This guide walk you through building an AI agent with function calling capability with ACI.dev

​
Sign up to the Platform

  • We are currently in a closed beta, so you’ll need a signup code to register and log in to our platform .
  • We’ll create a default project for you when you first log in to the platform.

​
Configure Your First App for Your AI Agent to Use

1

Configure Github in App Store

Navigate to App Store and find GITHUB . Then click on Configure App button to set up the app for your project.

2

Link your Github account

  • Navigate to App Configurations and find GITHUB App Configuration . Then click on Add Account button to link your Github account.

linked account owner id is the ID of the owner of the linked account. It’s up to you to decide which ID to use—it can be the unique end-user ID from your system. Or If you’re building an agent for yourself, you can choose any name you prefer. Later, you’ll need to provide this linked account owner id in your code to execute functions on behalf of the user.

  • Click Start OAuth2 Flow button to start the OAuth2 flow and link your Github account under the project.
3

Allow Your Agents to Access GITHUB

  • We took an opinionated approach to acommodate a multi-agent architecture. You can create multiple agents within the project, and each agent has its own API key and apps that they are allowed to use.
  • We created a default agent for you when you first log in to the platform.

Navigate to the project settings page: Manage Project . Then click the Edit button under ALLOWED APPS column of the agent to allow access to the GITHUB app.

4

Grab the API key

Each Agent is assigned an API key, which you can use to send requests to our platform—either via raw HTTP requests or our SDK. Later, you’ll need to include this API key in your code.

​
Code Example

The Aipolabs ACI Python SDK requires Python 3.10 or later. The example below uses poetry for package installation, but you can use pip or any other package manager of your choice.

The full example code is available at the end of this guide.

1

Install Aipolabs ACI Python SDK

  • Install the Aipolabs ACI SDK
bash
  poetry add aipolabs
  • To run the example, you’ll also need to install other required packages.
bash
  poetry add openai python-dotenv
2

Provide the API key to the SDK

You’ll need both the AIPOLABS API key and the OpenAI API key to run the example in this section. Create a .env file in the root of your project and add the following:

.env
  AIPOLABS_ACI_API_KEY=your_aipolabs_api_key
  OPENAI_API_KEY=your_openai_api_key
3

Create Aipolabs ACI Client and OpenAI Client

python 3.10+
  from dotenv import load_dotenv
  from aipolabs import ACI
  from openai import OpenAI

  load_dotenv()

  aci = ACI()
  openai = OpenAI()
4

Get the Function Definition

python 3.10+
  github_star_repo_function = aci.functions.get_definition("GITHUB__STAR_REPOSITORY")
5

Append the Function Definition to the LLM Request

python 3.10+
  response = openai.chat.completions.create(
    model="gpt-4o-2024-08-06",
    messages=[
        {
            "role": "system",
            "content": "You are a helpful assistant with access to a variety of tools.",
        },
        {
            "role": "user",
            "content": "Star the aipotheosis-labs/aipolabs-python githubrepository for me",
        },
    ],
    tools=[github_star_repo_function],
    tool_choice="required",  # force the model to generate a tool call for demo purposes
  )
6

Handle the Tool Call Response and Execute the Function via ACI.dev

Replace <LINKED_ACCOUNT_OWNER_ID> with the linked account owner id you used when linking your Github account.

python 3.10+
  tool_call = (
      response.choices[0].message.tool_calls[0]
      if response.choices[0].message.tool_calls
      else None
  )

  if tool_call:
      result = aci.handle_function_call(
          tool_call.function.name,
          json.loads(tool_call.function.arguments),
          linked_account_owner_id=<LINKED_ACCOUNT_OWNER_ID>
      )
7

Full Runnable Code

Here is the complete runnable code for the example above, you can copy and paste it into a file and run it.

  • Remember to provide api keys in the .env file.
  • Replace <LINKED_ACCOUNT_OWNER_ID> with the linked account owner id you used when linking your Github account.
python 3.10+
  import json
  from dotenv import load_dotenv
  from openai import OpenAI
  from aipolabs import ACI
  load_dotenv()

  openai = OpenAI()
  aci = ACI()

  def main() -> None:
      # For a list of all supported apps and functions, please go to the platform.aci.dev
      print("Getting function definition for GITHUB__STAR_REPOSITORY")
      github_star_repo_function = aci.functions.get_definition("GITHUB__STAR_REPOSITORY")

      print("Sending request to OpenAI")
      response = openai.chat.completions.create(
          model="gpt-4o-2024-08-06",
          messages=[
              {
                  "role": "system",
                  "content": "You are a helpful assistant with access to a variety of tools.",
              },
              {
                  "role": "user",
                  "content": "Star the aipotheosis-labs/aipolabs-python github repository for me",
              },
          ],
          tools=[github_star_repo_function],
          tool_choice="required",  # force the model to generate a tool call for demo purposes
      )
      tool_call = (
          response.choices[0].message.tool_calls[0]
          if response.choices[0].message.tool_calls
          else None
      )

      if tool_call:
          print("Handling function call")
          result = aci.handle_function_call(
              tool_call.function.name,
              json.loads(tool_call.function.arguments),
              linked_account_owner_id=<LINKED_ACCOUNT_OWNER_ID>,
          )
          print(result)


  if __name__ == "__main__":
      main()

​
Advanced Usage

For more advanced usage, please refer to the examples in our Python SDK Repository .