Frequently Asked Questions
Quick answers to common questions about using the HFIM Admin Panel.
Use the search bar at the top of this page or check the Glossary for term definitions!
General Questions
What is the HFIM Admin Panel?
The admin panel is a web-based tool that lets you manage the HFIM chatbot ("Sage"). You can:
- Create and edit cached responses for faster answers
- Review user conversations and feedback
- Customize the chatbot's behavior through system prompts
- Monitor performance metrics and usage statistics
Learn more: Getting Started
Who can access the admin panel?
Access is restricted to:
- HFIM program administrators
- Faculty members (with permission)
- IT support staff
- Designated staff with specific responsibilities
If you need access, contact your program coordinator or email support@uga.edu.
Learn more: Login & Security
Do I need technical skills to use this?
No! The admin panel is designed for non-technical users. This documentation uses plain language and step-by-step tutorials with screenshots to help you learn.
How often should I check the admin panel?
Recommended frequency:
- Daily (5-10 minutes): Check the dashboard for any negative feedback
- Weekly (30 minutes): Review popular questions, update cache entries as needed
- Monthly (1-2 hours): Analyze conversation trends, clean up unused cache entries, review prompt performance
Login & Access
I forgot my password. What do I do?
Contact IT support at support@uga.edu with your UGA email address. They'll send you password reset instructions within 1 business day.
Note: There's currently no self-service password reset option.
Learn more: Login & Security
Why can't I log in?
Common issues:
- Wrong password: Admin panel password ≠ UGA MyID password
- Caps Lock is on: Passwords are case-sensitive
- Too many failed attempts: Wait 15 minutes or contact support
- Browser issues: Try clearing cache or using a different browser
Learn more: Troubleshooting Login Issues
How long do I stay logged in?
Your session expires after 30 minutes of inactivity. You'll need to log in again if you see a "Session Expired" message.
Tip: Your changes are auto-saved, so you won't lose work if your session expires.
Cache Management
What are cached responses?
Cached responses are pre-written question-and-answer pairs that let the chatbot respond instantly without searching documents. They provide:
- ✅ Faster response times
- ✅ Consistent, verified answers
- ✅ Lower costs (fewer AI queries)
Example: Instead of searching documents every time someone asks "What is HFIM?", the chatbot uses your cached answer.
Learn more: Cache Overview
How do I know what questions to cache?
Good questions to cache:
- ✅ Frequently asked questions (check Popular Questions)
- ✅ Official information (program requirements, deadlines, contacts)
- ✅ Questions with consistent answers (program history, degree options)
- ✅ Questions users search multiple times (course descriptions, prerequisites)
Avoid caching:
- ❌ Rarely asked questions (< 5 times)
- ❌ Information that changes frequently
- ❌ Questions requiring current context (today's date, recent news)
Learn more: Cache Best Practices
Why isn't my cache entry being used?
Common reasons:
- Status is "Inactive": Edit the entry and set status to "Active"
- Confidence too low: Increase confidence score to 0.70 or higher
- Question variations don't match: Add more variations that users actually type
- Users ask differently: Check conversations to see exact wording users use
Check this: Look at "Times Served" - if it's 0, users aren't matching this entry
Learn more: Cache Troubleshooting
How many question variations should I add?
Recommended: 3-10 variations per cache entry
Include variations that:
- Use different wording ("What is HFIM?" vs "Tell me about HFIM")
- Include common misspellings ("professer" vs "professor")
- Use abbreviations ("HFIM 3000" vs "Hospitality 3000")
- Ask in different formats (questions vs statements)
Tip: Use the AI-powered variation generator to save time!
Learn more: Generating Variations
What's a good confidence score?
Recommended confidence levels:
| Score Range | Use For | Example |
|---|---|---|
| 0.95-1.0 | Official facts, verified info | Program requirements, contact information |
| 0.85-0.94 | Generally accurate, stable info | Course descriptions, faculty bios |
| 0.70-0.84 | Helpful but may vary | General advice, suggestions |
| Below 0.70 | Use with caution | Outdated info, uncertain answers |
Learn more: Glossary - Confidence Score
What TTL (Time-To-Live) should I use?
Recommended TTL values:
| TTL | Use For | Examples |
|---|---|---|
| 7 days | Changes frequently | Upcoming deadlines, current semester info |
| 30 days | Needs occasional updates | Course schedules, faculty office hours |
| 90 days | Stable, permanent info | Program history, degree types, mission statement |
Tip: Set reminders to review entries before their TTL expires!
Learn more: Glossary - TTL
Conversations
How do I find conversations with negative feedback?
- Go to Conversations in the sidebar
- Click the Filter dropdown
- Select "Negative Feedback Only"
- Click Apply Filters
Learn more: Filtering Feedback
Why should I review negative feedback?
Negative feedback helps you:
- 🔍 Identify incorrect or unhelpful responses
- 📝 Create new cache entries for common questions
- 🎯 Improve existing cache entries
- 🐛 Find bugs or system issues
Tip: Address negative feedback weekly to maintain chatbot quality.
Learn more: Filtering Feedback
Can I edit a user's feedback rating?
Yes! If a user accidentally clicked the wrong feedback button, you can correct it:
- Go to Conversations
- Click View on the conversation
- Click Edit Feedback
- Change the rating and optionally add a note
- Click Save Changes
Learn more: Editing Feedback
How do I turn a good conversation into a cached response?
