gemini

Gemini

Gemini is Google's native multimodal foundation model family, featuring native audio, image, and video comprehension alongside million-token context windows and fast tool-calling.

When to Use

• Massive Multimodal Understanding: Gemini 1.5 Pro / Flash processing up to 2 million tokens of text, hours of video, audio, and large codebases.
• Structured JSON Extraction: Enforcing rigid response schemas with Pydantic and native response_schema.
• High-Speed Real-Time Inference: Gemini 1.5 Flash for sub-second, cost-effective conversational assistants and summarization.
• Code & Repository Analysis: Analyzing entire GitHub repositories in a single context window.

Quick Start

from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="gemini-2.0-flash",
    contents="Explain quantum computing in one short paragraph.",
)

print(response.text)

Core Concepts

Structured Output with Pydantic & Google GenAI SDK

Enforcing type-safe JSON schema responses:

from google import genai
from google.genai import types
from pydantic import BaseModel, Field
import os

client = genai.Client(api_key=os.environ.get("GEMINI_API_KEY"))

class InvoiceExtraction(BaseModel):
    invoice_number: str
    vendor: str
    total_amount: float = Field(description="Total invoice amount in USD")
    items: list[str]
    is_paid: bool

response = client.models.generate_content(
    model='gemini-1.5-flash',
    contents='Invoice #INV-2026-908 from CloudCorp for $1,450.00. Included 2 Cloud Servers and 1 Storage bucket. Paid via Wire.',
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema=InvoiceExtraction,
        temperature=0.1
    ),
)

extracted: InvoiceExtraction = InvoiceExtraction.model_validate_json(response.text)
print(f"Extracted Invoice: {extracted.invoice_number}, Total: ${extracted.total_amount}")

Native Video & Audio Multimodal Analysis