Template Generation

Overview

Template Generation is an AI-powered feature that helps you create and improve clinical note templates without manual authoring. Instead of building templates from scratch, you can describe what you need in natural language or provide an existing template for refinement. The feature operates in two modes:

Mode	Purpose
Generation	Create a new template from a natural language description
Refinement	Improve an existing template based on feedback

This is particularly useful when you need custom note formats for specific specialties or workflows but do not want to manually define every section, prompt, and property.

Alpha Feature: Template generation is currently in alpha. The API may change without notice and should not be used in production environments without careful consideration. See the Alpha API Authentication guide for access details.

Authentication

The Alpha API uses Bearer token authentication:

Authorization: Bearer YOUR_API_KEY

This differs from the stable API which uses the X-API-Key header. Use the same API key you obtained from your Sully.ai dashboard.

Generation Mode

Generation mode creates a template from a natural language description. Describe the template structure, sections, and formatting preferences, and the AI will produce a valid NoteTemplate object.

Endpoint

POST /alpha/note-templates

Request Body

Field	Type	Required	Description
`mode`	string	Yes	Set to `"generation"`
`description`	string	Yes	Natural language description of the template you want

Example Request

const response = await fetch('https://api.sully.ai/alpha/note-templates', {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${process.env.SULLY_API_KEY}`,
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    mode: 'generation',
    description: 'Create a template for a cardiology consultation with sections for cardiac history, current symptoms, physical exam findings, EKG interpretation, and recommendations'
  })
});

const { data } = await response.json();
console.log('Job ID:', data.id);
console.log('Status:', data.status);

Response

The endpoint returns a job object with a pending status:

{
  "data": {
    "id": "template_abc123def456",
    "mode": "generation",
    "status": "pending",
    "created_at": "2024-01-15T10:30:00Z",
    "updated_at": "2024-01-15T10:30:00Z"
  }
}

Refinement mode improves an existing template based on your feedback. Provide the current template and describe what changes you want, and the AI will produce an updated version.

Endpoint

POST /alpha/note-templates

Request Body

Field	Type	Required	Description
`mode`	string	Yes	Set to `"refinement"`
`template`	object	Yes	The existing NoteTemplate to refine
`feedback`	string	Yes	Description of improvements you want

Example Request

const existingTemplate = {
  id: 'cardiology-consult-v1',
  title: 'Cardiology Consultation',
  global_prompt: 'Generate a cardiology consultation note.',
  sections: [
    {
      id: 'cardiac-history',
      type: 'heading',
      properties: { level: 1, text: 'Cardiac History' },
      children: [
        {
          id: 'history-text',
          type: 'text',
          prompt: 'Document cardiac history.',
          properties: { detail_level: 'standard', tone: 'formal' }
        }
      ]
    },
    {
      id: 'assessment',
      type: 'heading',
      properties: { level: 1, text: 'Assessment' },
      children: [
        {
          id: 'assessment-text',
          type: 'text',
          prompt: 'Provide clinical assessment.',
          properties: { detail_level: 'standard', tone: 'formal' }
        }
      ]
    }
  ]
};

const response = await fetch('https://api.sully.ai/alpha/note-templates', {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${process.env.SULLY_API_KEY}`,
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    mode: 'refinement',
    template: existingTemplate,
    feedback: 'Make the assessment section more detailed and add a section for follow-up instructions'
  })
});

const { data } = await response.json();
console.log('Job ID:', data.id);

Polling for Results

Template generation is asynchronous. After creating a job, poll the GET endpoint to check status and retrieve results.

Endpoint

GET /alpha/note-templates/{id}

Status Flow

pending → processing → completed | failed

Status	Description
`pending`	Job is queued for processing
`processing`	Currently generating or refining the template
`completed`	Success - `result.template` contains the generated NoteTemplate
`failed`	Error occurred - `result.error` contains the error message

Polling Pattern

async function waitForTemplate(jobId: string): Promise<any> {
  const maxAttempts = 30;
  const pollInterval = 2000; // 2 seconds

  for (let attempt = 0; attempt < maxAttempts; attempt++) {
    const response = await fetch(
      `https://api.sully.ai/alpha/note-templates/${jobId}`,
      {
        headers: { 'Authorization': `Bearer ${process.env.SULLY_API_KEY}` }
      }
    );

    const { data } = await response.json();

    if (data.status === 'completed') {
      return data.result.template;
    }

    if (data.status === 'failed') {
      throw new Error(`Template generation failed: ${data.result?.error || 'Unknown error'}`);
    }

    // Wait before next poll
    await new Promise(resolve => setTimeout(resolve, pollInterval));
  }

  throw new Error('Template generation timed out');
}

// Usage
const template = await waitForTemplate('template_abc123def456');
console.log('Generated template:', template);

Understanding the Output

When a job completes successfully, result.template contains a valid NoteTemplate object that can be used directly with the POST /v1/notes endpoint.

