v0.7.1 (6-Mar-2025)

Fixed supported python version in the installation package

• Supported python version are: 3.10.xx, 3.11.xx, 3.12.xx

v0.7.0 (12-Dec-2024)

New features introduced supporting Bhashik Text and Speech APIs:

• Entity Extraction API
  Enables extraction of personal identifiable information (PII) and named entities (NER) such as names, locations, organizations, and phone numbers from text, supporting multiple languages.
• Summarization API
  Generates concise summaries of text with customizable summary sizes.
• Translation API
  Facilitates text translation between multiple languages.
• Sentiment Analysis API
  Determines the sentiment of text, identifying emotions like joy, sadness, or neutrality.
• Text-to-Speech (TTS) API
  Converts written text into spoken audio, with support for multiple speakers.
• Text-to-Speech Translation API
  Translates text into a different language and then converts it into spoken audio.
• Speech-to-Text (STT) API
  Transcribes speech from audio files into text.
• Speech-to-Text Translation API
  Converts speech into text and translates it into a different language.
• Speech-to-Speech Translation API
  Translates speech from one language to another and provides the output as audio.
• Speech-to-Text Long Duration API
  Supports transcription of long-duration audio files.
• Speech-to-Text Translation Long Duration API
  Enables translation of long-duration audio into text in a different language.
• Language Detection API
  Detects the language of input text automatically.

V0.6.1 (15-Oct-2024)

Fixes added for proper error handling

v0.6.0 (26-Sep-2024)

New Feature:

Added support of Fine tuning and Inference Services for LLMs

v0.5.0 (19-Sep-2024)

Support added & updated for models:

Text to Video: THUDM/CogVideoX-2b
- Example

Multimodal:
Replaced Idefics2 with Idefics3 model

v0.4.1 (19-Sep-2024)

Enhancement: Added automatic version update after BYOM model deployment, enabling inference of deployed LLM models directly from the Deployment UI Playground.

v0.4.0 (13-Sep-2024)

New Features:

1. Bring Your Own Model (BYOM) Support for Text-to-Text LLM Models:

   Users can now upload their custom model files to their dedicated Krutrim S3 bucket. This provides greater flexibility to integrate custom LLM models into the Krutrim cloud platform.

   Example Notebook

2. Model File Management:

   List and Download Models: Users can easily list and download model files stored in Krutrim S3 storage through the SDK, making model management more seamless.
   List Model Versions: Users can list all versions of the models they have uploaded to Krutrim S3 storage, simplifying version tracking and management.

3. Model Deployment:

   Users can deploy models uploaded to the Krutrim S3 bucket directly via the SDK, streamlining the deployment process for custom LLM models.

4. Deployment Management:

   List Deployments: The SDK allows users to retrieve a list of all their deployments.
   Check Deployment Status: Users can check the status of any deployment using the SDK to track progress or identify issues.
   Cancel Deployment: If needed, users can cancel a specific deployment by providing the deployment ID, offering more control over model operations.

v0.3.1 (12-Sep-2024)

[Text-to-Text Model]

• Removed support of Reflection-Llama-3.1-70B as removed from Krutrim cloud

v0.3.0 (9-Sep-2024)

[Text-to-Text Models]

• Added support of Gemma-2-27B-IT & Reflection-Llama-3.1-70B for ChatCompletion API

[Text-to-Image Model]

• Updated input validation for input height and width of stable diffusion model parameters [min:256, max:2048]

v0.2.0 (29-Aug-2024)

Audio Models

• New Features:
  • Extended Audio Processing: Now supports processing of longer audio files more than 6 hours.
  • Video Input Support: Added functionality to handle video files by automatically converting video to audio for processing.
  • Audio Validation: Introduced a validation mechanism to ensure audio integrity and format compliance.
  • Audio Manipulation: Implemented support for cropping and concatenating audio files using PyDub, with export capabilities to local storage.

Image Models

• Examples Added:
  • Support for the OpenCV library.
  • Reading and saving images in various color spaces: RGB, RGBA, CMYK.
  • Image resizing with multiple algorithms.
  • Image manipulation functionalities including:
    • Rotation
    • Flipping
    • Grayscale conversion
    • Brightness and contrast adjustment

Text Models

• New Features:
  • Added a Prompt Registry with example usage for enhanced text model prompting.