Cyrene_v2

809 Commits 7 Branches 0 Tags

3 Commits

Include renames

Author	SHA1	Message	Date
AskaEth	a0b4c941e4	fix(loop): keep cross-batch voice success and gate speak without dispatcher (M57) 两处 review 必修（review-m57 r1）： P2-1 跨批覆盖：applyBatchVoice 原先无条件用后一批的语音结论覆盖前一批，于是「先成功合成 abc.wav、下一批换个参数再试却失败」会把 Result.Voice 退化成 tts_failed —— 音频已落盘，退回文本既无收益也违背那次表达。改为守 captureVoice 的同一条优先级：已落盘的音频不被后续批次的失败抹掉；先降级后成功仍然覆盖（后到的成功即最终语音）。 P2-2 nil-dispatcher 安全门：speak 由循环自己承载、无 dispatcher 也能执行（invokeOne 短路），但 applyToolAdmissions 在 dispatcher 缺席时整段跳过安全门，使「没配 dispatcher」成为绕过 design/11 §3 的旁路。新增 applyDispatcherlessVoiceGate：与策略引擎对未登记工具的口径一致（默认关闭、只有 owner 放行），拒绝走同一条 not_allowed 软结果与 tool.reject.not_allowed 审计通道；并改正该处过时注释。两条回归测试都做过改前红/改后绿： - TestApplyBatchVoice_KeepsCrossBatchSuccess：注释掉守卫即报「voice = ... FallbackReason:tts_failed, want abc.wav」 - TestApplyToolAdmissions_DispatcherlessSpeakGate：去掉 else 分支即报「env group_member run = true, want false」验收：go build ./... / go vet ./... / go test ./... 全绿。	2026-10-27 22:00:18 +08:00
AskaEth	c9da488749	refactor(loop): extract speak voice backfill into applyBatchVoice (M57)	2026-10-27 21:33:05 +08:00
AskaEth	1347a77d33	feat(loop): wire speak tool into the loop (M57) - speak_tool.go: the speak tool itself — rune-derived timeout budget, structured failure codes, persona/emotion stamped on the synthesis ctx, audio persisted through the store seam, structured result for the model. - voice_deps.go: the wiring seam (VoiceSynth / VoiceStore + VoiceStoreFunc adapter), voiceEnabled, withSpeakSpec (appended after every schema shrink), and captureVoice (success beats failure, last one wins). - options.go / loop.go: VoiceSynth, VoiceStore, SpeakBudget injection points. - request.go: advertise speak on both toolSchemas return paths. - tools.go / tools_invoke.go: carry the batch voice outcome; split invokeOne / emitToolExecution / spanStatusForToolPhase / invokeParallel into their own file (tools.go was at the 500-line ceiling) and let speak run without a dispatcher. - run.go: keep the voice outcome across tool batches (a later batch without speak must not erase it). Known, deliberate: speak bypasses the capacity allow-set and skill-round narrowing (it belongs to no built-in toolset), so it is also advertised in quiet/awake; execution still passes the tool security gate, where speak is not registered and non-owner actors are denied fail-closed.	2026-10-27 21:22:14 +08:00

Author

SHA1

Message

Date

AskaEth

a0b4c941e4

fix(loop): keep cross-batch voice success and gate speak without dispatcher (M57)

两处 review 必修（review-m57 r1）：

P2-1 跨批覆盖：applyBatchVoice 原先无条件用后一批的语音结论覆盖
前一批，于是「先成功合成 abc.wav、下一批换个参数再试却失败」会把
Result.Voice 退化成 tts_failed —— 音频已落盘，退回文本既无收益也
违背那次表达。改为守 captureVoice 的同一条优先级：已落盘的音频不被
后续批次的失败抹掉；先降级后成功仍然覆盖（后到的成功即最终语音）。

P2-2 nil-dispatcher 安全门：speak 由循环自己承载、无 dispatcher 也
能执行（invokeOne 短路），但 applyToolAdmissions 在 dispatcher 缺席
时整段跳过安全门，使「没配 dispatcher」成为绕过 design/11 §3 的
旁路。新增 applyDispatcherlessVoiceGate：与策略引擎对未登记工具的
口径一致（默认关闭、只有 owner 放行），拒绝走同一条 not_allowed
软结果与 tool.reject.not_allowed 审计通道；并改正该处过时注释。

两条回归测试都做过改前红/改后绿：
- TestApplyBatchVoice_KeepsCrossBatchSuccess：注释掉守卫即报
  「voice = ... FallbackReason:tts_failed, want abc.wav」
- TestApplyToolAdmissions_DispatcherlessSpeakGate：去掉 else 分支即报
  「env group_member run = true, want false」

验收：go build ./... / go vet ./... / go test ./... 全绿。

2026-10-27 22:00:18 +08:00

AskaEth

c9da488749

refactor(loop): extract speak voice backfill into applyBatchVoice (M57)

2026-10-27 21:33:05 +08:00

AskaEth

1347a77d33

feat(loop): wire speak tool into the loop (M57)

- speak_tool.go: the speak tool itself — rune-derived timeout budget, structured
  failure codes, persona/emotion stamped on the synthesis ctx, audio persisted
  through the store seam, structured result for the model.
- voice_deps.go: the wiring seam (VoiceSynth / VoiceStore + VoiceStoreFunc
  adapter), voiceEnabled, withSpeakSpec (appended after every schema shrink),
  and captureVoice (success beats failure, last one wins).
- options.go / loop.go: VoiceSynth, VoiceStore, SpeakBudget injection points.
- request.go: advertise speak on both toolSchemas return paths.
- tools.go / tools_invoke.go: carry the batch voice outcome; split invokeOne /
  emitToolExecution / spanStatusForToolPhase / invokeParallel into their own
  file (tools.go was at the 500-line ceiling) and let speak run without a
  dispatcher.
- run.go: keep the voice outcome across tool batches (a later batch without
  speak must not erase it).

Known, deliberate: speak bypasses the capacity allow-set and skill-round
narrowing (it belongs to no built-in toolset), so it is also advertised in
quiet/awake; execution still passes the tool security gate, where speak is not
registered and non-owner actors are denied fail-closed.

2026-10-27 21:22:14 +08:00