Template Structure

Field	Type	Description
`id`	string	Unique identifier for the template
`title`	string	Display title
`global_prompt`	string	Instructions that apply to all sections
`sections`	array	Ordered list of HeadingSection, TextSection, or ListSection objects

Example Output

{
  "data": {
    "id": "template_abc123def456",
    "mode": "generation",
    "status": "completed",
    "created_at": "2024-01-15T10:30:00Z",
    "updated_at": "2024-01-15T10:32:00Z",
    "result": {
      "template": {
        "id": "cardiology-consultation-generated",
        "title": "Cardiology Consultation Template",
        "global_prompt": "Generate a comprehensive cardiology consultation note with emphasis on cardiovascular assessment and risk stratification.",
        "sections": [
          {
            "id": "cardiac-history",
            "type": "heading",
            "properties": {
              "level": 1,
              "text": "Cardiac History",
              "bold": true
            },
            "children": [
              {
                "id": "cardiac-history-text",
                "type": "text",
                "prompt": "Document relevant cardiac history including prior diagnoses, procedures, and interventions.",
                "properties": {
                  "detail_level": "detailed",
                  "tone": "technical",
                  "formatting_style": "markdown"
                }
              }
            ]
          },
          {
            "id": "current-symptoms",
            "type": "heading",
            "properties": {
              "level": 1,
              "text": "Current Symptoms",
              "bold": true
            },
            "children": [
              {
                "id": "symptoms-text",
                "type": "text",
                "prompt": "Describe presenting cardiac symptoms including chest pain characteristics, dyspnea, palpitations, syncope, and edema.",
                "properties": {
                  "detail_level": "exhaustive",
                  "tone": "technical"
                }
              }
            ]
          },
          {
            "id": "physical-exam",
            "type": "heading",
            "properties": {
              "level": 1,
              "text": "Physical Examination",
              "bold": true
            },
            "children": [
              {
                "id": "vitals",
                "type": "text",
                "prompt": "Document vital signs including blood pressure (both arms if applicable), heart rate, and rhythm.",
                "properties": {
                  "detail_level": "standard",
                  "tone": "technical",
                  "label": "Vitals:",
                  "bold_label": true
                }
              },
              {
                "id": "cardiac-exam",
                "type": "text",
                "prompt": "Document cardiac examination findings: JVP, carotid examination, PMI, heart sounds, murmurs, and gallops.",
                "properties": {
                  "detail_level": "detailed",
                  "tone": "technical"
                }
              }
            ]
          },
          {
            "id": "ekg-interpretation",
            "type": "heading",
            "properties": {
              "level": 1,
              "text": "EKG Interpretation",
              "bold": true
            },
            "children": [
              {
                "id": "ekg-findings",
                "type": "text",
                "prompt": "Interpret EKG findings including rate, rhythm, intervals, axis, and any abnormalities.",
                "properties": {
                  "detail_level": "detailed",
                  "tone": "technical",
                  "hideIfEmpty": true
                }
              }
            ]
          },
          {
            "id": "recommendations",
            "type": "heading",
            "properties": {
              "level": 1,
              "text": "Recommendations",
              "bold": true
            },
            "children": [
              {
                "id": "recommendations-list",
                "type": "list",
                "prompt": "List specific recommendations including further testing, medications, lifestyle modifications, and follow-up plan.",
                "properties": {
                  "list_type": "numeric",
                  "detail_level": "detailed",
                  "tone": "instructional"
                }
              }
            ]
          }
        ]
      }
    }
  }
}

Using the Generated Template

Once you have the template, use it with the notes endpoint:

const noteResponse = await sully.notes.create({
  transcript: transcription,
  noteType: {
    type: 'note_template',
    template: generatedTemplate
  }
});

Iteration Workflow

The recommended workflow for building production-ready templates:

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Generate   │────▶│    Test     │────▶│   Refine    │
│  (describe) │     │ (with real  │     │ (feedback)  │
└─────────────┘     │  transcripts)│     └──────┬──────┘
                    └─────────────┘            │
                           ▲                   │
                           │                   │
                           └───────────────────┘
                                   │
                                   ▼
                            ┌─────────────┐
                            │    Done     │
                            │ (production)│
                            └─────────────┘

