📋 Quick Navigation

🚀 Getting Started - Your First Material

The quickest way to create your first therapy material:

  1. Go to the Create page
  2. Click any template from the Prompt Library sidebar (e.g., "Simple Reading Passage")
  3. The prompt appears in the text box - you can edit it if you want
  4. Click "Generate Material" button
  5. Wait a moment while AI creates your material
  6. Review the output, then copy, print, or download it
💡 That's it! You just created a personalized therapy material in under a minute. Now let's explore how to make them even better with patient profiles.

📊 Complete Workflow Guide

For best results, follow this recommended workflow:

Step 1: Create a Patient Profile (Recommended)

  1. Go to Patients page
  2. Click "New Patient"
  3. Fill in:
    • Name/Initials: Use initials for privacy (e.g., "J.S.")
    • Aphasia Type: Broca's, Wernicke's, Global, etc.
    • Reading Level: Elementary, Middle School, High School, or College
    • Background: Former occupation (e.g., "Retired teacher")
    • Interests: Hobbies and passions (e.g., "gardening, golf, cooking")
  4. Click "Save Patient Profile"

Step 2: Select Patient & Generate Material

  1. From Patients page, click "Use for Generation" on the patient card
  2. You're taken to Create page with patient context pre-loaded
  3. Browse Prompt Library and select a template
  4. Review the auto-populated prompt (includes patient details)
  5. Edit the prompt to specify the exact activity you want
  6. Click "Generate Material"

Step 3: Use & Save the Material

  1. Review the generated material for clinical appropriateness
  2. Use action buttons:
    • 📋 Copy: Copy to clipboard for pasting elsewhere
    • đŸ–¨ī¸ Print: Print directly for therapy session
    • 💾 Download: Save as .txt file to your computer
  3. Material is automatically saved to your Library (up to 100 recent)
  4. Find it later in Library page

✅ Common Tasks - How Do I...?

How do I save a custom prompt I use frequently?

  1. Type or edit your prompt in the text box on Create page
  2. Click "💾 Save as Custom" button
  3. Give it a descriptive title
  4. Your custom prompt now appears in the "Custom" tab of Prompt Library

How do I edit or delete a custom prompt?

  1. On Create page, click "âœī¸ Manage" button next to Prompt Library
  2. You'll see all your custom prompts
  3. Click âœī¸ to edit or đŸ—‘ī¸ to delete each prompt

How do I find materials I generated yesterday?

  1. Go to Library page
  2. Use the date filter: select "Last 7 Days" or "Last 30 Days"
  3. Use search box to search by patient name or content
  4. Click on any material to view, edit, or re-download it

How do I backup my patient data and materials?

💡 Backup Reminder: When you create or delete any data (materials, patients, prompts), a red pulsing dot appears on the Settings nav link to remind you to backup.
  1. Watch for the red dot badge on "âš™ī¸ Settings" in the navigation - it appears when backup is needed
  2. Go to Settings page
  3. Click "💾 Export All Data" to download everything as JSON
  4. Save the file to a secure location on your computer
  5. The red dot disappears after successful backup

Note: All export/import is now centralized in Settings for consistency. The red badge helps you remember to backup after any changes.

How do I change which patient I'm generating for?

  1. On Create page, if a patient is selected, click "✕ Clear" in the patient selector box
  2. Go to Patients page and select a different patient
  3. Or manually edit the prompt to change patient details

âœī¸ Writing Effective Prompts

The quality of your generated materials depends heavily on your prompt. Here's how to write prompts that produce excellent results:

Key Elements of a Good Prompt

  • Aphasia type: Specify Broca's, Wernicke's, Global, Anomic, etc.
  • Reading level: Elementary, Middle School, High School, or College
  • Material type: Reading passage, naming exercises, conversation prompts, etc.
  • Topic/Theme: Connect to patient's interests and background
  • Format preferences: Length, structure, visual elements

Example: Good vs. Bad Prompts

❌ BAD - Too Vague
Create a reading exercise.
✅ GOOD - Specific & Detailed
Create a reading passage for a patient with Broca's aphasia at a middle school reading level. The patient is a retired engineer who loves woodworking. Write a 150-word passage about building a birdhouse, using simple sentences (5-8 words each) and concrete language. Include 5 comprehension questions at the end.
❌ BAD - No Context
Make naming exercises.
✅ GOOD - Clear Requirements
Create a naming exercise for a patient with anomic aphasia. The patient was a chef and loves cooking. Generate 20 items in a semantic field related to kitchen tools and cooking techniques. Provide each item as a fill-in-the-blank sentence with a context clue. Organize by difficulty level.

Pro Tips for Better Prompts

  • Be specific about difficulty: "Simple 4-5 word sentences" vs. "easy reading"
  • Request formatting: "Number each item" or "Use bullet points"
  • Specify quantity: "10 questions" not "some questions"
  • Include purpose: "For practicing verb conjugation" helps AI understand the goal
  • Use patient interests: Gardening, sports, music, cooking, etc. increase engagement

🔒 PHI Detection & Privacy Protection

âš ī¸ Your Responsibility

You are ultimately responsible for protecting patient privacy. The PHI detection system is a defensive tool that helps catch common mistakes, but it is not foolproof. Always review your prompts carefully before generation.

