Configure handoff capacity and fix empty-output evidence selection - #3
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Empty search output previously triggered an unnecessary evidence-model call that generated nonexistent line numbers until its 160-token cap. Return empty evidence deterministically, expose public context/output and evidence/packet limits, and retain selection timings, raw responses, stop reasons, and rejected planning output in local artifacts. Defaults stay 4096/160; 8192/2048 is an explicit evaluation profile.
Three paired final runs through the public CLI used pinned sources, identical saved fresh cases, full workload warmups, alternating treatment blocks, PowerShell reference checks, and real artifact writes/reloads. Revised code passed 48/48 checks per profile versus 42/48 merged; all execution checks passed in every arm. The six merged failures were empty-output evidence selection. Pooled median/p95: merged 291.2/1186.1 ms, revised compatibility 288.0/438.1 ms, revised capacity 290.1/418.8 ms. No decode-throughput or whole-agent speed improvement is claimed.
Validation: 10 focused binding/contract checks; nine live CLI probes including 296-token selection of all 100 error lines, 5187-token input, actual PowerShell execution, explicit truncation/raw fallback, and rejected one-token planning with no execution. RUNTIME_RESULTS.md records per-run prompt/decode metrics, failures, cache limitations, and local evidence locations. Historical final-test corpora and model weights were untouched; raw diagnostic data remains ignored locally. No shared-service configuration changes, deployment, or automatic OpenCode2 integration.