active_google_api.yaml

pip install -U google-genai



import os
from google import genai

# Setup your API Key (Set this in your terminal/OS environment first)
# export GEMINI_API_KEY='your_key_here'
api_key = os.getenv("GEMINI_API_KEY")

def generate_ai_response(prompt):
    try:
        # Initialize the client
        client = genai.Client(api_key=api_key)
        
        # Using Gemini 3 Flash (Preview) - typical for 2026 free tier
        model_id = "gemini-3-flash-preview"
        
        response = client.models.generate_content(
            model=model_id,
            contents=prompt
        )
        
        return response.text
    
    except Exception as e:
        return f"An error occurred: {e}"

# Example Usage
if __name__ == "__main__":
    user_prompt = "Explain the concept of Kaizen in software development."
    print("--- AI Response ---")
    print(generate_ai_response(user_prompt))



pip install -U google-genai



import os
import PIL.Image
from google import genai
from pydantic import BaseModel, Field

# 1. Define the Schema for Structured Output
class VideoAnalysis(BaseModel):
    summary: str = Field(description="A brief summary of the visual content")
    detected_objects: list[str] = Field(description="List of key objects visible")
    action_description: str = Field(description="Description of any movement or action")
    is_safe: bool = Field(description="Safety check based on visual content")

# 2. Configure the Advanced Client
client = genai.Client(api_key=os.getenv("GEMINI_API_KEY"))

def analyze_multimodal_content(file_path):
    # Load the media file (Supports PNG, JPEG, WEBP, HEIC, MP4, etc.)
    media_file = PIL.Image.open(file_path)

    try:
        response = client.models.generate_content(
            model="gemini-3-flash-preview", # Current 2026 Free Tier Workhorse
            contents=[
                "Analyze this visual content and provide a structured JSON report.",
                media_file
            ],
            config={
                "system_instruction": "You are a professional vision analyst. Always return valid JSON.",
                "response_mime_type": "application/json",
                "response_schema": VideoAnalysis, # Forces the AI to follow your Pydantic class
                "temperature": 0.2, # Lower temperature for higher accuracy/consistency
            }
        )
        
        # Access the parsed JSON directly
        return response.parsed
    
    except Exception as e:
        print(f"Error: {e}")
        return None

# Example Execution
if __name__ == "__main__":
    # Ensure you have an image/frame named 'input.jpg' in your directory
    # result = analyze_multimodal_content("input.jpg")
    # if result:
    #     print(f"Summary: {result.summary}")
    #     print(f"Objects: {result.detected_objects}")
    print("Code ready. Please ensure GEMINI_API_KEY is set in your environment.")