What is PHI Detection?

The system automatically scans prompts for Protected Health Information (PHI) before sending them to the AI. When PHI is detected, you'll be notified and can review an anonymized version.

What Gets Detected

✅ PHI That Will Be Detected:

  • Names: Patient names in any format (John Smith, john smith, JOHN SMITH)
  • Contact Info: Email addresses, phone numbers
  • Addresses: Street addresses, cities with street numbers
  • Dates: Specific dates, dates of birth
  • Identifiers: Canadian SIN, passport numbers, medical record numbers

✅ Medical Terms That Won't Be Flagged:

The system is smart enough to NOT flag medical and therapy terminology:

  • Aphasia types: Broca's aphasia, Wernicke's aphasia, global aphasia, etc.
  • Speech disorders: Apraxia, dysarthria, stuttering, etc.
  • Therapy terms: Speech therapy, treatment plan, therapy session, etc.
  • Educational terms: Middle school, reading level, comprehension, etc.
  • Neurological conditions: Stroke, TBI, Parkinson's disease, etc.

How to Use the PHI Test Feature

  1. Write your prompt on the Create page
  2. Click "🔍 Test for PHI" button (optional but recommended)
  3. Review the results:
    • No PHI detected: Safe to generate
    • PHI detected: Review the anonymized version
  4. If PHI was found, you have two options:
    • Option 1: Click "Use Corrected Prompt" to replace your prompt with the anonymized version
    • Option 2: Click "Cancel" and manually remove the PHI yourself
  5. Proceed with generation using your PHI-free prompt

Understanding PHI Placeholders

What are [PERSON_NAME], [EMAIL_ADDRESS], etc.?

When PHI is detected, the system replaces it with placeholders like [PERSON_NAME] or [EMAIL_ADDRESS]. These placeholders:

  • Show you exactly what type of PHI was detected
  • Can be used directly in your prompt (the AI understands generic references)
  • Protect privacy by removing specific identifiers

Example:

❌ Original (contains PHI):
"Create exercises for Mary Johnson with Broca's aphasia"
✅ Anonymized (safe to use):
"Create exercises for [PERSON_NAME] with Broca's aphasia"

The AI will generate materials using generic references like "the patient" instead of the specific name.

Best Practices for Privacy

  • Use initials in patient profiles: Store "J.S." locally, never include in prompts sent to AI
  • Test before generating: Use the "Test for PHI" button on important prompts
  • Review manually: Don't rely solely on automated detection - review prompts yourself
  • Generic references work fine: "A 65-year-old patient" is better than "John Smith, age 65"
  • When in doubt, leave it out: If information isn't essential for material generation, don't include it

System Limitations

Important: The PHI detection system may:

  • Miss some PHI: Unusual formats or misspellings may not be detected
  • Flag false positives: Sometimes non-PHI may be flagged (you can proceed if it's safe)
  • Not catch context clues: Indirect identifiers may not be detected

This system is a safety net, not a guarantee. You must still review prompts carefully.

What Happens Behind the Scenes

Technical Details (for transparency):

  • The system uses Google Cloud DLP (Data Loss Prevention) for primary detection
  • A local fallback detector is used if DLP is unavailable
  • Custom exclusion rules prevent medical terminology from being flagged
  • Detection happens before any data is sent to the AI
  • The system is continuously improved based on feedback

For operations staff: See PHI_DETECTION_OPERATIONS_GUIDE.md for maintenance procedures.

đŸ‘Ĩ Patient Profile Best Practices

Why Create Patient Profiles?

Patient profiles enable automatic personalization. When you select a patient, their details are incorporated into prompts, creating materials that are:

  • More engaging (topics they care about)
  • Appropriately challenging (right reading level and aphasia type)
  • Culturally relevant (reflecting their background and life experience)

What to Include in Profiles

Name/Initials

Best Practice: Use initials only (e.g., "J.S." or "M.R.") for HIPAA compliance.

Avoid: Full names if data is stored on shared or unsecured devices.

Background & Former Occupation

Good Examples:

  • "Retired high school math teacher, taught for 35 years"
  • "Former software engineer at tech startup"
  • "Stay-at-home parent, previously worked as a nurse"
  • "Owned a small construction business"

Why it matters: AI can create materials with relevant vocabulary and scenarios.

Interests & Hobbies

Good Examples:

  • "Gardening (roses especially), birdwatching, cooking Italian food"
  • "Golf, watching football, reading mystery novels"
  • "Knitting, classical music, playing with grandchildren"
  • "Woodworking, fishing, restoring vintage cars"

Tip: The more specific, the better! "Roses" is better than "flowers."

