Agents and Functions Creation using APIs

Introduction

This guide shows how to create Agents and Agent Functions in Agentverse using APIs. With this guide, we set up a Python script that interacts with the Agentverse and help us creating Agents and Agent Functions. Before you begin, it is necessary that you satisfy the following requirements:

Python version greater than 3.9 and less than 3.11.
The requests library is installed. You can install it using: pip install requests.
Agentverse (opens in a new tab) credentials.

How to get Agentverse API Tokens

Log in to your Agentverse (opens in a new tab) account and go to the Profile section.
Click on button + New API Key.
Give name to your API key.
Click on write for Access to all resources in Agentverse and click on Generate API Key

How to create Agents and respective Functions

Open terminal and create a directory agents using: mkdir agents.
Create two Python files, agent.py and agent_create.py, in this directory and include the following sample scripts within them. You can do this using:

windows
echo. > agent.py

windows
echo. > agent_create.py

Fill in the scripts with the code presented here below for each one of them:

Script 1: `agent.py`

agent.py
 
import requests
import json
from ai_engine import UAgentResponse, UAgentResponseType
 
 
class Coordinates(Model):
    location: str
 
 
location_protocol = Protocol("Location Coordinates")
 
 
async def location_coordinates(latitude, longitude):
    url = "https://geocoding-by-api-ninjas.p.rapidapi.com/v1/reversegeocoding"
    querystring = {"lat": latitude, "lon": longitude}
 
    headers = {
        "X-RapidAPI-Key": "YOUR_API_KEY",
        "X-RapidAPI-Host": "geocoding-by-api-ninjas.p.rapidapi.com"
    }
 
    response = requests.get(url, headers=headers, params=querystring)
 
    data = response.json()[0]['name']
 
    return data
 
 
@location_protocol.on_message(model=Coordinates, replies=UAgentResponse)
async def location_coordinates_calculator(ctx: Context, sender: str, msg: Coordinates):
    ctx.logger.info(msg.location)
    latitude, longitude = map(str.strip, msg.location.split(','))
    message = location_coordinates(latitude, longitude)
    ctx.logger.info(f"Location found: {message}")
    await ctx.send(sender, UAgentResponse(message=message, type=UAgentResponseType.FINAL))
 
 
agent.include(location_protocol)

Script 2: `agent_create.py`

This script interacts with the Agentverse API to achieve the Agent and Function creation tasks.

Let's get started!

Import the required libraries and set up the authorization token:


# Importing Required libraries
import time
import requests
 
# Define access token
token = 'Bearer <Your_access_token>'

Take the agent name from user and store the agent's address:


# Take name of agent from user
name = input('Please give name of your agent? ')
 
# Create payload for agent creation request
agent_creation_data = {
    "name": name
}
 
# Post request to create an agent and store address
response_agent = requests.post("https://agentverse.ai/v1/hosting/agents", json=agent_creation_data, headers={
    "Authorization": token
}).json()
address = response_agent['address']
print(f'Agent Address : {address}')

Take code from agent.py file. Then, store it in a dedicated script for the created agent:


# Reading code to be placed in agent
with open('agent.py', 'r') as file:
    code = file.read()
agent_code_data = {
    "code": code
}
 
    # Creating agent.py script for created agent
    response_code_update = requests.put(f"https://agentverse.ai/v1/hosting/agents/{address}/code", json=agent_code_data, headers={
        "Authorization": token
    })
 
# Starting the agent
requests.post(f"https://agentverse.ai/v1/hosting/agents/{address}/start", headers={
    "Authorization": token
})
 
time.sleep(10) # waiting before getting agent's protocol

Requesting protocol digest for the created Agent:


# Request to get agent protocol digest
response_protcol = requests.get(f"https://agentverse.ai/v1/almanac/agents/{address}", headers={
    "Authorization": token
})
protocol_digest = response_protcol.json()['protocols'][1]
print(f'Protocol Digest : {protocol_digest}')
 
time.sleep(10) # Waiting before getting model_digest

Request Model digest and name using Almanac APIs:


# Request to get agent's model details
response_model = requests.get(f"https://agentverse.ai/v1/almanac/manifests/protocols/{protocol_digest}", headers={
    "Authorization": token
})
model = response_model.json()['models']
 
time.sleep(10) # Waiting before storing details to create functions

Now, save all the required details to create the Agent Function. Then, create the Agent Function on the basis of the details received:


# Taking inputs from user for details required to create a function
name_function = input("Please give function name: ")
description = input("Please enter function description: ")
field_name = input("Please enter field name: ")
field_description = input("Please enter field description: ")
tasktype = input("Please tell primary or secondary function: ").upper()
 
# Logging details provided by user
print(
    f"Function name: {name_function} \nFunction Description: {description} \nField Name: {field_name}\nField Description: {field_description}\nTask Type: {tasktype}"
)
 
# Storing model digest and name to be used for function creation
model_digest = response_model.json()["interactions"][0]["request"].replace("model:", "")
print(f"Model Digest : {model_digest}")
model_name = model[0]["schema"]["title"]
print(f"Model Name : {model_name}")
 
