neta/README.md

AI Asistan Kampusu

Sekilli Sukullu Ogrenci Baslangic Rehberi

Local AI + Cloud AI = Daha hizli ogrenme

  ____   ___    __  ___            _     _
 / __ | / _ |  /  |/  /___ ______ (_)___(_)____
/ /_/ |/ __ | / /|_/ / _ `/ __/ // (_-< / _ \
\____/_/ |_|/_/  /_/\_,_/_/  \_,_/___/_/\___/

Important

Gemini 3 preview "indirilen bir program" degil, Google Cloud Vertex AI uzerinden API ile kullanilan bir model ailesidir.

0) Yol Haritasi

flowchart LR
    A[Ollama kur] --> B[Local model calistir]
    B --> C[Google Cloud proje + auth]
    C --> D[Gemini 3 preview API cagir]

1) Ollama Kurulumu (Local AI)

Windows

1. OllamaSetup.exe ile kurulum yap:
   https://docs.ollama.com/windows
2. Terminali yeniden ac.
3. Kontrol et:

ollama -v

Linux

curl -fsSL https://ollama.com/install.sh | sh
ollama -v

Ilk modelini calistir

ollama run gemma3

Ollama API varsayilan olarak burada calisir: http://localhost:11434

Terminalde bunu da yazin (egitim notu)

ollama run gemini-3-flash-preview

Bu komutun amaci farki gostermek: gemini-3-flash-preview bir Ollama modeli degildir, Vertex AI modelidir.

2) Google Cloud Gemini 3 Preview (Vertex AI)

Once gerekli olanlar

• Google Cloud projesi
• Billing acik olmali
• Vertex AI API aktif olmali
• gcloud CLI kurulu olmali

gcloud giris ve kimlik

gcloud init
gcloud auth application-default login

Proje ve API ayari

gcloud config set project YOUR_PROJECT_ID
gcloud services enable aiplatform.googleapis.com

Python SDK kurulumu

pip install --upgrade google-genai

Ortam degiskenleri (PowerShell)

$env:GOOGLE_CLOUD_PROJECT="YOUR_PROJECT_ID"
$env:GOOGLE_CLOUD_LOCATION="global"
$env:GOOGLE_GENAI_USE_VERTEXAI="True"

Ilk Gemini 3 Preview istegi

from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="gemini-3-flash-preview",
    contents="Merhaba! Bana 3 maddede Python'da for dongusunu anlat.",
)

print(response.text)

3) Mini Ogrenci Challenge

1. Terminalde su komutu yaz: ollama run gemini-3-flash-preview
2. Sonra Ollama'da gecerli bir modelle sor: ollama run gemma3
3. Ayni soruyu Gemini 3 preview ile sor.
4. Cevaplari hiz, detay ve dogruluk acisindan karsilastir.

4) Hata Cozme Kisa Notlari

• 403 alirsan: Billing, Vertex AI API ve IAM rol (roles/aiplatform.user) kontrol et.
• 401 alirsan: gcloud auth application-default login komutunu yeniden calistir.
• Model not found alirsan: model ID'yi kontrol et (gemini-3-flash-preview, gemini-3-pro-preview, gemini-3.1-pro-preview).

Kaynaklar (Resmi)

• Ollama Quickstart: https://docs.ollama.com/quickstart
• Ollama Windows: https://docs.ollama.com/windows
• Ollama Linux: https://docs.ollama.com/linux
• Vertex AI Quickstart: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/start/quickstart
• Gemini 3 Baslangic: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/start/get-started-with-gemini-3
• Gemini 3 Pro Model: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro
• Gemini 3 Flash Model: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-flash