feat(tui,experiments): reverse search backward nav and subject error tolerance#4976
Merged
Conversation
bug-ops
enabled auto-merge (squash)
June 7, 2026 10:45
bug-ops
force-pushed
the
4856-tui-experiments
branch
2 times, most recently
from
June 7, 2026 10:51
8f7b420 to
ed27c3b
Compare
…tolerance Add `select_previous()` to `ReverseSearchState` with wrap-around semantics (matching bash/zsh Ctrl+R/Ctrl+S conventions) and wire it to Ctrl+S in `handle_reverse_search_key`. Closes #4691. Add `tolerate_subject_errors: bool` config field to `ExperimentConfig` (default false). When true, per-case Phase 1 failures are excluded from scoring instead of aborting the run — matching Phase 2 graceful degradation behavior. Failed cases increment `error_count` and set `is_partial = true`. `EvalReport::is_partial` documents the partial-anchor comparison caveat. Closes #4856.
bug-ops
force-pushed
the
4856-tui-experiments
branch
from
June 7, 2026 10:59
ed27c3b to
6612a0d
Compare
…-prompt tests skill_bodies_prepended_to_system_prompt and no_skills_does_not_add_fence_to_system_prompt used a hard-coded 200 ms sleep to wait for the mock provider to be called. Under CI load (shard 5/5, many concurrent tests) the async task was not scheduled in time, causing a spurious failure. Replace the sleep with a 5 s timeout polling loop that exits as soon as the provider is called — fast in practice (<50 ms), reliable under load.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
select_previous()toReverseSearchStatewith wrap-around semantics. WireCtrl+Sinhandle_reverse_search_keyto navigate toward newer matches, matching bash/zshCtrl+R/Ctrl+Sconventions.select_next()updated to also wrap around (was clamping).tolerate_subject_errors: booltoExperimentConfig(defaultfalse). Whentrue, per-case Phase 1 failures are excluded from scoring instead of aborting — matching Phase 2 graceful degradation. Failed cases logged atWARN, counted inerror_count, andis_partial = trueis set inEvalReport.EvalReport::is_partialdocuments the partial-anchor comparison caveat.Changed files
crates/zeph-tui/src/widgets/reverse_search.rsselect_previous()+ updatedselect_next()(wrap), 4 new testscrates/zeph-tui/src/app/keys.rsCtrl+Sbinding inhandle_reverse_search_keycrates/zeph-config/src/experiment.rstolerate_subject_errors: boolfield (default false)crates/zeph-experiments/src/evaluator.rswith_tolerate_subject_errors()buildercrates/zeph-core/src/agent/experiment_cmd.rsconfig.tolerate_subject_errorsintoEvaluatorCHANGELOG.md[Unreleased]entries for both featuresTest plan
cargo nextest run -p zeph-tui --lib --bins: 568 PASS (4 new reverse-search tests)cargo nextest run -p zeph-experiments -p zeph-config --lib --bins: 663 PASScargo nextest run -p zeph-core --lib --bins: 1533 PASS (wiring test included)cargo clippy -p zeph-tui -p zeph-experiments -p zeph-config -p zeph-core -- -D warnings: CLEANcargo +nightly fmt --check: CLEANCtrl+Skey — blocked (no interactive terminal in CI)--experiment-runwithtolerate_subject_errors = true— blocked (LLM credits unavailable)Closes #4691
Closes #4856