Add Ambient Clinical Documentation Medical Scribe

[sazed-bot]

Source URLs

• Candidate source: https://github.com/sampleapp-ai/assemblyai-medical-scribe-pipecat
• GitHub source: https://github.com/sampleapp-ai/assemblyai-medical-scribe-pipecat

Assessor score

• Score: 5
• Version: assessor_llm_v2

Synopsis

Pipecat + AssemblyAI Universal Streaming template for an ambient clinical documentation scribe: passively transcribes doctor-patient conversations with medical keyterm boosting, post-processes each turn through AssemblyAI's LLM Gateway (Claude Haiku 4.5) for terminology correction, and auto-generates structured SOAP notes at session end. Ships with Python pipeline, Next.js client, Dockerfile, and tuned VAD config (800ms stop_secs) for natural clinical pauses.

Reasoning

Clears all disqualifiers: public repo with concrete artifacts, named healthcare/medical-scribe outcome, multi-component stack (STT+LLM+orchestration), not a single-vendor hello-world. The README is highly schema-able with explicit components, configuration parameters, and a named model string. Forkable as a clone-and-run repo. The passive-listening scribe with SOAP-note generation is a specific, named clinical use case — exactly the shape SpeechStack wants. Web-only transport (no telephony/TTS) is appropriate for an ambient in-room scribe.

Claude raw output

{
  "id": "medical-scribe-ambient-clinical-documentation",
  "slug": "medical-scribe-ambient-clinical-documentation",
  "tags": [
    "medical-scribe",
    "ambient-documentation",
    "soap-notes",
    "clinical-workflow",
    "passive-listening",
    "medical-terminology",
    "real-time-transcription"
  ],
  "title": "Ambient Clinical Documentation Medical Scribe",
  "config": {
    "vad": {
      "stop_secs": 0.8,
      "confidence": 0.6,
      "min_volume": 0.4
    },
    "mode": "passive_listening",
    "output_format": "SOAP_notes",
    "medical_keyterms": [
      "hypertension",
      "diabetes mellitus",
      "coronary artery disease",
      "metformin 1000mg",
      "lisinopril 10mg",
      "atorvastatin 20mg",
      "chief complaint",
      "history of present illness",
      "review of systems",
      "auscultation",
      "palpation",
      "echocardiogram",
      "hemoglobin A1c",
      "blood pressure",
      "oxygen saturation"
    ],
    "llm_gateway_model": "claude-haiku-4-5-20251001",
    "silence_threshold_ms": 800
  },
  "source": "github_discovery",
  "license": "Proprietary",
  "metrics": {
    "cost_display": null,
    "latency_p50_ms": 300,
    "latency_p95_ms": null,
    "latency_display": "Sub-300ms streaming transcription",
    "cost_per_minute_usd_max": null,
    "cost_per_minute_usd_min": null
  },
  "demo_url": null,
  "industry": "healthcare",
  "pipeline": {
    "llm": "Claude Haiku 4.5",
    "stt": "AssemblyAI Universal-2",
    "tts": "Cartesia Sonic Turbo",
    "telephony": "Web Only"
  },
  "use_case": "transcription-and-summary",
  "framework": "Pipecat",
  "languages": [
    "en-US"
  ],
  "created_at": "2025-05-17T10:30:00-07:00",
  "raw_prompt": null,
  "updated_at": "2025-05-17T10:30:00-07:00",
  "contributor": {
    "github": "seanflyyy"
  },
  "description": "A passive voice AI agent that listens to doctor-patient conversations in real time, transcribes speech with medical terminology boosting, and automatically generates structured SOAP notes at session end without interrupting clinical workflow.",
  "prompt_file": null,
  "github_source_url": "https://github.com/sampleapp-ai/assemblyai-medical-scribe-pipecat"
}

Dedup decision

F6 dedup has not landed yet; no dedup decision available.

Reviewer notes

approve-to-pr running: