Skip to content

feat: TTS/STT integration — speak tool, provider transcription, telegram voice replies - #259

Merged
jkyberneees merged 8 commits into
mainfrom
feat/tts-stt-integration
Sep 23, 2026
Merged

jkyberneees merged 8 commits into
mainfrom
feat/tts-stt-integration

Conversation

@jkyberneees

Copy link
Copy Markdown
Contributor

What

  • Bump go-llm-sdk v0.4.0 → v0.6.0 (adds SDK Speak/Transcribe)
  • New tts/stt operator config (project-level rejected, like vision), load-time validation with fallback semantics
  • New speak tool: provider-backed TTS, config-pinned voice/format, artifacts-dir output, metadata-only result
  • transcribe backend dispatch: stt.backend: "provider" routes to provider STT; local whisper byte-identical when unconfigured (default)
  • tts.telegram_voice_replies: additive voice-note replies on the Telegram surface via the existing media allowlist path
  • Both provider paths classified NetworkEgress (vision parity); ASR output untrusted-wrapped
  • Local-only e2e (env-gated) + unit tests; new-code coverage 85-100%

Verification

  • go build/vet ./... clean; go test cmd/odek, internal/config, internal/llmclient green
  • Real-provider roundtrip e2e passed locally (TTS 2.6s / STT 2.4s)
  • Adversarial review findings (2 major, 3 minor) all fixed with regression tests

@jkyberneees
jkyberneees merged commit 7b695c2 into main Sep 23, 2026
9 of 10 checks passed
@jkyberneees
jkyberneees deleted the feat/tts-stt-integration branch September 23, 2026 15:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant