ISP Classifier with Reasoning

This advanced classifier not only classifies complaints but also explains the reasoning behind each decision.

Chain-of-Thought Reasoning Flow

flowchart TD
    A[Complaint] --> B[Extract Symptoms]
    B --> C[List Possible Causes]
    C --> D[Check Field Rules]
    D --> E{Evidence Found?}
    E -->|Yes| F[Confidence: High]
    E -->|No| G[Apply LLM Reasoning]
    G --> H[Step-by-Step Analysis]
    H --> I[Generate Justification]
    I --> J[Final Code + Explanation]

    style A fill:#e3f2fd
    style J fill:#fff3e0

Reasoning Pipeline

flowchart LR
    subgraph Input
        A[Complaint Text]
    end

    subgraph Analysis
        B[Symptom Detection]
        C[Pattern Matching]
        D[Context Evaluation]
    end

    subgraph Reasoning
        E[Cause Analysis]
        F[Rule Application]
        G[LLM Validation]
    end

    subgraph Output
        H[ISP Code]
        I[Confidence Score]
        J[Explanation]
    end

    A --> B
    B --> C
    C --> D
    D --> E
    E --> F
    F --> G
    G --> H
    H --> I
    I --> J

When to Use Reasoning

Scenario	Basic Classifier	Reasoning Classifier
Simple keywords	Yes	Yes
Multiple symptoms	No	Yes
Complex complaints	No	Yes
Audit requirements	No	Yes
Team training	No	Yes

Example Output

Complaint: "Customer says internet is slow and WiFi keeps dropping"

Reasoning Steps:
1. Two symptoms identified: slow speed + intermittent connection
2. Possible causes: router overload, interference, ISP throttling
3. Checking field conditions... WiFi dropping suggests router issue
4. Final analysis: Router configuration problem

Result: ISP-002 (Router Issue)
Confidence: 87%
Justification: WiFi drops + slow speed points to router problem

Running the Demo

python app-reasoning1.py
python app-reasoning2.py

Benefits

• Transparency: Shows why each classification was made
• Accuracy: Multi-step analysis catches edge cases
• Training: Helps new team members understand patterns
• Audit Trail: Complete documentation of decisions