Skip to main content

Glossary of Terms

This glossary explains technical terms used throughout the HFIM Admin Panel and documentation. Terms are organized alphabetically for easy reference.

Using This Glossary

Press Ctrl/Cmd + F to search for specific terms on this page!

A​

Active​

A status indicating that a cache entry or feature is currently enabled and being used by the chatbot.

Example: An "Active" cache entry will be served to users when they ask matching questions.

Activity Log​

A record of changes made to the admin panel, showing who made changes, when, and what was modified. Found at the bottom of most admin pages.

Also known as: Audit Log, Change History

Admin Panel​

The web-based dashboard you use to manage the HFIM chatbot. Includes features for cache management, viewing conversations, and configuring prompts.

URL: https://hospitalityrag-66h2e.kinsta.app/admin/

API Key​

A secret code used by the chatbot to communicate with external services (like OpenAI or Pinecone). Not typically visible to admin users.

B​

Bulk Operations​

Actions that can be performed on multiple items at once, such as generating variations for several cache entries simultaneously.

Example: Instead of editing 10 cache entries one at a time, you can select all 10 and generate variations in one operation.

C​

Cache / Cache Entry​

A pre-written question-and-answer pair stored for quick retrieval. When a user asks a question similar to a cached question, the chatbot returns the cached answer immediately without needing to search documents.

Benefits: Faster responses, consistent answers, reduced AI costs

Example: If you cache the question "What is the HFIM program?" with a specific answer, users asking similar questions will get that exact answer.

Cache Hit​

When a user's question matches a cached entry closely enough to return the cached response.

Opposite: Cache Miss (when no suitable cache entry is found)

Cache Hit Rate​

The percentage of user questions that were successfully answered using cached responses.

Example: A 75% hit rate means 3 out of every 4 questions were answered from cache.

Clear Redis​

An action that removes a specific cache entry from the Redis (L1) cache layer only. The entry remains in the PostgreSQL database and continues to function normally.

Use when: You've updated an entry and want to force a cache refresh, or when troubleshooting cache inconsistencies.

See also: Managing Entry Status and Deletion

Chatbot​

The AI-powered assistant that answers questions about the HFIM program. Also called "Sage" or "HFIM Assistant."

Confidence Score​

A number between 0 and 1 indicating how confident you are that a cached response is accurate and appropriate.

Scale:

  • 0.95-1.0 = Very High Confidence (verified facts, official information)
  • 0.85-0.94 = High Confidence (generally accurate, may need occasional updates)
  • 0.70-0.84 = Medium Confidence (helpful but may vary by context)
  • Below 0.70 = Low Confidence (use with caution)

Example: Official program requirements should have 0.95+ confidence, while subjective advice might be 0.80.

Conversation​

A single interaction between a user and the chatbot, including the question asked, the response given, and metadata like feedback and response time.

Also called: Chat Session, Interaction

D​

Dashboard​

The main overview page of the admin panel, showing key metrics and statistics about cache performance and chatbot usage.

Deactivate​

An action that sets a cache entry's status to "Inactive", temporarily disabling it without deleting it. The entry stops serving responses but remains in the database and can be reactivated later.

Use when: Information is temporarily outdated, needs verification, or is seasonal.

Opposite: Activate (re-enables an inactive entry)

See also: Managing Entry Status and Deletion

E​

Embedding​

A mathematical representation of text that allows the AI to understand meaning and find similar content. Not directly visible to admin users, but used behind the scenes for matching questions.

F​

Feedback​

User ratings (positive 👍 or negative 👎) on chatbot responses, helping you identify which answers are helpful or need improvement.

Types:

  • Positive Feedback: User found the answer helpful
  • Negative Feedback: User was unsatisfied (may include comments explaining why)
  • No Feedback: User didn't rate the response

G​

GPT / GPT-4 / GPT-5​

The AI language model used to generate chatbot responses. GPT-5 is the latest, most advanced version (as of January 2026).

Full name: Generative Pre-trained Transformer

H​

Hard Delete​

A permanent deletion action that completely removes a cache entry from the database. Unlike deactivation, a hard delete cannot be undone and all data (including variations, admin notes, and history) is lost forever.

Safety: Requires typing "DELETE" to confirm, preventing accidental deletion.

Use when: Entry was created by mistake, is a duplicate, or is permanently irrelevant.

See also: Managing Entry Status and Deletion

HFIM​

Hospitality and Food Industry Management - the UGA academic program this chatbot supports.