- Go to Conversations
- Find a conversation with a high-quality answer
- Click View on that conversation
- Click Convert to Cache Entry
- Review the auto-filled fields
- Add question variations
- Click Save
Learn more: Converting to Cache
Prompts & Configuration
What is a system prompt?
A system prompt is a set of instructions that tells the AI how to behave. It controls:
- The chatbot's tone and personality
- What topics it should focus on
- Rules for generating responses
- How it should handle uncertain questions
Think of it like: A employee handbook that tells the chatbot how to do its job.
Learn more: Prompt Overview
Can I change the chatbot's personality?
Yes! Edit the system prompt to adjust:
- Tone: Formal vs. casual, friendly vs. professional
- Style: Concise vs. detailed, academic vs. conversational
- Approach: Directive vs. suggestive, strict vs. flexible
Warning: Test changes in preview mode before saving to avoid unintended behavior.
Learn more: Editing Prompts
Will my prompt changes affect the live chatbot immediately?
Yes! The admin panel uses hot reload, meaning changes take effect within seconds. That's why it's critical to:
- Preview your changes before saving
- Test with sample questions
- Monitor the first few conversations after making changes
Learn more: Hot Reload
Can I undo a prompt change?
Yes! The admin panel keeps a complete change history:
- Go to Prompt Configuration
- Click Change History
- Find the version you want to restore
- Click View to see the old prompt
- Copy the content and create a new update
Learn more: Change History
What should I never change in the system prompt?
Avoid removing or modifying:
- Anti-hallucination rules (prevents making up information)
- Source citation requirements (ensures transparency)
- Safety guidelines (prevents inappropriate responses)
- Core HFIM program facts
Safe to change:
- Tone and personality
- Examples and scenarios
- Emphasis on certain topics
- Response format preferences
Learn more: Prompt Best Practices
Performance & Metrics
What's a good cache hit rate?
Target cache hit rate: 60-80%
- Above 80%: Excellent! Your cache is comprehensive
- 60-80%: Good - most common questions are covered
- 40-60%: Fair - consider adding more cache entries
- Below 40%: Low - review popular questions and create cache entries
Learn more: Cache Performance Metrics
Why is my success rate 0%?
Success rate only includes cache entries that have been served at least once. If you just created entries, they need time to match user questions. Wait 1-2 weeks and check again.
Tip: "Times Served = 0" entries don't count toward success rate.
Learn more: Performance Metrics
What's an acceptable response time?
Response time guidelines:
- 50-500ms: Excellent (cache hit)
- 2,000-5,000ms: Good (RAG search)
- 5,000-10,000ms: Acceptable (complex search)
- Over 10,000ms: Slow (may indicate issues)
Tip: More cache hits = faster average response times!
Learn more: Performance Metrics
Technical Issues
The admin panel won't load. What should I do?
Try these steps in order:
- Check internet connection
- Refresh the page (F5)
- Clear browser cache (Ctrl+Shift+Delete)
- Try a different browser (Chrome, Firefox, Safari, Edge)
- Disable browser extensions temporarily
- Contact support if still not working
Learn more: Authentication Troubleshooting
I clicked "Save" but nothing happened. Did it save?
Check for:
- ✅ Success message (green notification at top)
- ✅ Modal closed automatically (changes saved)
- ❌ Error message (red notification - changes not saved)
- ❌ Modal still open (likely validation error)
If no confirmation: Try saving again or refresh the page to check if changes persisted.
The Activity Log shows "Failed to load". Is something broken?
No! The activity log is a supplementary feature. If it fails to load:
- ✅ Main functionality still works
- ✅ Your changes are still saved
- ✅ Cache, conversations, and prompts operate normally
This just means the audit log data couldn't be retrieved at that moment. Try refreshing later.
Learn more: Understanding the Interface
Can I use the admin panel on my phone?
The admin panel is not optimized for mobile devices. For the best experience, use:
- ✅ Desktop computer (recommended)
- ✅ Laptop (1366×768 resolution or larger)
- ⚠️ Tablet (works but requires scrolling)
- ❌ Phone (too small, difficult to use)
Learn more: Responsive Design
Best Practices
How often should I update cache entries?
Review schedule:
- Weekly: Check entries with TTL < 14 days
- Monthly: Review entries with negative feedback or low usage
- Quarterly: Audit all cache entries for accuracy
- As needed: Update when program information changes
Tip: Set calendar reminders based on TTL values!
Learn more: Cache Best Practices
Should I delete unused cache entries?
Instead of deleting, consider:
- Set status to "Inactive" (preserves entry for future use)
- Review question variations (maybe they don't match user wording)
- Lower confidence score (if you're unsure about accuracy)
- Add better variations (improve matching potential)
Only delete if:
- Information is permanently outdated
- Entry was created by mistake
- Question is no longer relevant to HFIM
Learn more: Cache Best Practices
What's the most important thing I can do to improve the chatbot?
Top priority: Address negative feedback regularly.
Every week:
- Filter conversations by negative feedback
- Read user comments to understand issues
- Create cache entries for frequently problematic questions
- Update prompts if users consistently misunderstand the chatbot's tone
Learn more: Analyzing Interactions
Still Have Questions?
Search This Documentation
Use the search bar at the top of any page to find specific topics.
Check Other Resources
- Glossary - Definitions of technical terms
- Support - Contact information and support hours
- Getting Started - Introduction and quick start guide
Contact Support
Email: support@uga.edu
Include:
- Your UGA email address
- Detailed description of your question or issue
- Screenshots if applicable
- What you've already tried
Response time: Usually within 1 business day
Last Updated: January 2026