configuration

Set your API keys as environment variables:

# Option 1: OpenAI only
export OPENAI_API_KEY="your-openai-key"

# Option 2: Gemini only  
export GEMINI_API_KEY="your-gemini-key"

# Option 3: Both (set preference)
export OPENAI_API_KEY="your-openai-key"
export GEMINI_API_KEY="your-gemini-key"
export AI_DEFAULT_PROVIDER="openai"  # or "gemini"

Or create a .env file in your project:

OPENAI_API_KEY=your-openai-key
GEMINI_API_KEY=your-gemini-key
AI_DEFAULT_PROVIDER=openai

Core AI Functionality

The make_ai_request() function is the heart of AIWand - a unified interface for all AI providers:

import aiwand
from pydantic import BaseModel

# Basic text generation
response = aiwand.make_ai_request(
    messages=[{"role": "user", "content": "Explain quantum computing"}],
    model="gpt-4o"  # Automatically uses OpenAI
)

# Switch providers seamlessly
response = aiwand.make_ai_request(
    messages=[{"role": "user", "content": "Explain quantum computing"}],
    model="gemini-2.0-flash"  # Automatically uses Gemini
)

# Structured output with Pydantic models
class BlogPost(BaseModel):
    title: str
    content: str
    tags: list[str]

blog_post = aiwand.make_ai_request(
    messages=[{"role": "user", "content": "Write a blog post about AI"}],
    model="gpt-4o",
    response_format=BlogPost  # Returns parsed BlogPost object!
)
print(blog_post.title)  # Direct access to structured data

# Custom/preview models with explicit provider
response = aiwand.make_ai_request(
    model="gemini-2.5-flash-preview-05-20",  # New model not in our registry
    provider="gemini",  # Explicit provider specification
    messages=[{"role": "user", "content": "Hello from the future!"}]
)

# Advanced parameters
response = aiwand.make_ai_request(
    messages=[
        {"role": "system", "content": "You are a helpful coding assistant"},
        {"role": "user", "content": "Write a Python function to sort a list"}
    ],
    model="gpt-4o",
    temperature=0.3,  # More focused
    max_tokens=500,
    top_p=0.9
)

High-Level Convenience Functions

For common tasks, use these simplified functions:

import aiwand

# Text summarization
summary = aiwand.summarize("Your long text here...")

# AI chat with conversation history
response = aiwand.chat("What is machine learning?")

# Text generation from prompts
story = aiwand.generate_text("Write a poem about coding")

# Customized summarization
summary = aiwand.summarize(
    text="Your long text...",
    style="bullet-points",  # "concise", "detailed", "bullet-points"
    max_length=50,
    model="gpt-4o"  # Optional: specify model
)

# Chat with conversation history
conversation = []
response1 = aiwand.chat("Hello!", conversation_history=conversation)
conversation.append({"role": "user", "content": "Hello!"})
conversation.append({"role": "assistant", "content": response1})

response2 = aiwand.chat("What did I just say?", conversation_history=conversation)

# Text generation with custom parameters
text = aiwand.generate_text(
    prompt="Write a technical explanation",
    max_tokens=300,
    temperature=0.3  # Lower = more focused, Higher = more creative
)

# Structured data extraction from content and links
contact_info = aiwand.extract(content="John Doe, email: [e-mail]")

# Extract from URLs and files
data = aiwand.extract(links=["https://example.com/article", "/path/to/file.txt"])

# Combine content and links with structured output
from pydantic import BaseModel

class ContactInfo(BaseModel):
    name: str
    email: str
    phone: str

result = aiwand.extract(
    content="Meeting notes about John...",
    links=["https://company.com/about"],
    response_format=ContactInfo,  # Get structured Pydantic object
    temperature=0.1  # Lower for more consistent extraction
)

# Text classification and grading
grader = aiwand.create_binary_classifier(criteria="correctness")
result = grader(question="What is 2+2?", answer="4", expected="4")
print(f"Score: {result.score}, Choice: {result.choice}")

# Custom classifier with multiple grades
math_grader = aiwand.create_classifier(
    prompt_template="Grade this math answer: {question} -> {answer}",
    choice_scores={"CORRECT": 1.0, "PARTIAL": 0.5, "WRONG": 0.0}
)
result = math_grader(question="What is 5+3?", answer="8", expected="8")

# Helper utilities for testing and development
random_num = aiwand.generate_random_number(8)  # 8-digit number
unique_id = aiwand.generate_uuid()  # UUID4

🎯 Smart Provider Features

Automatic Model Detection

# AIWand automatically detects the right provider:
response = aiwand.make_ai_request(model="gpt-4o", ...)        # → OpenAI
response = aiwand.make_ai_request(model="gemini-2.0-flash", ...)  # → Gemini
response = aiwand.make_ai_request(model="o3-mini", ...)       # → OpenAI

# Pattern-based detection for unknown models:
response = aiwand.make_ai_request(model="gemini-experimental-123", ...)  # → Gemini

Explicit Provider Control

# Force a specific provider for custom models:
response = aiwand.make_ai_request(
    model="my-custom-model",
    provider="gemini",  # or AIProvider.GEMINI
    messages=[...]
)

# Works with both string and enum:
from aiwand import AIProvider
response = aiwand.make_ai_request(
    model="any-model",
    provider=AIProvider.OPENAI,
    messages=[...]
)

Structured Output Support

from pydantic import BaseModel

class ProductReview(BaseModel):
    rating: int
    pros: list[str]
    cons: list[str]
    recommendation: bool

# Works identically with both providers:
review = aiwand.make_ai_request(
    model="gpt-4o",  # or "gemini-2.0-flash"
    messages=[{"role": "user", "content": "Review this product: ..."}],
    response_format=ProductReview
)
# No manual JSON parsing needed - returns ProductReview object directly!

Configuration Management

import aiwand

# Show current configuration
aiwand.show_current_config()

# Interactive setup (optional)
aiwand.setup_user_preferences()

Error Handling

import aiwand

try:
    summary = aiwand.summarize("Some text")
except aiwand.AIError as e:
    print(f"AI service error: {e}")
except ValueError as e:
    print(f"Input error: {e}")

🔧 CLI Usage (Optional)

# Direct prompts (easiest way!)
aiwand "Ten fun names for a pet pelican"
aiwand "Explain quantum computing in simple terms" 

# Specific commands
aiwand summarize "Your text here" --style bullet-points
aiwand chat "What is machine learning?"
aiwand generate "Write a story about AI"

# Helper utilities
aiwand helper random --length 8        # Generate 8-digit random number
aiwand helper uuid --uppercase         # Generate uppercase UUID

# Setup preferences
aiwand setup
aiwand config

📚 Documentation

• API Reference - Complete function documentation
• CLI Reference - Command line usage
• Installation Guide - Detailed setup instructions
• Virtual Environment Guide - Best practices for Python environments

🛠️ Contributing

We welcome contributions from both AI assistants and human developers! Please see our comprehensive contributing guide:

• Contributing Guide - Standards, workflows, and best practices
• Development Guide - Technical details and advanced topics

Whether you're an AI assistant helping users or a human developer, these guides ensure consistency and quality across all contributions.

🤝 Connect

• GitHub: github.com/onlyoneaman/aiwand
• PyPI: pypi.org/project/aiwand
• X (Twitter): @onlyoneaman

📝 License

MIT License - see LICENSE file for details.

---

Made with ❤️ by Aman Kumar