Katonic SDK

A Python SDK for interacting with Katonic's AI/ML platform, providing easy access to language models, vision models, and comprehensive request logging capabilities.

Features

• 🤖 Language Model Integration: Generate completions from various LLM models
• 👁️ Vision Model Support: Process text and image inputs with vision models
• 📊 Request Logging: Comprehensive logging with token usage and cost tracking
• ⚡ Simple API: Easy-to-use Python functions instead of raw API calls
• 🔍 Monitoring: Track and analyze LLM requests across your applications

Installation

pip install katonic

Model Completion

Import

from katonic.llm import generate_completion

generate_completion

Generates a completion from a specified model.

Parameters

Returns

• Type: str
• Description: Model-generated response as plain text

Finding Your Model ID

To find your model ID:

1. Navigate to My Model Library under LLM Management in the Katonic UI

   

2. Copy the model ID from the interface

3. Use this ID in the model_id parameter

Examples

Text Model

from katonic.llm import generate_completion

result = generate_completion(
    model_id="688b552061aa55897ae98fdc",
    data={"query": "Tell me a fun fact about space."}
)

print(result)
# Output: "Space is completely silent because there is no atmosphere to carry sound waves."

Vision Model (Text + Image)

from katonic.llm import generate_completion

result = generate_completion(
    model_id="688b552061aa55897ae98fdc",
    data={
        "query": "Describe what is in this image.",
        "image_url": "https://example.com/photo.jpg"
    }
)

print(result)

Response Format

The response will contain the model's generated text.

Example:

"Space is completely silent because there is no atmosphere to carry sound waves."

Request Logging

Import

from katonic.llm.log_requests import log_request_to_platform

log_request_to_platform

Logs user queries, responses, token usage, and cost details for monitoring and analysis.

model_name can be fetched from the following section of the platform.

Parameters

Returns

• Success: Returns a message_id (string) for tracking the logged request
• Failure: Returns None

Example

from katonic.llm.log_requests import log_request_to_platform

# Log a request with comprehensive details
message_id = log_request_to_platform(
    input_query="tell me about katonic",
    response="Katonic is a modern MLOps platform that helps enterprises manage AI/ML workflows efficiently.",
    user_name="developer@company.com",
    model_name="Openai/gpt-5-nano",
    product_type="Ace",
    product_name="Ace",
    project_name="Ace",
    latency=0.42,  # API response time in seconds
    status="success",  # status of the request
    embedding_model_name=None
)

print(f"Message logged with ID: {message_id}")

Response Format

On successful logging:

✅ Cost has been added successfully.
Message logged with ID: 650e95d2a8c7b123f5c123ab

Complete Workflow Example

Here's how to use both methods together:

from katonic.llm import generate_completion
from katonic.llm.log_requests import log_request_to_platform
import time

# Step 1: Generate completion
start_time = time.time()
query = "Explain quantum computing in simple terms"
model_id = "688b552061aa55897ae98fdc"

result = generate_completion(
    model_id=model_id,
    data={"query": query}
)

# Step 2: Calculate latency
latency = time.time() - start_time

# Step 3: Log the request
message_id = log_request_to_platform(
    input_query=query,
    response=result,
    user_name="developer@company.com",
    model_name="Openai/gpt-4",
    product_type="Research",
    product_name="QA Assistant",
    project_name="Science Education",
    latency=latency,
    status="success"
)

print(f"Response: {result}")
print(f"Logged with ID: {message_id}")

Notes

Model Completion

• Use the correct model_id provided in the Katonic UI
• For image-based models, ensure image_url is accessible over the internet
• The SDK handles API calls internally, so you only need to focus on inputs and outputs

Request Logging

• The SDK automatically calculates token usage and cost (if pricing data is available)
• If the request fails, you will see a warning message in the console
• Use the returned message_id to track logged requests

Requirements

• Python 3.7+
• Internet connection for API calls
• Valid Katonic platform credentials

Support

For issues, questions, or feature requests, please contact the Katonic support team or refer to the official Katonic platform documentation.

License

Please refer to your Katonic platform agreement for licensing terms.