Improve Scenario Generation Using MCP Server

[cielonet]

This pull request improves the scenario/test case generation process by using a local mcp (stdio) format server to generate the response needed for scenarios.

Included in the examples directory are two modified files that use local llm's with local mcp servers.

[cielonet]

python3 generate_from_http_mcp.py 
======================================================================
🚀 MCP Scenario Generator
======================================================================

📍 Server: http://mcp-code-coverage:8000/mcp
🤖 LLM: gpt-oss
🎯 Scenarios: 10

🔍 Connecting to MCP server at http://mcp-code-coverage:8000/mcp...
✓ Connected successfully
✓ Found 4 tools and 1 resources

======================================================================
📋 Discovered Capabilities
======================================================================

🔧 Tools (4):
   1. submit_testcase
      
    Submit a test case for coverage analysis.
    
    Args...
   2. get_function_coverage
      
    Get detailed line-by-line coverage for a specific funct...
   3. get_file_coverage
      
    Get coverage report for a specific source file.
    
  ...
   4. get_coverage_summary
      
    Get overall coverage summary across all files.
    
   ...

📚 Resources (1):
   1. get_coverage_summary
      
    Real-time coverage summary updated after each test case...

======================================================================
🎯 Generating Scenarios
======================================================================

[18:21:18] OK    Using model: gpt-oss
[18:21:18] INFO  Available: 4 tool(s), 1 resource(s).
[18:21:18] STEP  Preparing prompt for scenario generation
[18:21:18] STEP  Calling model: gpt-oss
[18:21:25] OK    Model responded in 7.66s.
[18:21:25] INFO  Raw content length: 3452 chars.
[18:21:25] STEP  Parsing model output
[18:21:25] STEP  Connecting to MCP server for JSON formatting
[18:21:25] OK    Using MCP tool: format_scenario
[18:21:25] INFO  Formatted 5/10 scenarios
[18:21:25] INFO  Formatted 10/10 scenarios
[18:21:26] OK    Formatted 10 scenarios via MCP server.
[18:21:26] INFO  Difficulty distribution:
   1/5:   1  █
   2/5:   1  █
   3/5:   3  ███
   4/5:   3  ███
   5/5:   2  ██
   1. Submit a single HTML test case that intentionally omits the `<title>` tag and use `submit_testcase` to trigger coverage…  (difficulty 1/5)
   2. Run a JavaScript test case (`test_type="js"`) designed to exercise a rarely used branch in the `calculateDiscount` funct…  (difficulty 2/5)
   3. Submit a PNG rendering test (`test_type="png"`) that draws a complex UI component. After submission, use `get_file_cover…  (difficulty 3/5)
   4. Evaluate overall health of the codebase by invoking `get_coverage_summary`. Parse the returned statistics to produce a c…  (difficulty 3/5)
   5. Create a raw data test (`test_type="raw"`) that feeds malformed JSON to the `parseConfig` function. After running the te…  (difficulty 4/5)
[18:21:26] OK    Generated 10 scenarios in 7.73s total.

======================================================================
💾 Saving Scenarios
======================================================================
✓ Saved to: scenarios.json
✓ File size: 4,290 bytes

======================================================================
📊 Summary
======================================================================

📈 Statistics:
  Total scenarios: 10
  Average difficulty: 3.4/5
  Average task length: 306 characters

📊 Difficulty Distribution:
  1/5:   1 ( 10.0%)  ██████████
  2/5:   1 ( 10.0%)  ██████████
  3/5:   3 ( 30.0%)  ██████████████████████████████
  4/5:   3 ( 30.0%)  ██████████████████████████████
  5/5:   2 ( 20.0%)  ████████████████████

✅ Complete!
root@art:/workspace/code-coverage-agent# cat scenarios.json 
{
  "metadata": {
    "generated_at": "2025-11-01T18:21:26.002254",
    "mcp_server_url": "http://mcp-code-coverage:8000/mcp",
    "llm_model": "gpt-oss",
    "num_tools": 4,
    "num_resources": 1,
    "tool_names": [
      "submit_testcase",
      "get_function_coverage",
      "get_file_coverage",
      "get_coverage_summary"
    ],
    "resource_names": [
      "get_coverage_summary"
    ],
    "summary": {
      "total_scenarios": 10,
      "difficulty_distribution": {
        "1": 1,
        "2": 1,
        "3": 3,
        "4": 3,
        "5": 2
      },
      "avg_difficulty": 3.4,
      "avg_task_length": 305.5
    }
  },
  "scenarios": [
    {
      "task": "Submit a single HTML test case that intentionally omits the `<title>` tag and use `submit_testcase` to trigger coverage of the HTML validator. After execution, retrieve the real\u2011time summary from `coverage://summary` to confirm that the tidy warning count increased.",
      "difficulty": 1
    },
    {
      "task": "Run a JavaScript test case (`test_type=\"js\"`) designed to exercise a rarely used branch in the `calculateDiscount` function. Then call `get_function_coverage` for `calculateDiscount` to verify that the new branch was hit and see its line\u2011by\u2011line coverage diff.",
      "difficulty": 2
    },