Stackchan Alive v0.2.0

Stackchan: Alive v0.2.0

Commit: 996b7e4

This is the public v0.2.0 package for Stackchan: Alive, a character OS for Stackchan hardware. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and ships guarded autonomous motion in the production full firmware.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Production voice hashes are recorded in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md, data/voice_source_provenance.yaml, VOICE_SOURCE_STATUS.md, and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, hardware simulation, flashing, publishing, evidence capture, and package verification helpers are included under tools/.

Engine readiness quick check:

• Run tools/run_engine_probe.cmd -Json to check whether local model, STT, and TTS commands are configured.
• Run tools/run_engine_probe.cmd -RunModelSmoke -Json after exporting a runner command to capture the first real smoke result. This is setup evidence; full model selection still requires bridge/model_benchmark.py --require-runner with a passing summary.candidate_gate and recorded recommended_profile.
• Run python bridge/model_benchmark.py --profile gemma4-e2b-gguf --require-runner --json after the real runner is configured to write MODEL_BENCHMARK.md/json with candidate blockers, ready_profiles, and the fastest ready profile recommendation.
• Run tools/run_character_red_team.cmd -Json to regenerate the dry-run Character Lock red-team report. After a real runner is configured, run tools/run_character_red_team.cmd -RequireRunner -Json so the B7 gate is backed by an actual model instead of deterministic fallback responses.
• For the mobile/low-footprint brain path, configure STACKCHAN_LITERT_LM_COMMAND and use bridge/litert_lm_stackchan_wrapper.py as STACKCHAN_GEMMA4_E2B_LITERT_COMMAND before running the LiteRT-LM profile benchmark.

No-hardware simulation quick check:

• Run tools/run_prearrival_sim_check.cmd to create output/prearrival-sim/latest/PREARRIVAL_SIM_CHECK.md/json with the combined virtual CoreS3/LAN/audio proxy status, nested LAN smoke report, and engine-readiness status.
• After a real runner command is configured, run tools/run_prearrival_sim_check.cmd -RunModelBenchmark -Json to include nested model-benchmark/MODEL_BENCHMARK.md/json output and the model-benchmark-candidate gate in the same pre-arrival report.
• Run tools/run_lan_smoke.cmd to create output/lan-smoke/latest/LAN_SMOKE.md/json with a real local TCP/WebSocket bridge handshake, text turn, fake mic upload, fake STT, fake TTS, PCM16 binary downlink check, and visible thinking-latency timing while delayed speech is still running.
• Run tools/run_litert_lm_smoke.cmd to create output/litert-lm-smoke/latest/LITERT_LM_SMOKE.md/json with a deterministic two-layer LiteRT-LM wrapper contract check.
• Run tools/run_hardware_simulation.cmd to exercise the virtual Stackchan bridge proxy before the physical unit is available.
• The simulator proves bridge frame ordering, LAN text turns, fake mic PCM upload through fake STT, conversation timing, fake WAV TTS normalization to PCM16 downlink, speech-envelope handoff, binary TTS audio stream accounting, virtual CoreS3 input/display/speaker counters, offline command fallback, power-cycle recovery, bridge-kill recovery, and timeout failure behavior. It does not replace real hardware evidence.
• After an evidence packet has simulator output plus real display, speech, and bridge replay logs, run RUN_SIM_HARDWARE_COMPARE.cmd inside that packet to write advisory SIM_HARDWARE_COMPARE.md/json reports.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to review every included voice sample and production-voice artifact.
• Published releases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone quick previews.
• The exact production model.pth and model.index are included under media/voice/rvc/ and verified by SHA-256.
• Run tools/verify_tracked_rvc_assets.cmd to verify the bundled production RVC hashes.
• The exact active production model and index are included under media/voice/rvc.

Hardware validation is still required for each recipient unit. The repository's private paired reference robot
is tracked separately through exact-image evidence in docs/FIRST_DEPLOY_STATUS.md and
docs/ARRIVAL_DAY_RUNBOOK.md; that reference evidence does not validate another assembled unit,
power path, calibration, credentials, or local voice model.