# Creating payload for function creation
data = {
    "agent": address,
    "name": name_function,
    "description": description,
    "protocolDigest": protocol_digest,
    "modelDigest": model_digest,
    "modelName": model_name,
    "arguments": [
        {
            "name": field_name,
            "required": True,
            "type": "string",
            "description": field_description,
        }
    ],
    "type": tasktype,
}
 
# Requesting AI Engine function API to create a function with created payload and storing the response.
response_function = requests.post(
    "https://agentverse.ai/v1beta1/functions/",
    json=data,
    headers={"Authorization": token},
)
# Storing name of function and printing it to check if function was created successfully
name = response_function.json()["name"]
print(f"Function Created with name: {name}")

Complete Script

agent_create.py
 
# Importing Required libraries
import time
import requests
 
# Decode the refresh token
token = f'Bearer <Your_access_token>'
 
# Take name of agent from user
name = input('Please give name of your agent? ')
# Create payload for agent creation request
agent_creation_data = {
    "name": name
}
# Post request to create an agent and store address
response_agent = requests.post("https://agentverse.ai/v1/hosting/agents", json=agent_creation_data,
                               headers={"Authorization": token}).json()
 
address = response_agent['address']
print(f'Agent Address : {address}')
 
# Reading code to be placed in agent
with open('agent.py', 'r') as file:
    code = file.read()
agent_code_data = {
    "code": code
}
 
# Creating simple_function.py script for created agent
response_code_update = requests.put(f"https://agentverse.ai/v1/hosting/agents/{address}/code", json=agent_code_data,
                                    headers={"Authorization": token})
 
# Starting the agent
requests.post(f"https://agentverse.ai/v1/hosting/agents/{address}/start", headers={"Authorization": token})
time.sleep(10)  # waiting before getting agent's protocol
 
# Request to get agent protocol digest
response_protcol = requests.get(f"https://agentverse.ai/v1/almanac/agents/{address}", headers={"Authorization": token})
protocol_digest = response_protcol.json()['protocols'][1]
print(f'Protocol Digest : {protocol_digest}')
time.sleep(10)  # Waiting before getting model_digest
 
# Request to get agent's model details
response_model = requests.get(f"https://agentverse.ai/v1/almanac/manifests/protocols/{protocol_digest}",
                              headers={"Authorization": token})
model = response_model.json()['models']
time.sleep(10)  # Waiting before storing details to create functions
 
# Taking inputs from user for details required to create a function
name_function = input("Please give function name: ")
description = input("Please enter function description: ")
field_name = input("Please enter field name: ")
field_description = input("Please enter field description: ")
tasktype = input("Please tell primary or secondary function: ").upper()
 
# Logging details provided by user
print(
    f"Function name: {name_function} \nFunction Description: {description} \nField Name: {field_name}\nField Description: {field_description}\nTask Type: {tasktype}"
)
 
# Storing model digest and name to be used for function creation
model_digest = response_model.json()["interactions"][0]["request"].replace("model:", "")
print(f"Model Digest : {model_digest}")
model_name = model[0]["schema"]["title"]
print(f"Model Name : {model_name}")
 
# Creating payload for function creation
data = {
    "agent": address,
    "name": name_function,
    "description": description,
    "protocolDigest": protocol_digest,
    "modelDigest": model_digest,
    "modelName": model_name,
    "arguments": [
        {
            "name": field_name,
            "required": True,
            "type": "string",
            "description": field_description,
        }
    ],
    "type": tasktype,
}
 
# Requesting AI Engine function API to create a function with created payload and storing the response.
response_function = requests.post(
    "https://agentverse.ai/v1beta1/functions/",
    json=data,
    headers={"Authorization": token},
)
# Storing name of function and printing it to check if function was created successfully
name = response_function.json()["name"]
print(f"Function Created with name: {name}")

Steps to run the script

Open terminal and go to directory agents created above.
Make sure agent.py and agent_create.py are in this directory.
Head over to the Agentverse (opens in a new tab) and generate API keys .
Open script in editor and replace token field.
Run command python agent_create.py and enter the required details.
Provide Agent and Function details as asked, then, check agent and function on Agentverse.

Expected Output

Provide all details asked in the script:


abc@xyz-MacBook-Pro agents % python3 agents_create.py
Please give name of your agent? my first API agent
Agent Address : agent1q06l8hekn859e5rtwufewmyhwghe6j9y00g0wc8u7gcx05cjfk98jyf6lte
Protocol Digest : c7a6f160fd8d8b7cb357dad9b5be420510ce466dbb67051c07caf2b860216b01
Please give function name: location finder
Please enter function description: this function helps to find nearest city using coordinates
Please enter field name: location
Please enter field description: this describes the coordinates of the location in string format longitude latitude
Please tell primary or secondary function: primary
Function name: location finder
Function Description: this function helps to find nearest city using coordinates
Field Name: location
Field Description: this describes the coordinates of the location in string format longitude latitude
Task Type: PRIMARY
Model Digest : 10a2f843c4c92955688d5e7f22fabe79623869eabfcd97d97da83527b436d3e2
Model Name : Coordinates
Function Created with name: location finder

Agent created on Agentverse:

Function created on Agentverse:

Was this page helpful?

You can also leave detailed feedbackon Github

CI with AVCTL and Github Actions Secret Management APIs