-
Notifications
You must be signed in to change notification settings - Fork 232
Description
To the Developer of Extended OpenAI Conversation,
I am writing to formally request a review of the configuration user interface, specifically regarding the setup flow for local LLM backends (e.g., Ollama).
As a user attempting to integrate a local AI agent, I discovered that the current UI layout obscures two critical settings. If these are missed—which the current design makes likely—the integration fails completely, leading to hours of unnecessary troubleshooting.
-
The "Skip Authentication" Checkbox For local endpoints (like http://192.168.x.x:11434), authentication is rarely used. However, this setting is currently buried or unticked by default.
The Failure Mode: If a user enters their Local URL but misses this box, the connection fails silently or with generic errors, leading users to believe their networking or server configuration is at fault.
Recommendation: If a user enters a local IP or HTTP URL (non-HTTPS), please consider highlighting this box or enabling it by default.
-
The "Use Tools" Checkbox This setting is essential for the integration’s primary purpose: controlling home devices. Currently, it is located at the very bottom of the configuration scroll.
The Failure Mode: A user can successfully connect the AI, but without this box checked, the agent cannot control entities. The user experiences a "working" chat that refuses to execute commands, creating the impression that the prompt or the model is broken.
Recommendation: This should be enabled by default, or placed prominently at the top of the configuration alongside the Model selection.
The current workflow effectively hides the "Engine" (Skip Auth) and the "Steering Wheel" (Use Tools), setting users up for inevitable failure. Small changes to these defaults would significantly reduce user frustration and support volume.
Thank you for your work on this integration and for considering these UI improvements.
Best regards,
attackbat A frustrated but hopeful Home Assistant User
Gemini can make mistakes, so double-check it