Skip to main content

Viewing Conversations

Learn how to navigate the Conversations section, browse user interactions, and view detailed conversation information.

Accessing the Conversations Section​

From the Dashboard​

  1. Log in to the admin panel
  2. Look at the left sidebar
  3. Click "💬 Conversations"
  4. The conversations table loads automatically
Quick Access

Bookmark the direct URL for faster access: https://hospitalityrag-66h2e.kinsta.app/admin/conversations.html

Understanding the Conversations Table​

Table Layout​

The main conversations view shows a table with all user interactions, sorted by most recent first.

Default View:

  • 10 conversations per page (adjustable with pagination)
  • Most recent first (newest at top)
  • All feedback types included (positive, negative, and none)

Table Columns Explained​

1. ID Column​

What it shows: Unique identifier for each conversation

Format: Sequential number (1, 2, 3, etc.)

Why it matters: Use this ID when reporting issues or referencing specific conversations with support

Example: #142

2. Session Column​

What it shows: Session identifier linking related questions from the same user

Format: Shortened hash (e.g., abc123...)

Why it matters: Helps you identify multiple questions from the same user session for context

Example: Same session ID = user asked 3 follow-up questions

Session Continuity

Users can ask multiple related questions in one session. The chatbot remembers context within a session, so later questions may reference earlier ones.

3. Question Column​

What it shows: Exact text the user typed (truncated if long)

Format: First 50-80 characters of the question

Why it matters: Quick preview to identify conversation topics

Examples:

  • "What are the prerequisites for HFIM 3000?"
  • "How do I apply for an internship?"
  • "Tell me about career paths"
Click to See Full Question

If the question is truncated (ends with ...), click View to see the complete text in the detail modal.

4. Feedback Column​

What it shows: User rating for the response

Icons:

  • 👍 Positive - User found the answer helpful
  • 👎 Negative - User was unsatisfied (may include comment)
  • — No feedback - User didn't rate the response

Comment Indicator: If negative feedback includes a comment, you'll see 👎 followed by a message icon or preview text

Why it matters: Negative feedback is your #1 priority for improving the chatbot

5. Response Time Column​

What it shows: How long the chatbot took to generate the response

Format: Milliseconds (ms) - e.g., 2,450ms = 2.45 seconds

Color Coding (if present):

  • Green (< 1,000ms): Excellent - likely cache hit
  • Yellow (1,000-8,000ms): Normal - RAG search
  • Red (> 10,000ms): Slow - potential issue

Why it matters: Identifies opportunities for caching and spots performance problems

6. Timestamp Column​

What it shows: Date and time when the conversation occurred

Format: MMM DD, YYYY HH:MM AM/PM (e.g., "Jan 11, 2026 3:45 PM")

Why it matters: Helps you correlate conversations with events (deadlines, start of semester, etc.)

7. Actions Column​

What it shows: Buttons for interacting with the conversation

Available Actions:

  • View (👁️ icon) - Open full conversation details
  • Edit (✏️ icon) - Modify feedback rating or add notes

Using Pagination​

Controls at the bottom of the table:

  1. Page Numbers: Click a specific page number (1, 2, 3, etc.)
  2. Previous/Next Buttons: Navigate one page at a time
  3. First/Last Buttons: Jump to beginning or end
  4. Items Per Page: Change how many conversations show per page (10, 25, 50, 100)

Changing Items Per Page​

Steps:

  1. Look for "Show: [10]" dropdown (usually bottom-right)
  2. Click to expand options
  3. Select your preferred number (25, 50, or 100)
  4. Table reloads with new page size

When to use more items:

  • Searching for specific conversation by scanning
  • Exporting/analyzing large datasets
  • Reviewing feedback during weekly audits

When to use fewer items:

  • Slow internet connection
  • Quick browsing
  • Focused review of recent conversations

Opening Conversation Details​

Viewing Full Details​

Steps:

  1. Find the conversation in the table
  2. Click the "View" button (👁️ icon) in the Actions column
  3. Modal opens with complete conversation details

Closing the modal:

  • Click X in top-right corner
  • Click Close button at bottom
  • Press Esc key
  • Click outside the modal (on darkened background)

Scrolling the modal:

  • Use mouse wheel to scroll content
  • Use trackpad gestures to scroll
  • Use scroll bar on right side
Modal vs. Page

The main page behind the modal doesn't scroll when the modal is open. Only the modal content scrolls.

Understanding the Detail Modal​

The conversation detail modal shows all information about a single interaction.

Section 1: Header Information​

Located at: Top of modal

Shows:

  • Conversation ID: Unique number (e.g., #142)
  • Session ID: Full session identifier (e.g., abc123def456...)
  • Timestamp: Date and time (e.g., "January 11, 2026 at 3:45 PM EST")

Why it matters: Provides metadata for tracking and support

Section 2: Question​

Located at: Below header

Shows: Full text of the user's question

Format: Exactly as the user typed it (preserves capitalization, punctuation, typos)

Example:

what do i need before i can take 3000

Why it matters: Shows how users actually phrase questions, informing cache variation creation

Section 3: Answer​

Located at: Middle of modal

Shows: Complete response the chatbot generated

Format: Formatted text with:

  • Markdown rendering (bold, italic, lists)
  • Paragraph breaks
  • Bullet points or numbered lists
  • Link formatting (if present)

Example:

To enroll in HFIM 3000, you need to complete the following prerequisites:

• HFIM 2100 - Introduction to Hospitality Management
• HFIM 2200 - Food Service Operations

Additionally, you must have:
- Junior standing (60+ credit hours)
- Minimum 2.5 GPA in major courses

Why it matters: Lets you evaluate response quality, accuracy, and helpfulness

Section 4: Feedback​

Located at: After answer

Shows:

  • Rating: Positive 👍, Negative 👎, or None —
  • Comment: User's optional feedback comment (if provided)

Positive Feedback Example:

Rating: 👍 Positive
Comment: None

Negative Feedback Example:

Rating: 👎 Negative
Comment: "This doesn't answer my question about prerequisites."

