Secret AI SDK

The Secret AI SDK provides a simple and convenient way to access Secret Confidential AI models. With this SDK, you can easily integrate Secret's AI capabilities into your own applications and services.

Overview

The Secret AI SDK is a Python library that enables access to Secret Confidential AI models. The SDK provides a simple and intuitive API that allows you to send requests to Secret's AI models and receive responses in a variety of formats.

Features

• Access to Secret Confidential AI models
• Simple and intuitive API

Requirements

secret-ai-sdk has a list of dependencies as defined in requirements.txt file All, but secret-sdk-python, are listed in pyproject.toml file

You will need to install secret-sdk-python as shown below:

pip install 'secret-sdk>=1.8.1'

Installation

To install the Secret AI SDK, you can use pip:

pip install secret-ai-sdk

Usage

Here's an example of how to use the Secret AI SDK:

from secret_ai_sdk.secret_ai import ChatSecret
from secret_ai_sdk.secret import Secret

secret_client = Secret()
# Get all the models registered with the smart contracts
models = secret_client.get_models()
# For the chosen model you may obtain a list of LLM instance URLs to connect to
urls = secret_client.get_urls(model=models[0])
# You previosly exported the env var SECRET_AI_API_KEY=YOUR-API-KEY
secret_ai_llm = ChatSecret(
base_url=urls[0], # in this case we choose to access the first url in the list
model=models[0], # your previosly selected model
temperature=1.
)
# Define your messages you want to send to the confidential LLM for processing
messages = [
(
      "system",
      "You are a helpful assistant that translates English to French. Translate the user sentence.",
),
("human", "I love programming."),
]
# Invoke the llm
response = secret_ai_llm.invoke(messages, stream=False)
print(response.content)

If you do not need to select a specific LLM URL (let's say your agent does not need to worry about the contextual memory from previous sessions), you can make a simplified call to ChatSecret. Given the exported env var SECRET_AI_API_KEY='YOUR_API_KEY':

from secret_ai_sdk.secret_ai import ChatSecret

model = ChatSecret(model='llama3.1:70b')

response = model.invoke([system_message, human_message])

print(response)

You can reference test_secret_ai.py and test_secret.py to see how we tested the code as it may help you in writing your own implementation.

API Documentation

For more information on the Secret AI SDK API, please see our API documentation.

License

The Secret AI SDK is licensed under the MIT License.