v0.2.0

Release v0.2.0

New Features

• Azure OpenAI Service Integration: This release introduces support for Azure OpenAI services. Users can now configure the application to connect to their Azure OpenAI deployments, in addition to the standard OpenAI API.
  • Configuration for Azure services (API key, endpoint, and API version) can be set via environment variables or a .env file.
  • A new "Service Type" selector has been added to the UI, allowing users to switch between "OpenAI (Standard)" and "Azure OpenAI" at runtime.
  • The application will dynamically fetch available models or use appropriate defaults (e.g., gpt-35-turbo for Azure) based on the selected service.

Enhancements

• Improved Configuration:
  • A STARTUP_SERVICE_TYPE environment variable can now be set to define the default service (OpenAI or Azure) used when the application starts.
  • The README.md has been updated with detailed instructions for configuring and using both standard OpenAI and Azure OpenAI services.
• User Interface:
  • The "Service Type" dropdown will indicate if Azure OpenAI is not configured on the backend and will default to standard OpenAI in such cases.
  • Minor improvement to error message display on the frontend.

Technical Changes

• The backend OpenAIClient has been refactored to support initialization for both OpenAI and AzureOpenAI SDKs.
• API endpoints for fetching models and generating text now accommodate the selected service_type.
• The .env.example file has been updated with new variables for Azure configuration.

Release v0.1.0 - Initial Release

Release v0.1.0

This release introduces the Token Probability Visualizer, a web application designed to help users understand the token probabilities generated by OpenAI models.

Overview

The Token Probability Visualizer allows you to:

• Visualize token probabilities from OpenAI models in generated text.
• See color-coded tokens indicating different probability levels (high probability = green, low probability = red).
• Hover over tokens to view detailed probability information and alternative likely tokens.

Key Features

• Model Configuration: Select from available OpenAI models associated with your API key, adjust temperature, top_p, and maximum tokens.
• Interactive Prompting: Enter prompts and generate text with probability visualization.
• Detailed Insights: Hover over tokens for log probabilities and alternative token options.
• Customizable Legend: Default probability thresholds and colors are provided (details in config.py and models/token_processor.py).

Getting Started

1. Installation: Please refer to the Installation Guide in the README for detailed setup instructions.
2. Running the App: Follow the Running the Application steps in the README.
3. Usage: See the Using the Application section for a guide on how to use the visualizer.

Requirements

• Python 3.10+
• Flask
• OpenAI API key (configured in a .env file)

---

Please report any issues or provide feedback.