No Feedback Example:

Rating: — No Feedback
Comment: None
Negative Feedback Priority

Always read negative feedback comments carefully. They tell you exactly what went wrong and how to fix it.

Section 5: Performance Data​

Located at: Below feedback

Shows:

  • Response Time: Milliseconds (e.g., 2,450ms)
  • Cache Status: Hit or Miss (if shown)
  • Model Used: AI model (e.g., "gpt-4o", "gpt-5-mini")

Example:

Response Time: 2,450ms (2.45 seconds)
Cache Status: Miss (RAG search)
Model: gpt-4o

Why it matters: Helps identify caching opportunities and performance issues

Section 6: Sources​

Located at: Bottom of modal

Shows: JSON array of documents used to generate the response

Format:

[
  {
    "filename": "HFIM_Handbook_2026.pdf",
    "page": 12,
    "section": "Course Prerequisites",
    "relevance": 0.94,
    "content_preview": "HFIM 3000 requires completion of..."
  },
  {
    "filename": "Course_Catalog_2025-2026.pdf",
    "page": 45,
    "section": "HFIM 3000",
    "relevance": 0.88,
    "content_preview": "Hospitality Operations Management..."
  }
]

Key Fields:

  • filename: Source document name
  • page: Page number in PDF
  • section: Section or heading
  • relevance: Similarity score (0-1, higher = more relevant)
  • content_preview: Snippet of matched text

Why it matters:

  • Verify the chatbot used correct sources
  • Identify outdated or irrelevant documents
  • Understand why wrong answers occur
Empty Sources

If the sources array is empty [], the response was returned from cache (not RAG search).

Common Viewing Scenarios​

Scenario 1: Reviewing Recent Conversations​

Goal: Check what users asked today

Steps:

  1. Go to Conversations section
  2. Check timestamp column (default: newest first)
  3. Scan questions for common themes
  4. Click "View" on interesting conversations

Why: Stay aware of current user needs and issues

Scenario 2: Finding a Specific Conversation​

Goal: Locate conversation #142 mentioned in an email

Steps:

  1. Look at ID column
  2. Use pagination if needed (conversation may be on another page)
  3. Or use browser search (Ctrl/Cmd + F) and type #142

Why: Respond to user feedback or support tickets

Scenario 3: Reviewing a User Session​

Goal: See all questions from one user session

Steps:

  1. Note the Session ID from one conversation
  2. Scan the Session column for matching IDs
  3. View each conversation in sequence
  4. Understand the user's full interaction flow

Why: Identify follow-up question patterns and context issues

Example:

Session abc123:
1. "What is HFIM?" (2:00 PM)
2. "How do I apply?" (2:02 PM)
3. "What are the requirements?" (2:05 PM)

Scenario 4: Analyzing Response Times​

Goal: Identify slow responses that need caching

Steps:

  1. Scan Response Time column
  2. Note conversations > 8,000ms (8 seconds)
  3. Click "View" to see the question
  4. Consider creating cache entries for these questions

Why: Improve user experience by reducing wait times

Scenario 5: Checking Sources for Accuracy​

Goal: Verify chatbot is using current documents

Steps:

  1. Click "View" on any conversation
  2. Scroll to Sources section at bottom
  3. Check filenames and dates
  4. Look for outdated documents (e.g., "Handbook_2023.pdf" in 2026)

Why: Ensure responses use current information

What to Look For​

Red Flags 🚨​

Immediate attention needed:

  • ❌ Response time > 15 seconds
  • ❌ Empty sources array for non-cache responses
  • ❌ Negative feedback with comments
  • ❌ Same question asked 5+ times recently
  • ❌ Sources from old/outdated documents

Yellow Flags ⚠️​

Investigate when time allows:

  • ⚠️ Response time 8-15 seconds
  • ⚠️ Questions with no feedback (high volume)
  • ⚠️ Multiple sources with low relevance (< 0.70)
  • ⚠️ Follow-up questions indicating confusion

Green Flags ✅​

Good signs (but still monitor):

  • ✅ Response time < 3 seconds
  • ✅ Positive feedback
  • ✅ Relevant, current sources (relevance > 0.80)
  • ✅ Diverse questions (shows wide usage)
  • ✅ Cache hits (empty sources + fast time)

Next Steps​

Now that you can view conversations:

  1. Learn to filter by feedback - Find specific conversations quickly
  2. Edit incorrect feedback - Fix accidental ratings
  3. Convert to cache entries - Save excellent answers
  4. Analyze interaction patterns - Use data strategically
  5. Troubleshoot issues - Solve common problems

Remember: The Conversations section is your window into how users interact with the chatbot. Spend 5-10 minutes daily reviewing recent conversations to stay aware of user needs and issues!

Last updated on