Skip to content

dqa-agent-guide.md

Latest commit

 

History

History
282 lines (196 loc) · 8.12 KB

dqa-agent-guide.md

File metadata and controls

282 lines (196 loc) · 8.12 KB

DQA Agent Guide

This guide explains how to use the AI-powered Data Quality Assessment (DQA) Agent to get insights about your sensor data quality.

Overview

The DQA Agent is an AI-powered chat assistant that helps you:

  • Ask questions about your sensor data quality
  • Get insights and recommendations
  • Understand data quality metrics
  • Receive explanations of analysis results
  • Get help with data quality issues

Understanding the DQA Agent

What is the DQA Agent?

The DQA Agent is a conversational AI assistant powered by OpenAI that:

  • Understands your data quality context
  • Accesses your sensor data and metadata
  • Provides intelligent responses to questions
  • Offers recommendations and insights
  • Explains complex concepts in simple terms

Capabilities

The agent can help with:

  • Data Quality Questions: Ask about completeness, accuracy, consistency
  • Sensor Analysis: Get insights about specific sensors
  • Metric Explanations: Understand what quality metrics mean
  • Recommendations: Receive suggestions for improving data quality
  • Troubleshooting: Get help with data quality issues

Accessing the DQA Agent

  1. Open the application at http://localhost:8000
  2. Click DQA Agent in the sidebar or home page
  3. The chat interface opens automatically

Using the DQA Agent

Step 1: Select Machine Group (Optional)

  1. Choose a machine group from the dropdown
  2. This provides context for the agent
  3. Agent can access data for selected machine group

Note: You can chat without selecting a machine group for general questions.

Step 2: Start a Conversation

  1. Type your question in the message input box
  2. Press Enter or click Send
  3. Wait for the agent's response

Step 3: Continue Conversation

  1. Read the agent's response
  2. Ask follow-up questions
  3. Request clarifications
  4. Get more details on specific topics

Types of Questions You Can Ask

Data Quality Questions

Examples:

  • "What is the overall data quality for this machine group?"
  • "Which sensors have the most missing values?"
  • "Are there any accuracy issues I should be concerned about?"
  • "How does data quality compare across different sensors?"

Sensor-Specific Questions

Examples:

  • "Tell me about sensor 22PI102"
  • "What are the threshold values for pressure sensors?"
  • "Which sensors are most correlated?"
  • "Are there any sensors with high outlier rates?"

Metric Explanations

Examples:

  • "What does completeness percentage mean?"
  • "How is accuracy assessed?"
  • "What are outliers and why do they matter?"
  • "Explain correlation analysis"

Recommendations

Examples:

  • "What should I do about sensors with high missing values?"
  • "How can I improve data quality?"
  • "Which sensors need maintenance?"
  • "What are best practices for data quality?"

Troubleshooting

Examples:

  • "Why is my data quality assessment showing poor results?"
  • "What could cause high missing values?"
  • "How do I fix accuracy issues?"
  • "What should I check if sensors have many alarms?"

Best Practices for Effective Queries

Be Specific

Good:

  • "What is the missing value percentage for sensor 22PI102?"
  • "Which pressure sensors have accuracy issues?"

Less Effective:

  • "Tell me about data"
  • "What's wrong?"

Provide Context

Good:

  • "For the KT2201 machine group, which sensors have the most missing values?"
  • "In the last month, what is the overall data quality?"

Less Effective:

  • "What's the quality?"
  • "Tell me about sensors"

Ask Follow-Up Questions

Example Conversation:

  1. User: "What is the data quality for KT2201?"
  2. Agent: [Provides overview]
  3. User: "Which specific sensors have issues?"
  4. Agent: [Lists sensors with problems]
  5. User: "What should I do about sensor 22PI102?"

Use Natural Language

The agent understands natural language, so you can:

  • Ask questions conversationally
  • Use technical terms or plain language
  • Request explanations
  • Ask for recommendations

Understanding Agent Responses

Response Format

Agent responses may include:

  • Text Explanations: Detailed answers to your questions
  • Lists: Bulleted or numbered lists
  • Recommendations: Actionable suggestions
  • Code/Data: Technical details when relevant
  • Markdown Formatting: Structured information

Response Types

Direct Answers:

  • Straightforward responses to specific questions
  • Clear and concise information

Analytical Responses:

  • Analysis of your data
  • Comparisons and insights
  • Pattern identification

Recommendations:

  • Actionable suggestions
  • Best practices
  • Improvement strategies

Explanations:

  • Concept explanations
  • Metric definitions
  • Process descriptions

Limitations and Considerations

Data Access

  • Agent accesses data from selected machine group
  • May not have access to all historical data
  • Limited to data in the database

Response Accuracy

  • Responses are AI-generated and should be verified
  • May not always be 100% accurate
  • Use as guidance, not absolute truth

Context Understanding

  • Agent understands conversation context
  • May need clarification for complex questions
  • Works best with specific questions

API Requirements

  • Requires valid OpenAI API key
  • Internet connection needed
  • May have usage limits

Tips for Better Results

  1. Select Machine Group: Provides better context
  2. Be Specific: Ask clear, specific questions
  3. Use Follow-Ups: Build on previous responses
  4. Verify Information: Cross-check important findings
  5. Ask for Examples: Request examples when needed
  6. Request Clarification: Ask agent to explain if unclear

Example Conversations

Example 1: Data Quality Overview

User: "What is the overall data quality for KT2201?"

Agent: "Based on the data for KT2201, the overall data quality is good. The completeness is 94.5%, which is above the acceptable threshold. However, there are some sensors with accuracy issues that should be addressed..."

User: "Which sensors have accuracy issues?"

Agent: "The sensors with the most accuracy issues are: 22PI102 (15% alarm rate), 22TI111 (12% alarm rate), and 22PI103 (10% alarm rate). These sensors are exceeding their threshold values frequently..."

Example 2: Understanding Metrics

User: "What does completeness percentage mean?"

Agent: "Completeness percentage measures how much of the expected data is actually present. It's calculated as: (Actual Readings / Expected Readings) × 100. For example, if you expect 1000 readings but only have 950, the completeness is 95%..."

Example 3: Getting Recommendations

User: "How can I improve data quality for my sensors?"

Agent: "Here are some recommendations to improve data quality: 1. Address sensors with high missing values by checking hardware and connections. 2. Calibrate sensors with accuracy issues. 3. Review and adjust threshold settings if needed. 4. Implement regular maintenance schedules..."

Troubleshooting

Agent Not Responding

Problem: No response from agent Solutions:

  • Check OpenAI API key is configured
  • Verify internet connection
  • Check backend service is running
  • Review error messages

Inaccurate Responses

Problem: Agent provides incorrect information Solutions:

  • Verify data is loaded correctly
  • Check machine group selection
  • Ask more specific questions
  • Cross-check with other analyses

Slow Responses

Problem: Agent takes too long to respond Solutions:

  • Wait for response (may take time)
  • Check network connection
  • Verify backend performance
  • Reduce question complexity

Next Steps

After using the DQA Agent:

  1. Verify Findings: Cross-check with other analyses
  2. Take Action: Implement recommendations
  3. Re-assess: Check improvements after changes
  4. Explore More: Use other analysis features

Related Documentation

Note: The DQA Agent requires a valid OpenAI API key. See Getting Started Guide for setup instructions.