Advanced Features with Gemini

Gemini provides several advanced features for fine-tuning generation, handling streaming responses, and maintaining conversation context.

Generation Configuration

Control the output characteristics using generation configuration parameters.

Typescript

import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);
const model = genAI.getGenerativeModel({ 
  model: "gemini-2.0-flash",
  generationConfig: {
    maxOutputTokens: 100,
    temperature: 0.5,
    stopSequences: ["\n"]
  }
});

const result = await model.generateContent(
  "Explain quantum computing in simple terms."
);
console.log(result.response.text());

Python

from google import genai
from google.genai import types

# Initialize the model with generation config
model = genai.GenerativeModel(
    "gemini-2.0-flash",
    generation_config=types.GenerationConfig(
        max_output_tokens=100,
        temperature=0.5,
        stop_sequences=["\n"]
    )
)

response = model.generate_content(
    "Explain quantum computing in simple terms."
)
print(response.text)

For information about chat conversations, see the Chat Conversations documentation.

Code Execution Tool

Enable the model to execute Python code in a sandbox environment.

Typescript

import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);
const model = genAI.getGenerativeModel({ 
  model: "gemini-2.0-flash",
  tools: ["code_execution"]
});

const result = await model.generateContent(
  "Calculate the sum of the first 40 odd numbers. " +
  "Please write a Python program to do this and return the result."
);
console.log(result.response.text());

Python

from google import genai

# Initialize model with code execution tool
model = genai.GenerativeModel(
    "gemini-2.0-flash",
    tools=["code_execution"]
)

response = model.generate_content(
    "Calculate the sum of the first 40 odd numbers. "
    "Please write a Python program to do this and return the result."
)
print(response.text)

Configuration Options

Generation Config Parameters

• maxOutputTokens: Maximum length of the generated response
• temperature: Controls randomness (0.0 to 1.0)
• stopSequences: Array of strings where generation should stop
• topP: Nucleus sampling parameter
• topK: Top-k sampling parameter
• candidateCount: Number of candidate responses to generate

Best Practices

1. For Generation Config:

   • Use lower temperature (0.2-0.4) for factual responses
   • Use higher temperature (0.7-0.9) for creative tasks
   • Set appropriate maxOutputTokens based on your needs

2. For Code Execution:

   • Validate code before execution
   • Set appropriate timeouts
   • Handle execution errors gracefully

For information about streaming responses, see the Streaming Responses documentation.