OpenAI Automatic Caching Intermediate

# Request 1:
[System: 1500 tokens] + [Tools: 500 tokens] + [User: 100 tokens]
  → Full processing, prefix cached automatically

# Request 2 (same system + tools, different user message):
[System: 1500 tokens] + [Tools: 500 tokens] + [User: 80 tokens]
  → 2000 tokens served from cache (50% off)
  → 80 tokens processed normally

from openai import OpenAI

client = OpenAI()

# Both requests share the same system prompt prefix
# OpenAI caches it automatically

def ask_question(question):
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {
                "role": "system",
                "content": very_long_system_prompt  # 2000+ tokens
            },
            {
                "role": "user",
                "content": question
            }
        ]
    )
    # Check cache usage in response
    usage = response.usage
    print(f"Cached tokens: {usage.prompt_tokens_details.cached_tokens}")
    return response.choices[0].message.content

# First call: no cache (creates cache entry)
ask_question("What is Python?")

# Second call: cache hit on system prompt
ask_question("Explain decorators")