Your First Agent¶

You have a project. Now let’s write an actual agent with steps, a task, and an AI call.

Create the agent¶

abi-core add agent greeter \
  --description "Greets users and answers questions" \
  --with-web-interface

Tasks/skills when prompted:

greet_user

Understand the structure¶

agents/greeter/
├── app.py              ← AbiCore instance (import this everywhere)
├── agent_greeter.py    ← Agent class (identity + AI model config)
├── steps.py            ← Your step functions
├── tasks.py            ← Your task functions
├── prompts.py          ← All prompts live here
├── config/config.py    ← AI model, ports, env vars
├── web_interface.py    ← HTTP endpoints
├── main.py             ← Entry point: agent.run()
└── Dockerfile

Write a step¶

A step is a function that does one thing — in this case, ask the AI to generate a greeting. Edit agents/greeter/steps.py:

from app import agent
from config import config
from abi_core.agent.llm_provider import invoke
from prompts import GREET_PROMPT


@agent.step(name="greet")
async def greet(text):
    """Generate a greeting response."""
    result = await invoke(config.LLM_CONFIG, GREET_PROMPT.format(text=text))
    return {"response": result}

Write the prompt¶

Edit agents/greeter/prompts.py:

GREET_PROMPT = """You are a friendly assistant. The user said:

{text}

Respond warmly and helpfully. Keep it short.
"""

Write a task¶

A task runs steps and sends progress updates to the user. Edit agents/greeter/tasks.py:

import json
from app import agent
from abi_core.agent.agent_response import AgentResponse


@agent.task(name="greet_user", task_id="task-greet")
async def greet_user(query):
    """Greet the user."""
    data = json.loads(query) if isinstance(query, str) else query
    text = data.get("text", "")

    yield AgentResponse.status("Thinking...")
    result = await agent.execute_step("greet", text=text)
    yield AgentResponse.result(result)

Run it¶

docker compose up --build -d

Talk to it¶

curl -X POST http://localhost:8002/stream \
  -H "Content-Type: application/json" \
  -d '{"query": "Hey there!"}'

What happened¶

The web interface received your HTTP request
It wrapped your text into a JSON message: {"route": "user_request", "text": "Hey there!"}
The task greet_user ran
It called agent.execute_step("greet") which asked the AI model for a response
The response streamed back to you in real-time

Key rules¶

Steps are simple functions. They receive data, call the AI, return a result.
Tasks run steps and send progress updates to the user.
Prompts go in prompts.py. Never write them inside your functions.
Config goes in config/config.py. Never hardcode addresses or model names.

Next step¶

👉 Simple Chatbot