Add sandbox-only `aai speak` streaming TTS command#61

Merged

alexkroman merged 28 commits into

mainfrom

add-speak-command

Jun 11, 2026

Commits on Jun 10, 2026

docs(speak): design for sandbox-only streaming TTS command

alexkroman-assembly
and
claude
committed
docs(speak): implementation plan for the speak command

alexkroman-assembly
and
claude
committed
feat(speak): add streaming_tts_host to Environment

alexkroman-assembly
and
claude
committed
feat(speak): tts session availability + url building

alexkroman-assembly
and
claude
committed
feat(speak): tts websocket synthesize protocol

alexkroman-assembly
and
claude
committed
feat(speak): tts wav writing + pcm playback

alexkroman-assembly
and
claude
committed
feat(speak): add sandbox-only speak command

alexkroman-assembly
and
claude
committed
test(speak): refresh CLI help snapshots

alexkroman-assembly
and
claude
committed
docs(speak): document the tts subsystem and speak command

alexkroman-assembly
and
claude
committed
test(speak): cover default audio/ws factories + help order

alexkroman-assembly
and
claude
committed
test(speak): kill surviving mutants on changed lines

alexkroman-assembly
and
claude
committed
fix(speak): default voice/language so the bare command works

alexkroman-assembly
and
claude
committed
test(speak): satisfy strict mypy/pyright on tts tests

alexkroman-assembly
and
claude
committed
fix(speak): match the real streaming-TTS protocol

alexkroman-assembly
and
claude
committed
fix(speak): cancel playback immediately on Ctrl-C

alexkroman-assembly
and
claude
committed
docs(speak): design for speaker-aware multi-voice playback

alexkroman-assembly
and
claude
committed
docs(speak): implementation plan for speaker-aware multi-voice playback

alexkroman-assembly
and
claude
committed
feat(speak): parse speaker-labeled transcript text into segments
alexkroman-assembly
committed
test(speak): cover label-only continuation; drop no-op strip

alexkroman-assembly
and
claude
committed
feat(speak): resolve speaker voices via rotation + overrides

alexkroman-assembly
and
claude
committed
test(speak): make voice-rotation wrap test self-sufficient; parametrize malformed-pair
alexkroman-assembly
committed
feat(speak): add silence() PCM helper for inter-turn gaps
alexkroman-assembly
committed
feat(speak): synthesize_dialogue concatenates per-voice segments
alexkroman-assembly
committed
test(speak): pin dialogue duration, server rate, and empty-segments edge
alexkroman-assembly
committed
feat(speak): auto-detect diarized input and voice each speaker

alexkroman-assembly
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add sandbox-only `aai speak` streaming TTS command#61

Add sandbox-only `aai speak` streaming TTS command#61
alexkroman merged 28 commits into
mainfrom
add-speak-command

Commits on Jun 10, 2026

docs(speak): design for sandbox-only streaming TTS command

docs(speak): implementation plan for the speak command

feat(speak): add streaming_tts_host to Environment

feat(speak): tts session availability + url building

feat(speak): tts websocket synthesize protocol

feat(speak): tts wav writing + pcm playback

feat(speak): add sandbox-only speak command

test(speak): refresh CLI help snapshots

docs(speak): document the tts subsystem and speak command

test(speak): cover default audio/ws factories + help order

test(speak): kill surviving mutants on changed lines

fix(speak): default voice/language so the bare command works

test(speak): satisfy strict mypy/pyright on tts tests

fix(speak): match the real streaming-TTS protocol

fix(speak): cancel playback immediately on Ctrl-C

docs(speak): design for speaker-aware multi-voice playback

docs(speak): implementation plan for speaker-aware multi-voice playback

feat(speak): parse speaker-labeled transcript text into segments

test(speak): cover label-only continuation; drop no-op strip

feat(speak): resolve speaker voices via rotation + overrides

test(speak): make voice-rotation wrap test self-sufficient; parametrize malformed-pair

feat(speak): add silence() PCM helper for inter-turn gaps

feat(speak): synthesize_dialogue concatenates per-voice segments

test(speak): pin dialogue duration, server rate, and empty-segments edge

feat(speak): auto-detect diarized input and voice each speaker

Commits on Jun 11, 2026

test(speak): cover empty-labels path; pin voice fallback, speaker count, multi duration

test(speak): kill mutation survivors (Segment frozen, multi --out path); fix dialogue test typing

docs(speak): clarify _read_text stdin-fallback is intentional (omitted arg only)