Reading Level Selection Guide

  • Elementary (Grades 3-5): Very simple sentences, common words, lots of repetition
  • Middle School (Grades 6-8): Simple to moderate sentences, everyday vocabulary
  • High School (Grades 9-12): More complex sentences, broader vocabulary
  • College Level: Complex sentences, academic or professional vocabulary

Note: Start conservative and adjust based on patient performance.

🔧 Troubleshooting

Problem: Generated material quality is poor or off-topic

Solutions:

  • Make your prompt more specific (see "Writing Effective Prompts" above)
  • Include more details about difficulty level and format
  • Try a different template as a starting point
  • Regenerate - sometimes trying again produces better results

Problem: Generation failed or shows an error

Solutions:

  • Check your internet connection
  • Wait a moment and try again (temporary API issue)
  • If error says "API key not configured" - contact the administrator
  • Refresh the page and try again

Problem: I lost my patient data or saved materials

Causes & Solutions:

  • Used different browser: Data is stored per browser. Stick to one browser or export/import data.
  • Cleared browser data: This deletes localStorage. Always export backups regularly!
  • Used incognito/private mode: Data is deleted when you close the window. Use normal browsing mode.
  • Browser storage limit reached: Export old data, then delete some materials in Library.

Problem: Page is slow or freezing

Solutions:

  • Too many saved materials? Go to Library and delete old ones.
  • Close other browser tabs to free up memory.
  • Try a different browser (Chrome or Firefox recommended).
  • Clear browser cache: Settings → Privacy → Clear browsing data (keep localStorage).

Problem: Can't find a specific feature or button

Common locations:

  • Export data: Settings page only (all data in one backup file)
  • Manage custom prompts: Create page → "âœī¸ Manage" button next to Prompt Library
  • View saved materials: Library page in main navigation
  • Auto-save toggle: Settings page

❓ Frequently Asked Questions

Q: Is my patient data secure and private?
A: Yes. All patient data is stored locally in your browser only - it never leaves your device. The only data sent to the AI is the prompt you write (which you control). However, you are responsible for keeping your device secure. See the Privacy page for full details.
Q: Can I use this offline?
A: No, generating materials requires an internet connection to access the AI API. However, you can view saved materials offline if you've already loaded the page.
Q: How long does material generation take?
A: Generation is usually quick - typically under a minute. Complex or longer materials may take a bit longer. If generation seems stuck or is taking unusually long, there may be a connection issue - try refreshing and generating again.
Q: Can I edit generated materials?
A: Yes! In the Library page, click on any saved material and then click the "âœī¸ Edit" button. You can modify the text and save your changes.
Q: How many materials can I save?
A: The system automatically saves up to 100 most recent materials. Older materials are automatically deleted. You can export materials before they're deleted.
Q: What happens if I switch browsers or computers?
A: Your data won't transfer automatically because it's stored locally. To move data: (1) Export from Settings page on old browser, (2) Import on Settings page in new browser. For regular use, stick to one browser.
Q: Is this HIPAA compliant?
A: This is a prototype with local browser storage, not a HIPAA-compliant system. While data stays on your device, you must follow your organization's policies. Use initials instead of names, secure your device, and don't use on shared computers. A future commercial version would include HIPAA compliance.
Q: How much does each generation cost?
A: Approximately $0.015-0.020 per generation. For this prototype, the API cost is covered. A future commercial version would have a subscription model. You can see exact costs in the usage statistics shown after each generation.
Q: Can I share prompts or materials with colleagues?
A: Yes! Use the export/import features:
  • Go to Settings page and click "💾 Export All Data"
  • Important: Remove all PHI from patient data before sharing!
  • Send the .json backup file to your colleague
  • They can import it on their Settings page
The backup includes all materials, patients, and custom prompts.
Q: How do I provide feedback about this prototype?
A: Your feedback is invaluable! Please share your thoughts, suggestions, and issues via email or however the administrator has requested feedback. See the About page for more information about the prototype phase.

💡 Pro Tips & Best Practices

Clinical Best Practices

  • Always review before use: AI-generated materials should be reviewed for clinical appropriateness.
  • Start conservative: Begin with easier reading levels and increase difficulty as you observe patient performance.
  • Personalize iteratively: Use feedback from patient to refine their profile and prompts.
  • Save good prompts: When a prompt works well, save it as a custom template for future use.

Efficiency Tips

  • Batch generation: Generate multiple materials for the week in one session.
  • Template variations: Save variations of prompts for different difficulty levels.
  • Regular backups: Watch for the red dot on Settings link - backup when it appears.
  • Use search: In Library, search by patient name or topic to quickly find past materials.

Important Reminders

  • Never include patient names in prompts sent to the AI (only in profiles stored locally).
  • Export backups when prompted - a red dot on Settings link reminds you when backup is needed.
  • Don't use on public/shared computers - data persists in browser.
  • Review materials for accuracy - AI can occasionally generate incorrect information.

📧 Still Need Help?

This is a prototype, and your experience matters!

If you're experiencing issues, have suggestions, or just want to share your experience, please contact the administrator who provided you access to this prototype.

For more information about the prototype phase and future plans, visit the About page.