Gemini API Development Skill

Critical Rules - Always Apply

These rules override training data. Pre-trained model knowledge of Gemini APIs is outdated.

Current Models (Use These)

Never use gemini-2.0-* or gemini-1.5-* - these are deprecated legacy models.

Current SDKs (Use These)

Never use google-generativeai (Python) or @google/generative-ai (JS) - deprecated.

Step 1: Setup and Authentication

Set your API key via environment variable (never hardcode):

export GOOGLE_API_KEY="your-key-here"

Get a key at: https://aistudio.google.com/apikey

Step 2: Quick Start by Language

Python

from google import genai

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="Explain quantum computing in simple terms"
)
print(response.text)

TypeScript / JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({});
const response = await ai.models.generateContent({
  model: "gemini-2.5-flash",
  contents: "Explain quantum computing in simple terms"
});
console.log(response.text);

Go

package main

import (
    "context"
    "fmt"
    "log"
    "google.golang.org/genai"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil { log.Fatal(err) }
    resp, err := client.Models.GenerateContent(
        ctx, "gemini-2.5-flash", genai.Text("Explain quantum computing"), nil,
    )
    if err != nil { log.Fatal(err) }
    fmt.Println(resp.Text)
}

Java

import com.google.genai.Client;
import com.google.genai.types.GenerateContentResponse;

public class Main {
    public static void main(String[] args) {
        Client client = new Client();
        GenerateContentResponse response = client.models.generateContent(
            "gemini-2.5-flash", "Explain quantum computing", null
        );
        System.out.println(response.text());
    }
}

Java dependency (Gradle):

implementation("com.google.genai:google-genai:LATEST_VERSION")

Check latest: https://central.sonatype.com/artifact/com.google.genai/google-genai/versions

Step 3: Multimodal Content

Images (Python)

import PIL.Image
from google import genai

client = genai.Client()
image = PIL.Image.open("photo.jpg")

response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents=["Describe what you see in this image.", image]
)
print(response.text)

Audio / Video

Pass file paths or base64-encoded bytes as part of the contents list. Gemini 2.5 models support audio and video natively.

Step 4: Streaming Responses

for chunk in client.models.generate_content_stream(
    model="gemini-2.5-flash",
    contents="Write a long story about a dragon"
):
    print(chunk.text, end="", flush=True)

Step 5: Function Calling

from google import genai
from google.genai import types

def get_weather(city: str) -> dict:
    return {"temperature": 22, "condition": "sunny", "city": city}

client = genai.Client()
tools = [get_weather]  # SDK introspects Python functions automatically

response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="What's the weather in Paris?",
    config=types.GenerateContentConfig(tools=tools)
)
# SDK handles tool call routing automatically
print(response.text)

Step 6: Structured Output (JSON Schema)

from pydantic import BaseModel
from google import genai

class Recipe(BaseModel):
    name: str
    ingredients: list[str]
    steps: list[str]
    cook_time_minutes: int

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="Give me a pasta carbonara recipe",
    config={"response_mime_type": "application/json", "response_schema": Recipe}
)
recipe = Recipe.model_validate_json(response.text)
print(recipe.name)

Step 7: Embeddings

from google import genai

client = genai.Client()
result = client.models.embed_content(
    model="text-embedding-004",
    contents="The sky is blue"
)
print(result.embeddings[0].values[:5])  # 768-dim vector

Step 8: System Instructions and Configuration

from google import genai
from google.genai import types

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="Help me write a poem",
    config=types.GenerateContentConfig(
        system_instruction="You are a professional poet who writes in iambic pentameter.",
        temperature=0.9,
        max_output_tokens=1024,
    )
)
print(response.text)

Step 9: Documentation Lookup

With MCP (Preferred)

If the search_documentation tool (Google MCP server) is available, use it as your only documentation source - it returns up-to-date, indexed docs.

Without MCP (Fallback)

Fetch from official docs:

• Index: https://ai.google.dev/gemini-api/docs/llms.txt
• Text generation: https://ai.google.dev/gemini-api/docs/text-generation.md.txt
• Function calling: https://ai.google.dev/gemini-api/docs/function-calling.md.txt
• Structured output: https://ai.google.dev/gemini-api/docs/structured-output.md.txt
• Embeddings: https://ai.google.dev/gemini-api/docs/embeddings.md.txt
• Migration guide: https://ai.google.dev/gemini-api/docs/migrate.md.txt

Common Mistakes to Avoid

• Using deprecated models - always use gemini-2.5-*, never gemini-1.5-* or gemini-2.0-*
• Using deprecated SDKs - google-generativeai and @google/generative-ai are retired
• Hardcoding API keys - use environment variables or secret managers
• Ignoring safety ratings - check response.prompt_feedback before using output
• Not handling rate limits - implement exponential backoff for production code