Example: Weather Agent

This example builds a fully functional weather agent with a custom Python capability that calls a weather API. By the end, you’ll have an agent package you can test locally and publish to the marketplace.

Project structure

Directoryweather-agent/
- agent.yaml
- agent.md
- Directorycapabilities/
  - Directoryweather/
    manifest.yaml
    prompt.md
    Dockerfile
    requirements.txt
    server.py

Step-by-step

agent.yaml

id: weather-agent
name: Weather Agent
model:
  temperature: 0.5
routing: inline
capabilities:
  - id: weather
install_steps:
  - id: api_key
    type: input
    label: OpenWeatherMap API Key
    description: Get a free key at openweathermap.org/api
    store_as:
      scope: system
      capability_id: weather
      credential_name: OPENWEATHER_API_KEY

agent.md

You are Weather Agent, a helpful assistant that provides accurate,
real-time weather information.

## Instructions
- Use the `weather_lookup` tool to answer any weather question.
- Always ask for a location if the user doesn't provide one.
- Report temperature, conditions, humidity, and wind speed.

## Constraints
- Only answer weather-related questions.
- Never guess weather data — always use the tool.

manifest.yaml

id: weather
class: tool
image: weather-cap:latest
tools:
  - name: weather_lookup
    description: Get current weather conditions for a location.
    input_schema:
      type: object
      properties:
        location:
          type: string
          description: City name or coordinates
        units:
          type: string
          description: "metric or imperial"
      required: [location]
    recommended_policy: allow
network:
  mode: allowlist
  hosts:
    - "api.openweathermap.org:443"
credentials:
  - name: OPENWEATHER_API_KEY
    scope: system
    required: true
    description: API key from openweathermap.org
resources:
  max_memory_mb: 128
  max_cpu_fraction: 0.5
  max_cpu_seconds: 30

server.py — the gRPC capability

import grpc, json, requests
from concurrent import futures
import capability_pb2 as pb2
import capability_pb2_grpc as pb2_grpc

BASE_URL = "https://api.openweathermap.org/data/2.5/weather"

class WeatherService(pb2_grpc.CapabilityServicer):
    def Invoke(self, request, context):
        args = json.loads(request.args_json.decode())
        config = json.loads(request.config_json.decode())
        api_key = config["OPENWEATHER_API_KEY"]

        location = args.get("location", "")
        units = args.get("units", "metric")

        try:
            resp = requests.get(BASE_URL, params={
                "q": location, "units": units,
                "appid": api_key,
            }, timeout=5)
            resp.raise_for_status()
            data = resp.json()

            result = {
                "location": data["name"],
                "temperature": data["main"]["temp"],
                "conditions": data["weather"][0]["description"],
                "humidity": data["main"]["humidity"],
                "wind_speed": data["wind"]["speed"],
            }
            return pb2.InvokeResponse(
                result_json=json.dumps(result).encode()
            )
        except Exception as e:
            return pb2.InvokeResponse(
                result_json=json.dumps({"error": str(e)}).encode()
            )

def serve():
    server = grpc.server(futures.ThreadPoolExecutor(max_workers=4))
    pb2_grpc.add_CapabilityServicer_to_server(
        WeatherService(), server
    )
    server.add_insecure_port("[::]:50051")
    server.start()
    server.wait_for_termination()

if __name__ == "__main__":
    serve()

Dockerfile

FROM python:3.12-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY . .
USER 1000
EXPOSE 50051
CMD ["python", "server.py"]

Test it

Place the weather-agent/ directory inside your Selu agents/ folder and restart Selu.

Open http://localhost:3000, select the Weather Agent, and ask: “What’s the weather in Tokyo?”

Next steps

Ready to publish? Check the Marketplace Guidelines.