LLM API Tutorial

In plain terms

Talking to a model over API is like mailing a transcript of a play to an actor and asking 'write the next line'. The transcript lists who said what - the director's instructions (system), the customer's lines (user), the actor's past lines (assistant). The actor keeps nothing; you mail the whole transcript every time.

Why it matters

Raw models just continue text. The messages format gives conversations structure - who's speaking, what are the standing instructions - and became the universal contract across OpenAI, Anthropic, Google and every open-model server.

How it works

You send JSON: a model name, an optional system instruction, and messages: [{role, content}, ...]. The API returns the assistant's next message plus a usage token count. 'Memory' is an illusion you build: append each reply to your list and resend it all next turn.

When you use it

Every single interaction with a hosted model - chatbots, extraction, agents - flows through this shape.

In code

from anthropic import Anthropic
client = Anthropic()  # reads ANTHROPIC_API_KEY env var
history = []
while True:
    history.append({"role": "user", "content": input("you: ")})
    reply = client.messages.create(
        model="claude-sonnet-5", max_tokens=500,
        system="You are a concise assistant.",
        messages=history)
    text = reply.content[0].text
    history.append({"role": "assistant", "content": text})
    print("bot:", text)

FAQ