Deepgram Skill Pack

Speech-to-text, text-to-speech, and audio intelligence for Claude Code. 24 skills covering the full Deepgram platform: Nova-3/Nova-2 transcription, Aura-2 TTS, live streaming WebSocket, diarization, summarization, and enterprise deployment.

SDK: @deepgram/sdk (TypeScript) / deepgram-sdk (Python) API: createClient() (v3/v4) or new DeepgramClient() (v5) Models: Nova-3 (best accuracy), Nova-2 (proven), Base (fastest), Whisper (multilingual)

Installation

/plugin install deepgram-pack@claude-code-plugins-plus

Skills

Standard (S01-S12)

| # | Skill | What It Does | |---|-------|-------------| | S01 | deepgram-install-auth | Install SDK, configure API key, verify connection with getProjects() | | S02 | deepgram-hello-world | Transcribe audio URL in 5 lines — transcribeUrl with Nova-3 | | S03 | deepgram-local-dev-loop | Test fixtures, Vitest mocks, watch mode, integration tests | | S04 | deepgram-sdk-patterns | Singleton client, TTS with Aura-2, audio intelligence pipeline, Python patterns | | S05 | deepgram-core-workflow-a | Pre-recorded STT: file/URL transcription, diarization, batch with p-limit, callbacks | | S06 | deepgram-core-workflow-b | Live streaming: WebSocket, Sox mic capture, interim results, auto-reconnect, SSE | | S07 | deepgram-common-errors | HTTP error reference, WebSocket debugging, audio format validation, retry patterns | | S08 | deepgram-debug-bundle | Collect env/connectivity/audio diagnostics into sanitized support bundle | | S09 | deepgram-rate-limits | Concurrency-based limiting with p-limit, exponential backoff, circuit breaker | | S10 | deepgram-security-basics | Scoped keys, PII redaction (redact), temp keys for browsers, SSRF prevention | | S11 | deepgram-prod-checklist | Health check, Prometheus metrics, AlertManager rules, go-live timeline | | S12 | deepgram-upgrade-migration | SDK v3->v5 migration map, Nova-2->Nova-3 A/B testing, validation suite |

Pro (P13-P18)

| # | Skill | What It Does | |---|-------|-------------| | P13 | deepgram-ci-integration | GitHub Actions workflow, integration tests, smoke test script, key rotation | | P14 | deepgram-deploy-integration | Docker, Kubernetes (HPA + secrets), AWS Lambda (S3 trigger), Cloud Run | | P15 | deepgram-webhooks-events | Callback URL async transcription, signature verification, Redis job tracking | | P16 | deepgram-performance-tuning | ffmpeg preprocessing, model selection, streaming large files, caching | | P17 | deepgram-cost-tuning | Budget guardrails, silence removal savings, usage API dashboard, feature costs | | P18 | deepgram-reference-architecture | Sync REST, BullMQ queue, WebSocket proxy, hybrid router |

Flagship (F19-F24)

| # | Skill | What It Does | |---|-------|-------------| | F19 | deepgram-multi-env-setup | Typed config per environment, client factory, Docker profiles, K8s overlays | | F20 | deepgram-observability | Prometheus metrics, OpenTelemetry tracing, Pino logging, Grafana dashboards | | F21 | deepgram-incident-runbook | Automated triage script, SEV1-4 response, fallback queue, post-incident template | | F22 | deepgram-data-handling | PII redaction, S3 encrypted upload, retention policies, GDPR erasure | | F23 | deepgram-enterprise-rbac | 5-role model, scoped key provisioning, permission middleware, team management | | F24 | deepgram-migration-deep-dive | Adapter pattern, AWS/Google/Whisper migration, traffic shifting, validation |

Quick Start

import { createClient } from '@deepgram/sdk';

const deepgram = createClient(process.env.DEEPGRAM_API_KEY!);

// Pre-recorded transcription
const { result } = await deepgram.listen.prerecorded.transcribeUrl(
  { url: 'https://static.deepgram.com/examples/Bueller-Life-moves-702702706.wav' },
  { model: 'nova-3', smart_format: true, diarize: true }
);
console.log(result.results.channels[0].alternatives[0].transcript);

// Text-to-speech
const response = await deepgram.speak.request(
  { text: 'Hello from Deepgram.' },
  { model: 'aura-2-thalia-en' }
);

Deepgram Platform

| Product | API | Models | |---------|-----|--------| | Speech-to-Text | listen.prerecorded / listen.live | Nova-3, Nova-2, Base, Whisper | | Text-to-Speech | speak.request | Aura-2 (thalia, asteria, orion, luna, helios) | | Audio Intelligence | STT params | Summarize, Topics, Sentiment, Intent | | PII Redaction | redact: ['pci','ssn'] | Built into STT pipeline |

License

MIT