You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Dedicated Image Action: Added a new standalone action for generating images.
Provider Support: * Full support for OpenAI (DALL-E 3 & DALL-E 2).
Full support for Stable Diffusion (Local) via the Automatic1111 API.
Customization: Control image size, quality, and style parameters.
Flexibility: Added support for custom model identifiers.
New Feature: Function Calling (Tools)
Tools Tab: Added a dedicated "Tools" tab within the AI Chat action.
Custom Declarations: Define JSON function declarations that the AI can trigger autonomously.
Structured Output: New output variable %ai_tool_calls containing the structured JSON request data for Tasker to process.
New Feature: JSON Mode
Schema Enforcement: Added a "Force JSON Mode" toggle to ensure the AI returns strictly valid JSON.
Cross-Provider Compatibility: Supported across OpenAI, Gemini, and Groq.
Improved Provider Support
Gemini Updates: Migrated to the v1beta endpoint, enabling Gemini 1.5 Pro/Flash and the latest 2026 model iterations.
Claude Integration: Refined system prompt handling for better instruction following.
Vision Tasks: Improved multi-modal message construction for more reliable image analysis.
UI & Experience Enhancements
Tabbed Interface: Reorganized configuration into Manual, Variable, and Tools tabs for a cleaner workflow.
Auto-Detection: Added model ID auto-detection for supported providers.
Modern Design: Updated the message builder with a modernized card-based layout.
Quick Tips for v0.0.8:
JSON Parsing: When using JSON Mode, use Tasker’s native "JSON Data Yield" or "Variable Search/Replace" to grab specific data points from the response.
Stable Diffusion: Ensure your A1111 instance is launched with the --api flag, or the plugin won't be able to "handshake" with your local PC.
Gemini 1.5: With the v1beta update, you can now send much larger files (or even short video clips) if your Tasker profile supports the file-path variable.