tumeryk-guardrails 0.2

API Client for Tumeryk_Guardrails

Metadata-Version: 2.1 Name: tumeryk_guardrails Version: 0.2 Summary: API Client for Tumeryk_Guardrails Author: Tumeryk Author-email: [e-mail] Classifier: Programming Language :: Python :: 3 Classifier: Operating System :: OS Independent Requires-Python: >=3.6 Description-Content-Type: text/markdown Requires-Dist: requests

Metadata-Version: 2.1 Name: tumeryk_guardrails Version: 0.1.97 Summary: API Client for Tumeryk_Guardrails Author: Tumeryk Author-email: [e-mail] Classifier: Programming Language :: Python :: 3 Classifier: Operating System :: OS Independent Requires-Python: >=3.6 Description-Content-Type: text/markdown Requires-Dist: requests

Tumeryk Guardrails Client

The Tumeryk Guardrails Client is a Python package that provides an interface to the Tumeryk Guardrails API. The client allows you to easily use the API from your Python code.

Setup

To install the Tumeryk Guardrails Client, use pip:

pip install tumeryk_guardrails

Example .env File

TUMERYK_USERNAME=sample_username
TUMERYK_PASSWORD=sample_password
TUMERYK_POLICY=hr_policy

Simple Usage

You can use the Tumeryk Guardrails Client with minimal setup. The client will automatically load the configuration from the .env file if it exists. Here's an example of simple usage:

from dotenv import load_dotenv
load_dotenv()

import tumeryk_guardrails


messages = [{"role": "user", "content": "hi"}]

response = tumeryk_guardrails.tumeryk_completions(messages=messages)

print(response)

Manual Usage

The Tumeryk Guardrails Client uses chat.tmryk.com as the default base URL. However, you can change this URL if required. Here's how you can set a custom base URL:

import tumeryk_guardrails

# Set a custom base URL
tumeryk_guardrails.set_base_url("https://your-custom-url.com")

Configuration

The Tumeryk Guardrails Client uses environment variables to store credentials and policy. Here's an example of how to use environment variables:

import tumeryk_guardrails
import os
from dotenv import load_dotenv

# Load environment variables
load_dotenv()

# Retrieve credentials and policy from environment variables
username = os.getenv("TUMERYK_USERNAME")
password = os.getenv("TUMERYK_PASSWORD")
policy = os.getenv("TUMERYK_POLICY")

# Authenticate with Tumeryk Guardrails
tumeryk_guardrails.login(username, password)

# Retrieve available policies
policies = tumeryk_guardrails.get_policies()
print("Available Policies:", policies)

# Set the chosen policy
tumeryk_guardrails.set_policy(policy)

# Prepare a message for the guard service
messages = [{"role": "user", "content": "Example input to guard"}]

# Send a request to the guard service
response = tumeryk_guardrails.tumeryk_completions(messages)
print("Guard Response:")
print(response)

Response Structure

The guard service returns a structured response. Here's an example:

{
  {
  "messages": [
    {
      "role": "assistant",
      "content": "Hello! It's nice to meet you. How can I assist you today?",
      "stats": {
        "total_calls": 1,
        "total_time": 4.82445502281189,
        "total_tokens": 108,
        "total_prompt_tokens": 79,
        "total_completion_tokens": 29,
        "latencies": [2.410598039627075, 2.4138569831848145],
        "llm_process_time": 2.384185791015625e-07
      }
    }
  ],
  "metrics": {
    "violation": false,
    "jailbreak_detection": false,
    "trust_score": 674,
    "model_score": 674,
    "real_time_score": 0,
    "hallucination_score": 0.0,
    "bias_score": {
      "input": 0,
      "output": 0
    },
    "toxicity_scores": {},
    "llama_guard_allowed": {},
    "llama_guard_categories": {},
    "information": {
      "303": "Low Sensitive Information Disclosure Score"
    }
  },
  "log": {
    "general_stats": {
      "total_time": "2.43s",
      "dialog_rails_time": "99.84%",
      "llm_calls": 2,
      "llm_duration": "4.83s",
      "prompt_tokens": 158,
      "completion_tokens": 58,
      "total_tokens": 216
    },
    "detailed_stats": {
      "dialog": {
        "time": "2.43s",
        "intent_generation": {
          "actions": 1,
          "llm_calls": [
            {
              "time": "2.41s"
            },
            {
              "time": "2.41s"
            }
          ]
        }
      }
    }
  }
}

}

Available Methods

The Tumeryk Guardrails Client provides the following methods to interact with the Tumeryk Guardrails API:

• login(username, password): Authenticate and store access token.
• get_policies(): Fetch available policies and return a list.
• set_policy(config_id): Set the configuration/policy to be used by the user.
• tumeryk_completions(messages): Send user input to the Guard service.
• get_base_url(): Get the current base URL.
• set_base_url(base_url): Set a new base URL.

Dependencies

• requests: Used for making HTTP requests to the API.