Getting Started with Gemini 3.6 Flash API for Text Generation

The Gemini 3.6 Flash model offers fast, cost-effective text generation through Google's API. Here's a practical guide to integrate it into your Python projects.

Prerequisites

• Python 3.8+
• Google Cloud account with Vertex AI enabled
• API key or service account credentials

Installation

pip install google-cloud-aiplatform vertexai

Basic Setup

import vertexai
from vertexai.generative_models import GenerativeModel

# Initialize Vertex AI
vertexai.init(project="your-project-id", location="us-central1")

# Load the Flash model
model = GenerativeModel("gemini-1.5-flash-001")

Text Generation with Parameters

def generate_text(prompt: str, max_tokens: int = 1024, temperature: float = 0.7) -> str:
    """Generate text using Gemini 3.6 Flash with configurable parameters."""

    generation_config = {
        "max_output_tokens": max_tokens,      # Limit response length
        "temperature": temperature,           # Creativity (0.0-1.0)
        "top_p": 0.95,                       # Nucleus sampling
        "top_k": 40,                         # Top-k sampling
    }

    safety_settings = {
        "HARM_CATEGORY_HATE_SPEECH": "BLOCK_MEDIUM_AND_ABOVE",
        "HARM_CATEGORY_DANGEROUS_CONTENT": "BLOCK_MEDIUM_AND_ABOVE",
        "HARM_CATEGORY_SEXUAL_CONTENT": "BLOCK_MEDIUM_AND_ABOVE",
        "HARM_CATEGORY_HARASSMENT": "BLOCK_MEDIUM_AND_ABOVE",
    }

    response = model.generate_content(
        prompt,
        generation_config=generation_config,
        safety_settings=safety_settings
    )

    return response.text

# Example usage
result = generate_text(
    "Explain quantum computing in simple terms",
    max_tokens=500,
    temperature=0.8
)
print(result)

Key Parameters Explained

Parameter	Range	Effect
max_output_tokens	1-8192	Maximum tokens in response
temperature	0.0-1.0	Higher = more creative, lower = more deterministic
top_p	0.0-1.0	Nucleus sampling threshold
top_k	1-40	Top-k sampling limit

Streaming Responses

def generate_stream(prompt: str):
    """Stream tokens as they're generated for real-time UX."""
    response = model.generate_content(
        prompt,
        generation_config={"max_output_tokens": 2048},
        stream=True
    )

    for chunk in response:
        if chunk.text:
            yield chunk.text  # Process incrementally

# Usage
for token in generate_stream("Write a short story about a robot"):
    print(token, end="", flush=True)

Cost Optimization Tips

1. Use Flash over Pro — 10x cheaper for most tasks
2. Set max_output_tokens — Prevent runaway generation
3. Batch requests — Combine multiple prompts when possible
4. Cache responses — Store frequent queries locally

Error Handling

from google.api_core.exceptions import ResourceExhausted, GoogleAPICallError

try:
    result = generate_text(prompt)
except ResourceExhausted:
    print("Rate limit hit — implement exponential backoff")
except GoogleAPICallError as e:
    print(f"API error: {e}")

Conclusion

Gemini 3.6 Flash provides excellent price-performance for text generation. Start with the basic config above, tune parameters for your use case, and implement proper error handling for production workloads.

---

Published via IncomeAutomaton tutorial pipeline