v0.4.5-alpha release

Purpose

• Added o1-assistant Configuration:

  • The o1-assistant configuration loads by default in the OPEN_AI configuration.
  • The assistant can be used to obtain chat completions from o1-mini and o1-preview models (or other models specified by the user) while still utilizing threads, images, and tools functionalities, which are currently limited with o1 models.
  • The o1 model response can be controlled via #o1-mini and/or #o1-preview user commands.

• Added the Following System Functions:

  • get_openai_chat_completion(prompt: str, model: str) -> str: Generates chat completions using OpenAI models.
  • get_azure_openai_chat_completion(prompt: str, model: str) -> str: Generates chat completions using Azure OpenAI.

• Note: Due to the use of multiple models, token consumption and latencies may be higher.

• Version Updated to 0.4.5a1

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[ ] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.5a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.4.4-alpha release

Purpose

• Change ConversationThreadClient so that old image files are kept on thread
• Change ConversationMessage and AsyncConversationMessage classes to support multiple FileMessage instances per instance
• Change max tokens default value to 1000
• Add contains_file_id method and remove contains_image_file_id from Conversation and AsyncConversation classes
• Version updated to 0.4.4-alpha

Does this introduce a breaking change?

The file_message property of the Message and AsyncMessage classes has been changed to file_messages and now returns a list of FileMessage objects. To go through individual file messages you can iterate through file_messages.

The contains_image_file_id method has been removed from the Conversation and AsyncConversation classes. To check if a conversation contains any file with the specified file id, use the new contains_file_id method

[X] Yes
[ ] No

Release Type

[X] Bugfix
[ ] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.4a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.4.3-alpha release

Purpose

• Improved validity checking for MultiTask class for sync and async
• Improved multi-agent sample
• Version updated to 0.4.3-alpha

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[X] Bugfix
[ ] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.3a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.4.2-alpha release

Purpose

• Fixed issues #36 and #35. Consecutive conversation messages with image attachments can now be processed
• Version updated to 0.4.2-alpha

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[X] Bugfix
[ ] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.2a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.4.1-alpha release

Purpose

• Support added for image attachments and image URLs as input for Assistant Tool and SDK middleware library. Image input support is added for both AssistantClient(OpenAI assistant API based) and ChatAssistantClient (OpenAI chat completion API based) types.
  • With Assistant Tool, images can be attached as user input by copy pasting the image into the text input or by right clicking the thread list item and adding the image file.
• Added new Attachment class to make use of attachments easier
• Change ConversationMessage and AsyncConversationMessage classes to support multiple ImageMessage instances per instance.
• Removed image_messages parameter from Conversation and AsyncConversation classes.
• Version updated to 0.4.1-alpha

Does this introduce a breaking change?

There is a change in create_conversation_thread_message method for ConversationThreadClient class for attachments, which now takes a list of Attachment instances instead of dictionary. If you were not using attachments, there should not be a break.

There is a change in the ConversationMessage and AsyncConversationMessage classes such that image_message parameter is now a list, image_messages, of ImageMessage instances instead of a single ImageMessage.

The image_messages parameter of the Conversation and AsyncConversation classes has been removed. To go through image messages within a conversation, you can go through all messages from the messages property and look at each messages image messages.

[X] Yes
[ ] No

Release Type

[X] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.1a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.
• If using Mac, image copy and paste does not work due to potential pyside6 issue

v0.4.0-alpha release

Purpose

• Refactored Conversation and TextMessage, FileMessage and ImageMessage classes. Main reason is that the AssistantClientCallbacks implementation was insufficient and did not provide information about other than text messages in on_run_update method. In streaming scenario, it is important to get information about text, file citiations, files, images while streaming the content.
  • Conversation/AsyncConversation classes
    • Creates ConversationMessage/AsyncConversationMessage instances and provide convenience methods to get messages (including text, file and image contents) and text messages
  • New ConversationMessage/AsyncConversationMessage classes
    • Creates and provide properties for TextMessage, FileMessage and ImageMessage instances with contents
  • ConversationThreadClient/AsyncConversationThreadClient classes
    • AssistantClientCallbacks on_run_update method has been modified to pass ConversationMessage, which is used to provide convenient way to get text, file, image content information from the OpenAI assistant messages
    • New method retrieve_message added to ConversationThreadClient to convert OpenAI Message to ConversationMessage with parsed information.
  • Fixed bugs in async context when testing inside Azure Container App environment
  • Updated samples and templates for required changes due to refactoring
• 2024-05-01-preview API version is default for Azure OpenAI unless specified in AZURE_OPENAI_API_VERSION env variable
• Streaming has been now enabled as default in Assistant Tool. If you want to switch streaming off, you can do that in Settings -> General menu.
• Update the version to 0.4.0-alpha.

