Skip to content

Quick Start

Follow these steps to start developing your first agent:

1. Clone/Download the Repository

git clone https://github.com/codeligence/codeligence-agents.git
cd codeligence-agents
(If you received a zip archive, just unzip and cd into the folder.)

2. Copy .env.sample to .env

cp .env.sample .env
- Update .env with your credentials and environment variables: - CODELIGENCE_URL=https://.../v1 - CODELIGENCE_TOKEN=ci-... Get a token from the Codeligence Admin interface. - Any other needed variables like WEATHER_API_KEY=xxx, JIRA_API_KEY=yyy, etc.

conda create --name codeligence python=3.12 -y
conda activate codeligence
(Alternatively, you can use python -m venv venv and source venv/bin/activate on most platforms.)

4. Install Requirements

pip install --upgrade pip
pip install -r requirements.txt
This will install FastAPI, uvicorn, Pydantic AI, and other libraries needed by the server and your agents.

5. Run the Dev Script

Inside the src/ folder, there is a rundev.py file for local testing: 

cd src
python rundev.py
- By default, it might reference an example agent (e.g., agents/weather/agent.py) in code. If you don’t have that agent, you can replace the import statement and the call with your own agent’s functions.

6. Create Your First Agent

Inside /src/agents/, create a subdirectory for your new agent, for example weather. 

src/agents/weather/
    agent.py
In agent.py, you need to define:

from typing import Optional, List

from utils.codeligence_utils import AgentInfo, Message, ButtonResponse
from utils.codeligence_utils import codeligence_report_status, codeligence_report_task_output


def run(message: str, history: Optional[List[Message]] = None, button: ButtonResponse = None):
    # 1. Send a status message to the user
    codeligence_report_status("Starting the Weather Agent...")

    # 2. Do whatever logic or AI calls you want
    #    You can talk to an external API, a local DB, or a Pydantic AI model.

    codeligence_report_task_output("Final Result", "The weather is sunny with 20°C.")
    # The final user-visible text is included in the "Final Result" output.

def get_config():
    # This determines if the agent is visible and which name is recognized by the server
    return AgentInfo(enabled=True, name="Weather Agent")
  • codeligence_report_status("..."): A progress update to the chat’s sidebar or log.
  • codeligence_report_task_output("...", "..."): The actual output the user sees. Use taskname to group partial/final outputs.

Note: The agent folder name (weather) becomes the crewName or agent identifier. The server looks for agents.weather.agent.

7. Test Your Agent

Update rundev.py to import and call your weather.agent.run function: 

import asyncio
from dotenv import load_dotenv

load_dotenv(verbose=True)

from agents.weather.agent import run


async def execute():
    await run("What's the weather in Berlin today?")

asyncio.run(execute())
Run again: 
cd src
python rundev.py
You should see logs indicating your weather agent was called. You can insert print statements or logging calls in your agent to debug.