Noise Reduction Integration
Developmentnoise-reduction-integrationNoise Reduction Integration is an Agent Skill for development published in the rapidaai/voice-ai repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Add or modify noise-reduction providers in assistant-api with denoiser factory wiring, packet safety, and UI option compatibility.
QUICK START
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/rapidaai/voice-ai/blob/HEAD/.claude/skills/noise-reduction-integration/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/noise-reduction-integration, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Noise Reduction Integration Skill
Mission
Integrate denoisers that suppress background noise while preserving speech quality, low latency, and packet ordering.
Inputs expected from user
- Runtime type: native library, SDK, or service API.
- Input audio assumptions: sample rate, channels, encoding.
- Tradeoff target: stronger suppression vs natural voice.
If user does not answer:
- Assume 16kHz LINEAR16 mono internal input.
- Keep current chunk cadence and low-latency defaults.
Hard boundaries
In scope:
api/assistant-api/internal/denoiser/internal/<provider>/...api/assistant-api/internal/denoiser/denoiser.go- optional compatibility updates in
api/assistant-api/internal/type/packet.go ui/src/providers/<provider>/noise.json
Out of scope:
- STT/TTS provider internals
- VAD/EOS internal algorithms
- telephony transport internals
Packet contract
Input:
DenoiseAudioPacket
Required outputs:
DenoisedAudioPacketwith stable context and chunk ordering- preserve downstream cadence expectations for VAD/STT/EOS
- on processing errors, fallback to original audio with
NoiseReduced=false
Optional diagnostics:
ConversationEventPacket{Name:"denoise"}for provider/debug context
Implementation workflow
- Pick baseline (
rn_noise,krisp). - Implement provider under
internal/<provider>/. - Register provider in
denoiser.gofactory switch. - Ensure lifecycle safety (
Close, context cancellation, cleanup). - Wire provider config/metadata and
noise.jsonwhere needed. - Add unit tests for selection/fallback/chunk integrity.
Done criteria
- Provider selectable by
microphone.denoising.provider. - No edits under VAD/EOS provider internals.
- Denoised packet behavior validated by tests.
Validation commands
go test ./api/assistant-api/internal/denoiser/...go test -run TestGetDenoiser ./api/assistant-api/internal/denoiser/...cd ui && yarn test providers./.claude/skills/noise-reduction-integration/scripts/validate.sh --check-diff --provider <provider>