prodigy 0.4.4

Turn ad-hoc Claude sessions into reproducible development pipelines with parallel AI agents
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120

name: debtmap-parallel-elimination
mode: mapreduce

# Setup phase: Analyze the codebase and generate debt items
setup:
  timeout: 900  # 15 minutes for coverage generation
  commands:
    # Generate coverage data with tarpaulin
    - shell: "just coverage-lcov"

    # Run debtmap with coverage data to establish baseline
    - shell: "debtmap analyze src --lcov target/coverage/lcov.info --output .prodigy/debtmap-before.json --format json"

# Map phase: Process each debt item in parallel with planning and validation
map:
  # Input configuration - debtmap-before.json contains items array
  input: .prodigy/debtmap-before.json
  json_path: "$.items[*]"

  # Commands to execute for each debt item with planning and validation
  agent_template:
    # Phase 1: Create implementation plan
    - claude: "/prodigy-debtmap-plan --item '${item}' --output .prodigy/plan-${item_id}.md"
      capture_output: true
      validate:
        commands:
          - claude: "/prodigy-validate-debtmap-plan --item '${item}' --plan .prodigy/plan-${item_id}.md --output .prodigy/validation-${item_id}.json"
        result_file: ".prodigy/validation-${item_id}.json"
        threshold: 75
        on_incomplete:
          commands:
            - claude: "/prodigy-revise-debtmap-plan --gaps ${validation.gaps} --plan .prodigy/plan-${item_id}.md"
          max_attempts: 3
          fail_workflow: false

    # Phase 2: Execute the plan
    - claude: "/prodigy-debtmap-implement --plan .prodigy/plan-${item_id}.md"
      commit_required: true
      validate:
        commands:
          - shell: "just coverage-lcov"
          - shell: "debtmap analyze src --lcov target/coverage/lcov.info --output .prodigy/debtmap-after-${item_id}.json --format json"
          - shell: "debtmap compare --before .prodigy/debtmap-before.json --after .prodigy/debtmap-after-${item_id}.json --plan .prodigy/plan-${item_id}.md --output .prodigy/comparison-${item_id}.json --format json"
          - claude: "/prodigy-validate-debtmap-improvement --comparison .prodigy/comparison-${item_id}.json --previous-validation .prodigy/debtmap-validation-${item_id}.json --output .prodigy/debtmap-validation-${item_id}.json"
        result_file: ".prodigy/debtmap-validation-${item_id}.json"
        threshold: 75
        on_incomplete:
          commands:
            - claude: "/prodigy-complete-debtmap-fix --plan .prodigy/plan-${item_id}.md --validation .prodigy/debtmap-validation-${item_id}.json --attempt ${validation.attempt_number}"
              commit_required: true
            - shell: "just coverage-lcov"
            - shell: "debtmap analyze src --lcov target/coverage/lcov.info --output .prodigy/debtmap-after-${item_id}.json --format json"
            - shell: "debtmap compare --before .prodigy/debtmap-before.json --after .prodigy/debtmap-after-${item_id}.json --plan .prodigy/plan-${item_id}.md --output .prodigy/comparison-${item_id}.json --format json"
          max_attempts: 5
          fail_workflow: true

    # Phase 3: Verify tests pass
    - shell: "just test"
      on_failure:
        claude: "/prodigy-debug-test-failure --output ${shell.output}"
        max_attempts: 5
        fail_workflow: true

    # Phase 4: Check formatting and linting
    - shell: "just fmt-check && just lint"
      on_failure:
        claude: "/prodigy-lint ${shell.output}"
        max_attempts: 5
        fail_workflow: true

  # Parallelization settings
  max_parallel: 5  # Run up to 5 agents in parallel

  # Process high-score items first
  # Note: Items are wrapped in either File or Function enums
  # File items: item.File.score
  # Function items: item.Function.unified_score.final_score
  # NULLS LAST ensures Files (with non-null File.score) sort before Functions (with null File.score)
  filter: "File.score >= 10 OR Function.unified_score.final_score >= 10"
  sort_by: "File.score DESC NULLS LAST, Function.unified_score.final_score DESC NULLS LAST"  # Process highest score items first
  max_items: 10  # Limit to 10 items per run

# Reduce phase: Aggregate results and verify overall improvements
reduce:
  # Phase 1: Run final tests across all changes
  - shell: "just test"
    on_failure:
      claude: "/prodigy-debug-test-failure --output ${shell.output}"
      max_attempts: 5
      fail_workflow: true

  # Phase 2: Check formatting and linting
  - shell: "just fmt-check && just lint"
    on_failure:
      claude: "/prodigy-lint ${shell.output}"
      max_attempts: 5
      fail_workflow: true

  # Phase 3: Re-run debtmap to measure cumulative improvements
  - shell: "just coverage-lcov"

  - shell: "debtmap analyze src --lcov target/coverage/lcov.info --output .prodigy/debtmap-after.json --format json"

  # Phase 4: Compare debt metrics and create final commit with summary
  - write_file:
      path: ".prodigy/map-results.json"
      content: "${map.results}"
      format: json
      create_dirs: true

  - claude: |
      /prodigy-compare-debt-results \
        --before .prodigy/debtmap-before.json \
        --after .prodigy/debtmap-after.json \
        --map-results-file .prodigy/map-results.json \
        --successful ${map.successful} \
        --failed ${map.failed} \
        --total ${map.total}
    commit_required: true