Python quickstart

This guide will walk you through creating and deploying your first AI agent using the aiXplain Python SDK. You'll learn how to specify a Large Language Model (LLM), equip the agent with tools, and integrate the agent into your application.

Create and export an API Key

Create an API key on the Integrations page on Studio. Once generated,

1. export it as an environment variable in your terminal or
2. set it directly in your Python project - you can use the os module or the python-dotenv package & .env file.

MacOS / Linux

export TEAM_API_KEY="your_api_key_here"

Windows

setx TEAM_API_KEY="your_api_key_here"

os module

import os
os.environ["TEAM_API_KEY"] = "your_api_key_here"

python-dotenv package & .env file

from dotenv import find_dotenv, load_dotenv # pip install python-dotenv
load_dotenv(find_dotenv()) # Load environment variables from .env file

Install the aiXplain SDK

To get started, install the aiXplain package using pip:

pip install aixplain

1. Build and deploy an Agent

Agent

With the the SDK installed, copy the code below into a python file (example.py) or jupyter notebook (example.ipynb) code block. After a few moments, you should see the output of the agent!

from aixplain.factories import AgentFactory
from aixplain.modules.agent import ModelTool

agent = AgentFactory.create(
    name="Agent",
)



agent_response = agent.run("What's an agent?")
print(agent_response)

Show output

+ choose LLM

When creating an agent, you can specify which large language model (LLM) to use. The default is OpenAI's GPT-4o, but you can choose from other LLMs available in the aiXplain marketplace.

from aixplain.factories import AgentFactory
from aixplain.modules.agent import ModelTool

agent = AgentFactory.create(
    name="Agent",
    llm_id="66b2708c6eb5635d1c71f611", # groq llama 3.1 70B
)


agent_response = agent.run("What's an agent?")
print(agent_response)

Show output

+ add tools

You can also add tools to your agent, such as models or pipelines, to extend its capabilities beyond text-based responses. Tools enable the agent to perform specialized tasks, like generating audio or processing images.

from aixplain.factories import AgentFactory
from aixplain.modules.agent import ModelTool

agent = AgentFactory.create(
    name="Agent",
    tools=[
        ModelTool(model="6633fd59821ee31dd914e232", # speech synthesis
    ],
)
agent_response = agent.run("What's an agent? Answer with audio.")
print(agent_response)

Show output

tip

You can delete an agent using the delete method.

agent.delete()

tip

You can instantiate your agent by searching for its ID (see below) and using the get method.

agent = AgentFactory.get("66f744f390118d8653adcd8c")

2. Choose an LLM and add tools

You can explore the aiXplain marketplace to find the best models to integrate into your agent use case. This could be selecting a task-specific model or choosing the right large language model (LLM) to serve as your agent's core. You may also find models that are powerful enough for your use case without the need to create an agent, or that you'd prefer to combine into a pipeline.

2.1 Browse for models and tools

There are two ways to browse the assets available on marketplace: via Studio or the SDK.

• How to search the Marketplace (SDK)
• Search with Discover (Studio)

Here are three short examples of searching for assets in the SDK.

Models

from aixplain.factories import ModelFactory
from aixplain.enums import Function

model_list = ModelFactory.list(function=Function.TEXT_GENERATION, page_size=50)["results"]

for model in model_list:
    print(model.id, model.name, model.supplier)

Show output

Pipelines

from aixplain.factories import PipelineFactory


pipeline_list = PipelineFactory.list()["results"]

for pipeline in pipeline_list:
    print(pipeline.__dict__)

Show output

Agents

from aixplain.factories import AgentFactory


agent_list = AgentFactory.list()["results"]

for agent in agent_list:
    print(agent.id, agent.name)

Show output

2.2 Try a model

As with browsing, there are also two ways to try models (and pipelines):

• How to call and asset (SDK)
• Search with Discover - Try it out (Studio)

Here are three examples of calling models in the SDK.

Text generation

Suppose we searched TEXT_GENERATION functions and want to try

• GPT-4o Mini 669a63646eb56306647e1091.

model = ModelFactory.get("669a63646eb56306647e1091")

response = model.run("What is the capital of France?")
response

Show output

Image generation

Suppose we searched TEXT_TO_IMAGE_GENERATION functions and want to try

• Stable Diffusion XL 1.0 - Standard (1024x1024) 663bc4f76eb5637aa56d6d31.

model = ModelFactory.get("663bc4f76eb5637aa56d6d31")

response = model.run("A dog in a bathtub wearing a sailor hat.")
response

Show output

Speech synthesis

Suppose we searched SPEECH_SYNTHESIS functions and want to try

• English - Premium - Ivy (Child) 618ba6eae2e1a9153ca2a3ba
• English (India) - Premium 6171eec2c714b775a4b48caf
• English - Premium - Aria 618ba6e8e2e1a9153ca2a3b4

responses = [
    ModelFactory.get(model.id).run("Hi! Hope you're having a lovely day!")
    for model in models
]

responses

Show output

3. Next

Once you’ve built your first agent, there are additional features you may want to add to enhance its capabilities.

3.1 Multi-Agent Systems

You can orchestrate multiple agents to collaborate on complex tasks, forming what we call a "Team Agent." These agents can work together by delegating tasks, improving efficiency, and ensuring thorough task execution. For more detailed guidance, visit our Team Agent guide.

3.2 Memory

Agents can benefit from memory, allowing them to retain context over long interactions and improve their decision-making. You can read more on adding memory to your agents in the Agent Memory guide.

Ready to get started?

If you’re interested in building a multi-agent system or adding memory to your agent, check out the How to Build an Agent guide for a step-by-step process to set up these advanced features.