🎯 Coverage Gap Analyzer

An AI-powered assistant that analyzes functional coverage reports and suggests targeted test scenarios to close coverage gaps.

---

📋 Overview

Design verification engineers spend significant time analyzing coverage reports to identify test gaps and create new test scenarios. This tool automates that process using LLM-powered analysis.

Key Features:

• 📊 Parses functional coverage reports (multiple formats supported)
• 🔍 Identifies uncovered bins and cross-coverage gaps
• 💡 Suggests specific test scenarios to hit uncovered cases
• 🤖 Uses LangGraph for multi-step agentic reasoning
• 📝 Generates detailed analysis reports in Markdown

---

🏗️ Architecture

Coverage Report → Parse Gaps → Analyze Each Gap → Generate Suggestions → Final Report
                      ↓              ↓                    ↓
                   LLM Extract    Context Builder    Stimulus Generator

Current Version (MVP):

• Static coverage report analysis
• Mock data for initial testing
• Single-step LLM analysis

Planned Expansions:

1. RAG integration with protocol specifications
2. Integration with actual simulators (Verilator, Icarus)
3. Iterative simulation loop with coverage feedback
4. Multi-protocol support

---

🚀 Quick Start

Prerequisites

• Python 3.11+
• Anthropic API key (get one here)

Installation

# 1. Clone or download this project
cd coverage-gap-analyzer

# 2. Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. Set up API key
echo "ANTHROPIC_API_KEY=your-key-here" > .env

Run the Analyzer

# Run with sample coverage report
python agent.py

# Run with your own coverage report
python agent.py --report path/to/your/coverage_report.txt

# Interactive mode
python agent.py --interactive

---

📁 Project Structure

coverage-gap-analyzer/
├── agent.py              # Main LangGraph agent implementation
├── coverage_parser.py    # Coverage report parsing utilities
├── mock_data.py          # Sample coverage reports generator
├── utils.py              # Helper functions
├── requirements.txt      # Python dependencies
├── .env                  # API keys (not committed)
├── data/
│   ├── sample_coverage.txt        # Example coverage report
│   └── sample_coverage_complex.txt # More complex example
├── outputs/
│   └── analysis_report.md         # Generated analysis reports
└── README.md

---

🎯 Usage Examples

Example 1: Basic Analysis

from agent import CoverageAgent

agent = CoverageAgent()
result = agent.analyze("data/sample_coverage.txt")
print(result['report'])

Example 2: Custom Coverage Report

Create your own coverage report in this format:

COVERAGE REPORT - Simple FIFO Design
=====================================

Coverage Group: fifo_operations
  - empty_to_full: 85% (17/20 bins hit)
    UNCOVERED: consecutive_writes_max_depth
    UNCOVERED: write_when_full
    UNCOVERED: alternating_rd_wr_full
  
  - full_to_empty: 90% (18/20 bins hit)
    UNCOVERED: consecutive_reads_empty
    UNCOVERED: read_when_empty

Cross Coverage: fifo_states x operations
  - 75% coverage (12/16 combinations hit)
    UNCOVERED: (FULL, WRITE)
    UNCOVERED: (EMPTY, READ)
    UNCOVERED: (ALMOST_FULL, WRITE)
    UNCOVERED: (ALMOST_EMPTY, READ)

---

🛠️ Customization

Adding Your Own Protocol

1. Update mock_data.py with your protocol's coverage structure
2. Modify coverage_parser.py to handle your report format
3. Customize prompts in agent.py for protocol-specific context

Adding Simulation Integration

See docs/SIMULATION_INTEGRATION.md (coming soon) for:

• Connecting to Verilator/Icarus
• Parsing VCS/Questa UCDB files
• Building iterative coverage closure loop

---

🎓 Learning Path

This project is designed to teach you:

Phase 1: Core Concepts (Current MVP)

• ✅ LangChain basics and prompt engineering
• ✅ LangGraph for multi-step agent workflows
• ✅ Coverage report parsing and analysis

Phase 2: RAG Integration (Next)

• 📚 Vector databases (ChromaDB/Pinecone)
• 🔍 Document retrieval and chunking
• 🧠 Context-aware stimulus generation

Phase 3: Simulation Loop (Advanced)

• 🔄 Subprocess management for simulators
• 📊 Coverage delta calculation
• 🎯 Iterative test generation

Phase 4: Production Features

• 🌐 Web UI with Streamlit/Gradio
• 📈 Metrics and benchmarking
• 🔐 Multi-user support

---

📊 Sample Output

# Coverage Gap Analysis Report

**Total gaps analyzed:** 7
**Analysis date:** 2024-01-15

---

## 1. `consecutive_writes_max_depth`

**Gap Description:** This bin tracks the scenario where writes occur 
continuously until FIFO reaches maximum depth without any intervening reads.

**Suggested Test Scenario:**
Create a directed sequence that:
1. Initializes FIFO to empty state
2. Performs continuous write operations (no reads)
3. Writes exactly FIFO_DEPTH transactions
4. Verifies full flag assertion on final write

**Sample Code Sketch:**
```systemverilog
class fifo_max_depth_seq extends base_sequence;
  task body();
    for (int i = 0; i < FIFO_DEPTH; i++) begin
      `uvm_do_with(req, {operation == WRITE;})
    end
  endtask
endclass

---

---

## 🤝 Contributing

This is a portfolio/learning project. Suggestions welcome!

---

## 📜 License

MIT License - feel free to use and modify.

---

## 👤 Author: KP Aditya Nukala

Built following the coverage-driven verification roadmap.

**Next Steps:**
1. ⭐ Star this repo if you find it useful
2. 📧 Open an issue with questions or suggestions
3. 🔄 Fork and customize for your own protocols