What's Changed
Voici la version finale de la release note avec l'ajout du support de GPT sur OpenAI :
Hosting Options for Prompts and Subprompts
We’ve added hosting options within the framework for prompts and subprompts, providing flexibility for deployment pipelines. With this update, it's now possible to bundle code, custom actions, and prompts into a single image, separating prompt management from backend operations. This enhances the backend’s focus on data operations.
New Configuration Option:
LOAD_PROMPTS_FROM_BACKEND
: Set toTrue
orFalse
to choose prompt hosting mode.
Integration of OpenAI GPT Support
In addition to the existing support for Azure OpenAI, we’ve now added support for OpenAI’s GPT directly. This allows users to trigger text generation via OpenAI’s API, giving more flexibility for different deployment environments and use cases.
New Event Processing Plugins
We’ve implemented and standardized queue management for certain user interaction processes. This decouples GenAI calls from messaging delivery in Slack and other clients, enabling faster generation times by reducing dependencies on external APIs.
Supported Queue Systems:
- Local simulation with
file_system_queue
- Azure Blob Storage
Slack notifications (messages, reactions) are now parallelized across threads, with each thread’s messages ordered to ensure conversational coherence.
Enhanced GenerateText Action
The GenerateText
action has been updated to trigger generation using a specific model, with options to include or exclude conversation context, add supplementary context, and inject a specific main prompt. This gives more control over text generation and enables tailored interactions based on specific needs.
Updated UserInteraction for Message Targeting
UserInteraction
now supports sending messages directly to a specified channel or thread, distinct from the current conversation. This adds flexibility in managing interactions across different channels and threads, improving coordination and communication in diverse scenarios.
Enhanced Session Recording Model
We've enriched and standardized JSON logs for conversations, adding extensive metadata around exchanges (e.g., costs, triggered actions, processing times, models used). This lays the groundwork for comprehensive Quality & Performance monitoring for our bots throughout October.
Full Changelog: v0.7-beta.1...v0.7-beta.2