mirror of
https://github.com/QwenLM/qwen-code.git
synced 2026-08-03 05:14:52 +00:00
4aa7d1ec0b
* feat(cli): adopt Goal v3 in interactive TUI * fix(cli): restore Goal CI contracts * fix(cli): stop a queued message from stalling the Goal loop (#8005) Reserve the next Goal turn only for a user batch the queue will actually release: peekNextUserBatchKey now mirrors the two-lane drain gate, so an active Goal turn no longer binds a permit to a held plain message and the loop keeps running. Also release the Goal turn binding on the background-capacity early return, drop a redundant getSteerInput spread, recognise the legacy /goal clear aliases in interactive mode, and restore the rollback assertion for loadPausedBackgroundAgents. * fix(cli): address review feedback on Goal queue and retry hint (#8005) * fix(cli): restore Goal trust gate and address review feedback (#8005) * fix(cli): address Goal review feedback on cancel release, icons, claim gate (#8005) * fix(cli): address Goal review feedback on non-interactive guard, icons, and drain retry (#8005) * fix(cli): address review feedback on Goal v3 TUI (#8005) - Strip 'set' keyword before forwarding to legacy path in non-interactive /goal set, preventing the keyword from leaking into the goal condition - Add goalTerminalErrorRef to prevent post-stream cleanup from wiping Goal turn terminal errors - Add 50-turn continuation budget to Goal runtime, matching the legacy MAX_GOAL_ITERATIONS cap - Display verifier rejection status cards (verifier_reject cause) - Clear stale lastReason when editing a goal objective * fix(test): align client-goal turn count with continuation budget (#8005) * fix(cli): address Goal set-clear and delivery-error review feedback (#8005) Non-interactive `/goal set <clear-keyword>` now sets a literal objective instead of clearing the active goal, by bypassing the clear-keyword check for explicit set operations. Goal-turn stream errors now fire onDeliveryFailed rather than onDelivered by including goalTerminalErrorRef in the post-stream delivery dispatch. * fix(cli): rename misleading missingActiveGoalContext variable (#8005) * fix(cli): stop Goal queue from stranding input and orphaning prompts (#8005) Only an active Goal turn holds ordinary input now; paused, blocked and usage_limited states drain the queue normally so a user whose Goal is merely paused (e.g. via Escape) is never stranded waiting for /goal clear. A cancelled Goal continuation turn also strips its synthetic "no new real user input" prompt from the chat history. Previously the auto-restore branch bailed before its orphan strip ran (Goal turns add no UI user item), so the preamble survived and appendCuratedContent merged the user's next real message into it. * fix(core): address Goal resume-budget and mid-turn clear review feedback (#8005) Resuming a Goal that exhausted its 50-turn continuation budget was accepted and reported as `active`, then immediately re-transitioned to `usage_limited` without running a turn: the resume branch kept the exhausted `turnCount`, which `queueContinuation` re-checks. Resume now resets `turnCount` so an explicit resume grants a fresh continuation budget and the reported outcome matches the settled one. A mid-turn `/goal clear` with no active Goal was silently swallowed because its causeless `goal_control` result rendered only when idle. The handler now renders any causeless result (a `status` read or a no-Goal `clear`), which never broadcasts and so cannot double-render. * test(cli): cover Goal tool-batch fail-close guards (#8005) * fix(core,cli): guard budget-exhaustion identity and pass goalContext to tool-result recording (#8005) The budget-exhaustion callback in queueContinuation only checked goal status, not identity. A replace dispatched during the journal append could create a fresh goal that the stale callback then incorrectly usage-limited. Capture goalId/revision at enqueue time and return early on mismatch, matching the existing identity-guard pattern used by handleStartFailure and recordVerificationOutcome. The TUI recorded tool results without goalContext, so the evidence catalog never admitted tool-result records and any Goal whose objective depended on external state could not be verified. Pass request.goalContext at both recordToolResult call sites in useGeminiStream, tagging get_goal/update_goal results as goal_runtime to match the CoreToolScheduler pattern. --------- Co-authored-by: qwen-code-dev-bot <269191875+qwen-code-dev-bot@users.noreply.github.com> Co-authored-by: Shaojin Wen <shaojin.wensj@alibaba-inc.com> Co-authored-by: Qwen Code Bot <qwen-code-bot@users.noreply.github.com> Co-authored-by: Qwen Code Autofix <qwen-code-autofix@users.noreply.github.com> Co-authored-by: qwen-code-dev-bot <qwen-code-dev@service.alibaba.com> Co-authored-by: qwen-code-dev-bot <qwen-code-dev-bot@users.noreply.github.com> Co-authored-by: qwen-code-ci-bot <qwen-code-ci-bot@users.noreply.github.com> |
||
|---|---|---|
| .. | ||
| src | ||
| .gitignore | ||
| index.ts | ||
| package.json | ||
| test-setup.ts | ||
| tsconfig.json | ||
| vitest.config.ts | ||