Step-by-Step

Generate: Start with a description of the template you need
Test: Use the generated template with actual transcriptions
Review: Evaluate the note output for missing sections or formatting issues
Refine: Use refinement mode with specific feedback
Repeat: Continue testing and refining until satisfied
Deploy: Use the final template in production

Example Iteration

// Step 1: Initial generation
const genResponse = await createTemplateJob({
  mode: 'generation',
  description: 'Orthopedic consultation template with injury history, imaging review, and surgical recommendations'
});
const initialTemplate = await waitForTemplate(genResponse.data.id);

// Step 2: Test with real data
const testNote = await sully.notes.create({
  transcript: sampleTranscript,
  noteType: { type: 'note_template', template: initialTemplate }
});

// Step 3: Identify issues (e.g., missing physical exam detail)
// Step 4: Refine
const refineResponse = await createTemplateJob({
  mode: 'refinement',
  template: initialTemplate,
  feedback: 'Add a detailed physical examination section with range of motion, strength testing, and special tests. Make the surgical recommendations section use a numbered list format.'
});
const refinedTemplate = await waitForTemplate(refineResponse.data.id);

// Step 5: Test again and repeat if needed

Best Practices

Writing Effective Descriptions (Generation Mode)

Do	Do Not
Be specific about sections	Use vague terms like “good template”
Name sections explicitly	Assume AI knows your preferences
Mention detail levels	Leave formatting to chance
Specify formatting preferences	Skip specialty context

Good description:

“Create a dermatology consultation template with sections for: skin lesion description (using ABCDE criteria), location and distribution, relevant history, differential diagnosis with at least 3 possibilities, biopsy recommendations, and follow-up plan. Use bullet points for differentials and numbered items for the plan.”

Poor description:

“Make a skin doctor template”

Do	Do Not
Reference specific sections	Give vague feedback
Describe exact changes	Say “make it better”
Specify formatting changes	Assume context is understood
Explain the reasoning	Provide contradictory instructions

Good feedback:

“Change the Assessment section prompt to include severity scoring on a 1-10 scale. Add a ‘Red Flags’ subsection under History that specifically asks about warning symptoms. Convert the Plan section from paragraph format to a numbered list with categories: Diagnostics, Medications, Referrals, Follow-up.”

Poor feedback:

“Make the assessment better and add more sections”

General Guidelines

Review before production: Always test generated templates with real transcriptions before deploying
Iterate incrementally: Make small, focused refinements rather than large changes
Document your templates: Keep notes on what each template is designed for
Version control: Track template versions to revert if needed
Combine with manual edits: You can manually adjust the generated template JSON before use

Note Template Schema

Complete schema reference for NoteTemplate objects

Note Customization Guide

Deep dive into note styles and templates

Create Note Template Job

Full API reference for the create endpoint

Alpha Authentication

Authentication details for alpha endpoints

Overview

Alpha Guides

Alpha API Reference

Template Generation

Overview

Authentication

Generation Mode

Endpoint

Request Body

Example Request

Response

Refinement Mode

Endpoint

Request Body

Example Request

Polling for Results

Endpoint

Status Flow

Polling Pattern

Understanding the Output

Template Structure

Example Output

Using the Generated Template

Iteration Workflow

Step-by-Step

Example Iteration

Best Practices

Writing Effective Descriptions (Generation Mode)

Writing Effective Feedback (Refinement Mode)

General Guidelines

Note Template Schema

Note Customization Guide

Create Note Template Job

Alpha Authentication

Overview

Alpha Guides

Alpha API Reference

​Overview

​Authentication

​Generation Mode

​Endpoint

​Request Body

​Example Request

​Response

​Refinement Mode

​Endpoint

​Request Body

​Example Request

​Polling for Results

​Endpoint

​Status Flow

​Polling Pattern

​Understanding the Output

​Template Structure

​Example Output

​Using the Generated Template

​Iteration Workflow

​Step-by-Step

​Example Iteration

​Best Practices

​Writing Effective Descriptions (Generation Mode)

​Writing Effective Feedback (Refinement Mode)

​General Guidelines

​Related Resources

Note Template Schema

Note Customization Guide

Create Note Template Job

Alpha Authentication

Overview

Authentication

Generation Mode

Endpoint

Request Body

Example Request

Response

Refinement Mode

Endpoint

Request Body

Example Request

Polling for Results

Endpoint

Status Flow

Polling Pattern

Understanding the Output

Template Structure

Example Output

Using the Generated Template

Iteration Workflow

Step-by-Step

Example Iteration

Best Practices

Writing Effective Descriptions (Generation Mode)

Writing Effective Feedback (Refinement Mode)

General Guidelines

Related Resources