despite configuration of AI online endpoint, the chat complains that it isn't configured #33

Closed
opened 2026-07-19 15:08:22 +00:00 by hugo · 1 comment
Owner

I get the error "validation error: AI unavailable - configure online endpoint in Settings" despite just having configured just that in preferences and loaded a model selection. It seems the actual change wasn't stored in the database on saving (save button below AI settings).

Also that error message is not translated but shown in english.

Use my localhost endpoint for model endpoint: http://127.0.0.1:9000/v1 with api key "dummy" and when selecting any model, use mlx-community--gemma-4-12B-8bit

I get the error "validation error: AI unavailable - configure online endpoint in Settings" despite just having configured just that in preferences and loaded a model selection. It seems the actual change wasn't stored in the database on saving (save button below AI settings). Also that error message is not translated but shown in english. Use my localhost endpoint for model endpoint: http://127.0.0.1:9000/v1 with api key "dummy" and when selecting any model, use mlx-community--gemma-4-12B-8bit
hugo added the bug label 2026-07-19 15:08:22 +00:00
Author
Owner

Re-tested with the rebuilt macOS app using the configured online endpoint at port 9000; MLX Server was not used. Changing the AI endpoint URL in the GUI was persisted immediately, and bds-cli config list showed the same ai.endpoint.online.url value both for the temporary localhost value and after restoring http://127.0.0.1:9000/v1. An adjacent GUI regression was found and fixed: SettingsChanged had rebuilt only part of Settings state, which discarded the in-memory AI configuration and made chat falsely report AI unavailable until restart. Settings now fully rehydrates while preserving the active AI section. The missing saved-toast translations and the raw AI-unavailable configuration error were also fixed. In the rebuilt bundle, Save stays on AI, shows the localized saved toast, and chat immediately reaches the configured endpoint without restart. The endpoint returned its own HTTP 400 for the test prompt, which is separate from the configuration persistence issue. cargo test --workspace and cargo build --workspace pass.

Re-tested with the rebuilt macOS app using the configured online endpoint at port 9000; MLX Server was not used. Changing the AI endpoint URL in the GUI was persisted immediately, and bds-cli config list showed the same ai.endpoint.online.url value both for the temporary localhost value and after restoring http://127.0.0.1:9000/v1. An adjacent GUI regression was found and fixed: SettingsChanged had rebuilt only part of Settings state, which discarded the in-memory AI configuration and made chat falsely report AI unavailable until restart. Settings now fully rehydrates while preserving the active AI section. The missing saved-toast translations and the raw AI-unavailable configuration error were also fixed. In the rebuilt bundle, Save stays on AI, shows the localized saved toast, and chat immediately reaches the configured endpoint without restart. The endpoint returned its own HTTP 400 for the test prompt, which is separate from the configuration persistence issue. cargo test --workspace and cargo build --workspace pass.
hugo closed this issue 2026-07-20 09:02:42 +00:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: hugo/RuDS#33