How to add custom REST endpoints to your Agent

Introduction

uAgents now support custom REST endpoints using the on_rest_get() and on_rest_post() decorators, enabling them to handle HTTP GET and POST requests directly. With this addition, agents can define specific routes, create request and response models, and interact seamlessly with REST clients. This feature enhances the flexibility of agents, allowing them to communicate with external systems using standard web protocols while ensuring that all responses conform to predefined models.

Let's get started!

Prerequisites

Make sure you have read the following resources before going on with this guide:

⚠️

Please note that this feature is only available at the 'agent level', meaning that you cannot add REST endpoints to uAgents protocols.

Usage

To use the new REST API feature in your agent, follow these steps to define custom GET and POST routes, run the agent, and query the endpoints.

Define Custom GET and POST Routes

GET Endpoint

Use the @on_rest_get() decorator to define a custom GET endpoint. You will need to specify the endpoint route and the response model.


@agent.on_rest_get("/custom_get_route", Response)
async def handle_get(ctx: Context) -> Dict[str, Any]:
    return {
        "field": <value>,
    }

POST Endpoint

Use the @on_rest_post() decorator to define a custom POST endpoint. You need to provide both a request model (for input) and a response model (for output).


@agent.on_rest_post("/custom_post_route", Request, Response)
async def handle_post(ctx: Context, req: Request) -> Response:
    ctx.logger.info(req)
    return Response(...)

ℹ️

For querying the agent you have to make sure that:

You use the correct REST method ("GET" or "POST")
You address the agent endpoint together with its route e.g http://localhost:8000/custom_route

Example of custom GET and POST Routes

agent.py
 
import time
from typing import Any, Dict
 
from uagents import Agent, Context, Model
 
class Request(Model):
    text: str
 
class Response(Model):
    timestamp: int
    text: str
    agent_address: str
 
# You can also use empty models to represent empty request/response bodies
class EmptyMessage(Model):
    pass
 
agent = Agent(name="Rest API", seed="your_seed_phrase", port=8000, endpoint=["http://localhost:8000/submit"])
 
@agent.on_rest_get("/rest/get", Response)
async def handle_get(ctx: Context) -> Dict[str, Any]:
    ctx.logger.info("Received GET request")
    return {
        "timestamp": int(time.time()),
        "text": "Hello from the GET handler!",
        "agent_address": ctx.agent.address,
    }
 
@agent.on_rest_post("/rest/post", Request, Response)
async def handle_post(ctx: Context, req: Request) -> Response:
    ctx.logger.info("Received POST request")
    return Response(
        text=f"Received: {req.text}",
        agent_address=ctx.agent.address,
        timestamp=int(time.time()),
    )
 
if __name__ == "__main__":
    agent.run()

Run the example

Run the agent: python agent.py

Query the agent directly through your predefined interfaces:


curl -d '{"text": "test"}' -H "Content-Type: application/json" -X POST http://localhost:8000/rest/post

Expected output

Curl Response:


{"timestamp": 1725610956, "text": "Received: test", "agent_address": "agent1q2qavahvzm2yw237g2cq40pe8p590ppysclaffp2dd0gtk9evtxag7c8djd"}

Agent Logs:


WARNING:  [Rest API]: No endpoints provided. Skipping registration: Agent won't be reachable.
INFO:     [Rest API]: Starting server on http://0.0.0.0:8000 (Press CTRL+C to quit)
INFO:     [Rest API]: Received POST request

Was this page helpful?

You can also leave detailed feedbackon Github

Building a RAG Agent Public and private Agents