Repository navigation
Refuse runs on an unreachable endpoint, TUI input history, selection and notifications - #54
Merged
Merged
Conversation
…and notifications A run started with the local model server down spent six turns of retries on every branch before the provider-error limit stopped it. Starting or resuming a run now asks the endpoint first and answers 503 naming it when nothing answers (gates.edn :endpoint-preflight). The llama.cpp probe no longer reads "connection refused" as "not llama.cpp": what an endpoint says is remembered per endpoint only once it answers, so a server started after the harness still gets prefix caching and the right fold role. A project whose gates.edn predates the key starts runs as before instead of a 500. The client retries only IOExceptions; an error raised by harness code inside a call is fatal and no longer counted as provider trouble. The supervisor's notice prints its measurements as words, the warm-up warning no longer talks about https, and the front ends wait longer for a run start than the server does. TUI: Up/Down at the compose box's edges walk the sent history and keep the line being typed; text can be drag-selected (the box included) and is copied to the clipboard on release; notifications are raised by the terminal itself, so clicking one goes to the tab instead of Script Editor.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Starting or resuming a run asks the model endpoint first and answers 503 naming it when nothing answers (gates.edn :endpoint-preflight), instead of every branch retrying into the provider-error limit. The llama.cpp probe remembers an endpoint's answer only once it answers, so a model server started after the harness is still identified. Projects whose gates.edn predates the key behave as before.
The client retries only IOExceptions; harness errors inside a call are fatal and not counted as provider trouble. Front ends now wait longer than the server's run-start deadline.
TUI: Up/Down at the input's edges walk history (draft kept), drag-selection works in the input and copies on release (OSC 52), and notifications are raised by the terminal so clicking goes to the tab. Needs jlt-commons/ftxui-jolt#3.