ainqa-clinical-copilot

Clinical AI assistant for any web or Electron application. Provides chat, clinical document intelligence, and doctor-patient conversation capture. Powered by DeepInfra LLaMA.

Install

npm install ainqa-clinical-copilot ainqa-voice-assistant

Quick Start

import {
  CopilotProvider,
  CopilotChat,
  DocumentUploader,
  ConversationRecorder,
} from 'ainqa-clinical-copilot';

<CopilotProvider config={{
  deepInfraApiKey: 'YOUR_KEY',
  clinicalContext: { patientAge: 68, condition: 'AF' },
  onFieldsExtracted: (result) => console.log(result),
  onDataExtracted: (fields) => console.log(fields),
}}>
  <CopilotChat />
  <DocumentUploader />
  <ConversationRecorder />
</CopilotProvider>

Features

| Feature | Component | |---------|-----------| | Chat (streamed) | CopilotChat | | Document intelligence (PDF/image → structured JSON) | DocumentUploader | | Doctor-patient conversation capture | ConversationRecorder |

Theming

The package ships with a built-in light theme — drop CopilotProvider into any host application and the UI renders with a correct baseline look. No CSS variable configuration is required from the host.

Under the hood, CopilotProvider renders a root container element (.ainqa-copilot-root) with light-theme CSS custom properties applied as inline style:

--color-surface-base, --color-surface-raised, --color-surface-overlay,
--color-navy-900, --color-navy-800,
--color-text-primary, --color-text-secondary, --color-text-muted,
--color-border, --color-accent, --color-teal-500,
--color-clinical-red, --color-clinical-green, --color-clinical-amber

Custom theming

To override any variable, define it on a descendant element inside CopilotProvider (a closer ancestor wins the cascade for CSS custom properties). For example, to apply a dark theme:

<CopilotProvider config={{ ... }}>
  <div
    style={{
      '--color-surface-raised': '#1e293b',
      '--color-navy-900': '#0f172a',
      '--color-border': '#334155',
      '--color-text-primary': '#f1f5f9',
      '--color-text-secondary': '#94a3b8',
      '--color-text-muted': '#64748b',
    } as React.CSSProperties}
  >
    <CopilotChat />
  </div>
</CopilotProvider>

Document Pipeline

1. Format detection — native PDF / scanned PDF / image
2. Text extraction — PDF.js text layer or Tesseract.js OCR
3. Classification — LLaMA categorises into one of 8 clinical document types
4. Type-specific extraction — fields tailored to each document type
5. Structured output — typed JSON returned to host app