Does this introduce a breaking change?

Due to Conversation class restructure, there is breaking change. Please see the updated samples and sample templates for the required changes.

[X] Yes
[ ] No

Release Type

[X] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[X] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.4.0a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.3.7-alpha release

Purpose

• Add AI client keyword arguments for client classes to enable different authentication methods (key, default Azure credential chain).
• Fix conversation thread creation for ChatAssistantClients to avoid using attachments in threads, as they are not supported with chat completion.
• Refactor the assistant configuration saving process to default to the home folder and save the configuration to a file locally only if the user has configured it. With the assistant tool, this happens when the AI client changes or when the application closes.
• Fixes to defaults to enable azure-ai-assistant library usage when deployed to Azure Container Apps
  • Remove writing to the assistant.log file by default and require enabling it by using the ASSISTANT_LOG_TO_FILE environment variable.
  • Remove writing to the threads.json file as default. The config folder needs to be set if the user wants to store thread information locally.
  • Fix a bug where if the AZURE_OPENAI_API_VERSION environment variable is set to an empty string, the default API version was not selected.
• Fixed async stream event handler awaits for callbacks.
• Improved installation of the tool using requirements.txt
• Update the version to 0.3.7-alpha.

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[X] Bugfix
[ ] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.3.7a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.3.1-alpha release

Purpose

• Fix UI bugs when updating and pre-loading assistant information using assistant config dialog
• Update system assistant configurations to the new structure
• Fix cancelling run functionality to be more reliable
• Version update to 0.3.1-alpha

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[X] Bugfix
[ ] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.3.1a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.3.0-alpha release

Purpose

• Support for Assistant V2 API for following features:
  • Added support to assistant tool UI and middleware library for temperature, top_p, response_format, truncation_strategy, max_completion_tokens, max_prompt_tokens setting
  • Added support for tool resources for code interpreter and file search tools for the assistant tool UI and middleware (assistant creation and update)
  • Added support for attachments (files for code interpreter and file search tools) when adding message to thread.
    • Issue reported to OpenAI https://community.openai.com/t/assistant-cannot-access-to-attached-file-in-thread-message-when-using-file-search-as-tool/731347
  • Updated samples and tests for new V2 structure
• Migration instructions
  • Assistant configuration yaml files require update for tool_resources and file_search tools, see the updated samples for required changes.
• Minor UI updates to the tool
• Version update to 0.3.0-alpha

Does this introduce a breaking change?

Yes, primarily due to OpenAI V2 API changes. NOTE: AzureOpenAI does not yet support OpenAI V2 changes.

To use V1 version, do following steps:

1. Take latest v0.2.12-alpha branch content from https://github.com/Azure-Samples/azureai-assistant-tool repository for UI
2. Install azure-ai-assistant library 0.2.12a version.
3. Install openai 1.20.0 version or earlier.

[X] Yes
[ ] No

Release Type

[X] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[X] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Other Information

• Please install new version azure_ai_assistant-0.3.0a1-py3-none-any.whl for middleware SDK from this release. See the README for installation details.

v0.2.12-alpha release

Purpose

• Added async versions for task management APIs
• Updated multi-agent orchestration sample to use async APIs for better usability and added new agent for file creation in the sample
• Fixed bugs in asynchronous code execution
• Version update to 0.2.12-alpha

Does this introduce a breaking change?

[ ] Yes
[X] No

Release Type

[X] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[X] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[X] Other... Please describe: Sample for Multi-Agent Code Orchestration

Other Information

• Please install new version azure_ai_assistant-0.2.12a1-py3-none-any.whl for middleware SDK. See the README for details.