Careeroza

Multi-turn Conversations

The API Is Stateless

The model does not remember earlier calls. To have a conversation, your app keeps a list of messages and sends the whole list every time: user, assistant, user, assistant, ... After each reply, append it to the list.

History in Action

Example

A stand-in for the model replies from simple rules, but the history handling is exactly what real code does. Watch the list grow:

def fake_model(messages):
    """Stand-in for the model: it can only 'remember' what is in messages."""
    last = messages[-1]["content"].lower()
    names = [m["content"].split("is ")[-1] for m in messages
             if m["role"] == "user" and "my name is" in m["content"].lower()]
    if "my name" in last and "?" in last:
        return f"Your name is {names[0]}." if names else "I don't know your name yet."
    return "Nice to meet you!" if names else "Hello! How can I help?"

history = []
def chat(text):
    history.append({"role": "user", "content": text})
    reply = fake_model(history)
    history.append({"role": "assistant", "content": reply})
    print(f"user: {text}\nbot:  {reply}   (messages sent: {len(history) - 1})")

chat("Hi")
chat("My name is Priya")
chat("What is my name?")

print("\nWithout history:", fake_model([{"role": "user", "content": "What is my name?"}]))

Output

user: Hi
bot:  Hello! How can I help?   (messages sent: 1)
user: My name is Priya
bot:  Nice to meet you!   (messages sent: 3)
user: What is my name?
bot:  Your name is Priya.   (messages sent: 5)

Without history: I don't know your name yet.

With the Real API

history = []

def chat(user_text: str) -> str:
    history.append({"role": "user", "content": user_text})
    response = client.messages.create(
        model="claude-sonnet-5-5",
        max_tokens=500,
        system="You are the ShopEase support assistant.",
        messages=history,
    )
    reply = next(b.text for b in response.content if b.type == "text")
    history.append({"role": "assistant", "content": reply})
    return reply

print(chat("I ordered running shoes, size 9."))
print(chat("Can I exchange them for size 10?"))   # the model sees the first message too

Tip: Every turn resends the whole history, so input tokens (and cost) grow with each turn. Long chats need trimming or summarizing (medium level).