Kibo.AgenticAdmin

Conversation History

Overview

The Conversation History feature provides comprehensive access to all past interactions between users and your agent. This powerful tool enables you to review conversations, analyze patterns, identify improvement opportunities, and ensure quality service delivery.

Accessing Conversation History

Navigate to 💬 Conversations in the sidebar to access the conversation history interface.

Conversations List

Conversation List Interface

Filter Options

The conversation list provides multiple filtering capabilities:

  1. Session ID Search
    • Direct lookup by unique conversation identifier
    • Useful for investigating specific interactions
    • Supports partial matching
  2. Conversation Type
    • All Types (default)
    • Text: Text-based conversations
    • Audio: Voice interactions
    • Undetermined: Type not yet classified
    • Unspecified: Legacy or system conversations
  3. Language Code
    • Filter by conversation language
    • Example: en, es, fr, de
    • Helps in language-specific analysis
  4. Date Range
    • Start Time From: Beginning date/time
    • Start Time To: End date/time
    • Useful for period-based analysis

Conversation Table

The main table displays conversations with the following columns:

Column Description Sortable
Session ID Unique conversation identifier ✓
Start Time When the conversation began ✓
Duration Total conversation length ✓
Turns Number of user-agent exchanges ✓
Channel Communication channel (web, mobile, etc.) ✓
Type Conversation type (TEXT, AUDIO) ✓
Actions View or Delete options ✗

Pagination

Conversation Detail View

Clicking “View” on any conversation opens the detailed view showing the complete interaction.

Conversation Detail

Conversation Structure

Each conversation is broken down into turns, displaying:

  1. Turn Header
    • Turn number (sequential)
    • Turn type (text, tool_response, etc.)
  2. User Input Section
    • Exact message from the user
    • Timestamp (if available)
    • Any metadata
  3. Agent Response Section
    • Complete agent response
    • Tool invocations (if any)
    • Response formatting

Tool Call Visualization

When tools are used in a conversation:

Tool Request:

🔧 Tool Call Request
[Tool Name]
Status: PENDING

▶ Input Parameters
{
  "parameter1": "value1",
  "parameter2": "value2"
}

Tool Response:

✅ Tool Call Response
[Tool Name]

▶ Output Parameters
{
  "result": "data",
  "status": "success"
}

Conversation Flow Example

Turn 1 - text
─────────────
User: Hello, I'm looking for running shoes

Agent: Hello! I'd be happy to help you find the perfect 
running shoes. What type of running do you do?

Turn 2 - text
─────────────
User: Mostly road running, about 20 miles per week

Agent: Great! For road running at that volume, you'll want 
good cushioning and durability. Let me search our inventory.

🔧 Tool Call: Product Search API
Input: { "query": "road running shoes", "category": "footwear" }

Turn 3 - tool_response
─────────────
✅ Tool Response: Product Search API
Output: { "products": [...], "count": 15 }

Agent: I found 15 great options for road running shoes. 
Here are my top recommendations...

Conversation Metrics

At the bottom of each conversation detail, comprehensive metrics are displayed:

Basic Metrics

Advanced Metrics

Analyzing Conversations

Pattern Recognition

Look for common patterns in conversations:

  1. Successful Flows
    • Short duration with resolution
    • High confidence scores
    • Positive user feedback
  2. Problem Areas
    • Multiple tool failures
    • Low confidence responses
    • User frustration indicators
  3. Improvement Opportunities
    • Repeated questions
    • Unclear agent responses
    • Missing functionality

Quality Metrics

Monitor these key indicators:

  1. Resolution Rate
    • Conversations ending successfully
    • No handoff required
    • User goal achieved
  2. Efficiency
    • Average turns to resolution
    • Tool usage effectiveness
    • Response accuracy
  3. User Satisfaction
    • Sentiment analysis
    • Explicit feedback
    • Conversation abandonment

Export and Reporting

Data Export Options

  1. Individual Conversations
    • Export as JSON
    • Export as CSV
    • Include full transcript
  2. Bulk Export
    • Date range selection
    • Filter criteria applied
    • Scheduled exports

Report Generation

Create reports for:

Privacy and Compliance

Data Retention

Access Control

Sensitive Data

Best Practices

Regular Review

  1. Daily Checks
    • Monitor for failures
    • Check unusual patterns
    • Verify tool performance
  2. Weekly Analysis
    • Trend identification
    • Performance metrics
    • User feedback review
  3. Monthly Deep Dive
    • Comprehensive analysis
    • Playbook optimization
    • Training improvements

Using Insights

  1. Playbook Improvements
    • Identify missing intents
    • Refine conversation flows
    • Add error handling
  2. Tool Optimization
    • Monitor response times
    • Check failure rates
    • Update configurations
  3. Agent Training
    • Collect training examples
    • Identify knowledge gaps
    • Improve responses

Troubleshooting Conversations

Common Issues

  1. Incomplete Conversations
    • Check for timeouts
    • Verify connection stability
    • Review error logs
  2. Tool Failures
    • Examine tool responses
    • Check authentication
    • Verify API availability
  3. Low Confidence Responses
    • Review intent matching
    • Check training data
    • Analyze context handling

Investigation Process

  1. Identify Problem
    • Find affected conversations
    • Note common patterns
    • Document symptoms
  2. Analyze Root Cause
    • Review conversation flow
    • Check tool interactions
    • Examine agent logic
  3. Implement Solution
    • Update playbooks
    • Fix tool configurations
    • Retrain if necessary

Advanced Features

Search Capabilities

Beyond basic filters:

Conversation Threading

Real-time Monitoring

Integration Options

Analytics Platforms

Export data to:

CRM Integration

Ticketing Systems