Hit Rate​

See Cache Hit Rate

Hot Reload​

The ability to update system prompts without restarting the chatbot. Changes take effect immediately.

I​

Inactive​

A status indicating that a cache entry is disabled and won't be used by the chatbot, but is still stored in the system for future reference.

Use case: Temporarily disable outdated information without deleting it completely.

Ingestion​

The process of adding new documents and information into the chatbot's knowledge base. Typically handled by technical staff.

M​

Metadata​

Additional information about a cache entry or conversation, such as timestamps, user IDs, or source documents.

Modal / Modal Window​

A popup window that appears on top of the main page, used for editing cache entries, viewing conversation details, or configuring settings.

Example: When you click "Edit" on a cache entry, an edit modal appears.

P​

Pagination​

Dividing large lists into multiple pages. Use page numbers or "Next/Previous" buttons to browse.

Example: If there are 100 cache entries, they might be shown as 10 pages of 10 entries each.

Pinecone​

The vector database service used to store and search the chatbot's knowledge base. Not directly visible to admin users.

Frequently asked questions identified by analyzing user interactions, helping you prioritize which questions to cache.

Prompt / System Prompt​

Instructions given to the AI that control how it behaves, what tone it uses, and what rules it follows when generating responses.

Example: A system prompt might say "Always be friendly and helpful" or "Never make up information you're not sure about."

Q​

Question Variations​

Alternative ways to ask the same question, helping the cache match more user queries.

Example: For "What is the HFIM program?", variations might include:

  • "Tell me about HFIM"
  • "What does HFIM stand for?"
  • "Explain the Hospitality and Food Industry Management program"

R​

RAG (Retrieval-Augmented Generation)​

The technology that allows the chatbot to search documents and use that information to generate accurate responses. Combines document search with AI generation.

How it works: Question → Search documents → Use found information → Generate answer

Redis​

A fast in-memory caching system (L1 cache) used to store temporary data and speed up the chatbot. Redis is checked first before PostgreSQL (L2 cache) for cached responses.

Admin action: The "Clear Redis" button removes a specific entry from Redis only, forcing a fresh lookup from PostgreSQL.

See also: Clear Redis, Two-Tier Cache

Reframing​

The process of rephrasing a user's question to improve search results and response quality.

Example: "Who's the prof for 3000?" might be reframed to "Who teaches HFIM 3000?"

Response Time​

How long it took (in milliseconds) for the chatbot to generate and return a response.

Typical ranges:

  • Cache hit: 50-500ms (very fast)
  • RAG search: 2,000-8,000ms (normal)
  • Over 10,000ms: Slow (may indicate issues)

S​

Session​

A unique identifier for a user's interaction with the chatbot, allowing the chatbot to remember context across multiple questions.

Also called: Session ID, Conversation ID

The navigation menu on the left side of the admin panel, providing links to all major features.

Soft Delete​

See Deactivate - setting an entry to "Inactive" status rather than permanently removing it. The entry remains in the database but stops serving responses.

Opposite: Hard Delete (permanent removal)

Source / Sources​

The documents or information used to generate a chatbot response. Shown as citations with page numbers and filenames.

Example: "HFIM_Handbook.pdf (page 12)"

Success Rate​

The percentage of cache entries that successfully matched user questions (excluding entries with no matches).

Formula: (Entries with at least 1 match) ÷ (Total active entries) × 100%

System Prompt​

See Prompt / System Prompt

T​

TTL (Time-To-Live)​

How long a cached response remains valid before it should be reviewed and updated.

Common values:

  • 7 days: Information that changes frequently (dates, deadlines)
  • 30 days: Stable information that may need occasional updates
  • 90 days: Permanent information (program history, foundational concepts)

Example: A cache entry about "Spring 2026 application deadline" should have a short TTL (7 days) so it's reviewed regularly.

Times Served​

How many times a cache entry has been used to answer user questions.

Example: A "Times Served" count of 150 means that cache entry has been returned 150 times.

U​

Upsert​

A technical term meaning "insert or update" - adding new data or updating existing data in the database. Used during document ingestion.

V​

Variations​

See Question Variations

Vector​

A mathematical representation of text used for similarity matching. Not directly visible to admin users, but enables the chatbot to find relevant information.

The technology that finds relevant documents by comparing the meaning of the question to the meaning of stored content (rather than just matching keywords).

Additional Resources​

Still Have Questions?​

Suggest a New Term​

If you encounter a term that's not defined here, please email support@uga.edu and we'll add it to the glossary!

Last Updated: January 2026

Last updated on