Secret AI SDK is an advanced security layer designed to integrate seamlessly with DeepSeek, ensuring a secure and efficient API interaction. By requiring an API Key for access, it provides an additional layer of authentication and data protection, safeguarding sensitive communications.

This security protocol enhances the reliability and safety of DeepSeek's operations, offering developers confidence in managing and deploying their applications within secure boundaries. With Secret AI SDK, users can ensure that their interactions remain private and protected against unauthorized access.

Secret AI SDK caters to diverse application needs by supporting two types of clients: Synchronous and Asynchronous. Each client type is designed to check for the environment variable SECRET_AI_API_KEY, which must be correctly set to facilitate secure communication with the confidential LLM.

This configuration ensures that all interactions are authorized and protected, offering a robust framework for developing secure, reliable AI-driven applications.

Secret AI SDK is an advanced security layer designed to integrate seamlessly with DeepSeek, ensuring a secure and efficient API interaction. To begin utilizing Secret AI SDK, follow these installation and configuration steps:

Install Requirements

To follow along with the guide, we will be using Anaconda and Python 3.12.

1. Install Anaconda for Windows, Mac, or Linux.

2. Clone the Secret AI getting started repository here:

git clone https://github.com/SecretFoundation/secret-ai-getting-started.git

Create Anaconda virtual environment

In your text editor, create an Anaconda virtual environment with Python 3.12:

conda create -n menv python=3.12

Activate the virtual environment:

conda activate menv

Install Secret AI dependencies

pip install -r requirements.txt && pip install 'secret-sdk>=1.8.1' && pip install secret-ai-sdk

Set developer key

export SECRET_AI_API_KEY=bWFzdGVyQHNjcnRsYWJzLmNvbTpTZWNyZXROZXR3b3JrTWFzdGVyS2V5X18yMDI1

Understanding the application

Before you run secret_ai_getting_started.py, let's examine the code.

1. Importing the Required Modules

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

• ChatSecret: Handles communication with the LLM, such as sending messages and receiving responses.

• Secret: Manages secure access to LLM models and provides metadata about available models.

2. Initializing the Secret Client

secret_client = Secret()

• Creates an instance of Secret, which is used to interact with the Secret backend to fetch available LLM models and their connection details.

3. Fetching Models and URLs

models = secret_client.get_models()
urls = secret_client.get_urls(model=models[0])

• get_models(): Retrieves a list of all LLM models registered with the backend.

• get_urls(model=models[0]): For the first model in the list (models[0]), fetches a list of instance URLs where the model can be accessed.

4. Setting Up the LLM Client

secret_ai_llm = ChatSecret(
    base_url=urls[0], 
    model=models[0]', 
    temperature=1.0
)

• base_url=urls[0]: Specifies the first URL in the list as the endpoint to connect to the LLM.

• model=models[0]: Specifies the model to use (e.g., Llama or Deepseek, etc).

• temperature=1.0: Configures the "creativity" of the model's responses. Higher values produce more varied outputs, while lower values make responses more deterministic.

5. Defining the Chat Messages

messages = [
    ("system", "You are my therapist. Help me with my issues."),
    ("human", "I miss my cat."),
]

• Messages are defined as a list of tuples:

  • ("system", ...): Instructions for the model about its role (e.g., "therapist").

  • ("human", ...): A message from the user (e.g., expressing sadness about missing their cat).

6. Invoking the Model

response = secret_ai_llm.invoke(messages, stream=False)

• invoke(messages, stream=False): Sends the messages to the LLM for processing.

• messages: The list of chat messages.

• stream=False: Indicates that the response should be returned all at once rather than streamed in parts.

Manually Setting Up node_url

If you experience issues with the default node_url (SECRET_NODE_URL_DEFAULT config variable), instead of declaring the client like this:

secret_client = Secret()

you can manually specify a node_url when instantiating the Secret client:

from secret_ai_sdk.secret import Secret

secret_client = Secret(chain_id='pulsar-3', node_url=<LCD_NODE_URL>)

Alternatively, you can set the node_url via an environment variable:

export SECRET_NODE_URL=<LCD_NODE_URL>

For more details on available endpoints and a list of LCD nodes, refer to the official Secret Network documentation: Connecting to the Network - Testnet Pulsar-3

Run the application

To run the sample application:

python secret_ai_getting_started.py

Upon successful installation, you should see a response:

It's completely normal to feel a deep sense of loss and longing when we're separated from our beloved pets, especially ones as special as cats.

Was it a recent separation, or is this a lingering feeling you've been carrying around for a while?

Conclusion

The Secret AI SDK provides developers with an early glimpse into integrating confidential LLMs, enabling secure and innovative applications. As a developer preview, it is designed for testing and exploration, not production use. Feedback is encouraged to help refine the SDK for future releases.