Package and recipient gates:

1. Display-only flash and 10-minute idle run.
2. Speech-mouth demo evidence: logs/speech_mouth_demo_serial.log with streamed speech envelope commands, speech clear, and completion, plus logs/speak_all_intents_serial.log proving every packaged speech intent, earcon, and audio-output handoff.
3. Supervised servo-enable test.
4. Yaw classification and calibration.
5. 30-minute mixed idle/listen/speak soak.
6. Power-cycle recovery: USB power-cycle observation marked pass.
7. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
8. Verify the bundled production voice hashes.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.146-standalone-release-assets-manifest

Stackchan Alive v0.1.146-standalone-release-assets-manifest

Commit: 8f1640b

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.145-published-asset-manifest-check

Stackchan Alive v0.1.145-published-asset-manifest-check

Commit: c941be9

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.144-release-assets-manifest

Stackchan Alive v0.1.144-release-assets-manifest

Commit: 38e7107

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.143-release-asset-contract-gate

Stackchan Alive v0.1.143-release-asset-contract-gate

Commit: 3b63d68

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.142-release-asset-contract

Stackchan Alive v0.1.142-release-asset-contract

Commit: 565d01c

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.141-release-asset-allowlist

Stackchan Alive v0.1.141-release-asset-allowlist

Commit: d3bcfbc

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.140-release-asset-republish

Stackchan Alive v0.1.140-release-asset-republish

Commit: 135c4eb

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.139-publish-transient-cleanup

Stackchan Alive v0.1.139-publish-transient-cleanup

Commit: 2979f10

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.

Stackchan Alive v0.1.138-voice-audition-index

Stackchan Alive v0.1.138-voice-audition-index

Commit: 034ce55

This is a device-ready prerelease package. It is built, native-tested, compile-checked, includes preview media plus an expression QA sheet, and keeps servo output disabled by default.

Dependency provenance is recorded in DEPENDENCIES.md and dependency_lock.json, with copied build inputs under provenance/. Voice source provenance is staged in docs/VOICE_SOURCE_PROVENANCE_TEMPLATE.md and data/voice_source_provenance.yaml; voice approval status is summarized in VOICE_SOURCE_STATUS.md and voice_source_status.json. Readiness status is recorded in READINESS_REPORT.md and readiness_report.json. GitHub Actions status is recorded in GITHUB_ACTIONS_STATUS.md and github_actions_status.json. Preflight, flashing, manual publishing, evidence capture, evidence progress checking, hardware evidence verification, and package verification helpers are included under tools/.

Voice audition quick check:

• Run tools/open_voice_audition.cmd from the extracted package to open the local MP3 audition page.
• Run tools/open_voice_audition.cmd -All to generate one local page with both Stackchan Spark and RVC MP3 audition samples.
• Published prereleases upload stackchan_spark_audition_bright_robot_greeting.mp3 and stackchan_spark_thinking.mp3 as standalone release assets for one-click review.
• RVC review copies stackchan_rvc_bright_robot.mp3, stackchan_rvc_thinking_neutral.mp3, and stackchan_rvc_safety_neutral.mp3 are included in media/voice/rvc/ and uploaded as release assets for browser playback. Run tools/open_voice_audition.cmd -Rvc or open media/voice/rvc/RVC_AUDITION.html for the local RVC audition page.
• Run tools/verify_tracked_rvc_assets.cmd to verify the checked-in RVC MP3 review page without regenerating the full RVC WAV set.
• These are prototype voice-direction samples; consumer rollout still requires licensed or owned production voice-source provenance.

Hardware validation is still required before consumer rollout:

1. Display-only flash and 10-minute idle run.
2. Supervised servo-enable test.
3. Yaw classification and calibration.
4. 30-minute mixed idle/listen/speak soak.
5. USB power-cycle recovery test.
6. Target-speaker audio evidence: completed AUDIO_REVIEW.md plus a real-device speaker recording under audio/.
7. Licensed or owned production voice source.

See docs/DEVICE_BRINGUP.md and docs/PRODUCTION_READINESS.md.