0.8.0

What's Changed

• Image generation fixed by @s-nagaev in #31

Full Changelog: v0.7.0...v0.8.0

0.7.0

Added

• Anthropic provider support.
• MistralAI provider support.
• Optional HIDE_MODELS setting to hide models option from the bot menu.
• Optional HIDE_IMAGINE setting to hide imagine option from the bot menu.

Changed

• Project structure significantly refactored.
• README updated: environment variables description significantly improved.

0.6.2

Added

• Optional MODELS_WHITELIST setting to limit the number of available GPT models.

0.6.1

Added

• Optional IMAGE_GENERATIONS_LIMIT setting to limit the number of images a user can generate within a month using
  DALL-E (to avoid excessive spending). The default value is 0, which means the feature is turned off, and no limits are
  applied.
• Optional IMAGE_GENERATIONS_WHITELIST setting for a list of user IDs exempt from image generation limits.

0.6.0

Fixed

• Fixed an issue preventing the bots' use with local data storage
• Resolved a problem where the bot would lose asynchronous tasks during long server response times from OpenAI
• Fixed a bug that sometimes prevented sending images to users

Changed

• Updated key project dependencies and adapted code for integration with the latest version of the OpenAI library
• Significantly improved the informativeness of logs
• Implemented the ability to send large responses across multiple messages
• Added the option to select the model for generating images (currently only through bot setting using environment variables)
• Refactored the code
• Base Docker image updated to python:3.11-alpine
• REDIS_PASSWORD environment variable is deprecated.

0.5.3

Changed

• Project dependencies updated.

0.5.2

Fixed

• A bug when users whitelist couldn't work with the telegram user ID.

0.5.1

Added

• LOG_PROMPT_DATA setting. If true, the application will log user's prompts and GPT answers. The default value is False.

Changed

• The log of disallowed user actions now includes the user's ID.

0.5.0

Fixed

• A bug when the bot didn't answer messages quoting the bot's ones.
• A bug when the bot checked the user whitelist using the user's username, not the user's name.
• A bug when setting the application via .env was impossible.

Added

• Now it's possible to connect to Redis via password (using the REDIS_PASSWORD setting).

Changed

• Settings refactored.

0.4.0

Added

• "Public mode" - now the bot can run without a "master" OpenAI API Key, and every user will have to provide their own.

Changed

• Codebase has been refactored.