Minimal Example

The shortest complete AmritaCore program. Copy, paste, run.

What you will see: the agent's reply streaming out token by token. The three lines before the loop (minimal_init, create_agent, get_chatobject) are the entire setup — everything else is AmritaCore doing the work.

import asyncio
import os

from amrita_core import create_agent, minimal_init


async def main() -> None:
    await minimal_init()
    agent = create_agent(
        base_url="https://api.openai.com/v1",
        api_key=os.environ["OPENAI_API_KEY"],
        model="gpt-4o-mini",
    )
    chat = agent.get_chatobject("Hello! Who are you?")
    async with chat.begin():
        async for msg in chat.io_stream.get_response_generator():
            print(msg, end="", flush=True)


if __name__ == "__main__":
    asyncio.run(main())

What Just Happened

Line	What it does
minimal_init()	Initializes the global config (required once per process)
create_agent(...)	Builds an Agent factory with an LLM adapter bound to your endpoint
agent.get_chatobject(text)	Creates a ChatObject — the basic unit of a dialogue
chat.begin()	Runs the workflow; the agent answers inside this context
get_response_generator()	Streams the response token by token

Notes

• api_key can be omitted if you use OPENAI_API_KEY/ANTHROPIC_API_KEY env vars with the matching base_url.
• For DeepSeek or other OpenAI-compatible providers, just change base_url and model.
• Anthropic? Use protocol="anthropic" — see Adapters.

Next

Basic Example — add streaming metadata, tools and sessions.