Files
Claude-Code-Workflow/.codex/skills/numerical-analysis-workflow/specs/phase-topology.md
catlog22 f389e3e6dd feat: add dynamic pipeline detection for team sessions v5 architecture
Backend:
- Fix readLegacyFiles to handle { name, prefix }[] role format
- Add roles backfill in getEffectiveTeamMeta when meta.json exists
- Ensure pipeline_stages and roles flow correctly to API response

Team Skills:
- Add pipeline metadata initialization to all 16 team skill coordinator roles
- Each skill now reports pipeline_stages and roles to meta.json at session init

Documentation:
- Update command references and component documentation
- Add numerical-analysis-workflow skill spec
- Sync zh/en translations for commands and components
2026-03-04 14:52:07 +08:00

7.6 KiB

Phase Topology — Diamond Deep Tree

Wave coordination patterns for the Numerical Analysis Diamond Workflow (NADW).

Purpose

Phase Usage
Phase 1 (Decomposition) Reference when assigning waves and dependencies
Phase 2 (Execution) Context flow between waves
Phase 3 (Aggregation) Structure the final report by topology

1. Topology Overview

The NADW uses a Staged Diamond topology — six sequential waves, each with 3 parallel tracks. Context flows cumulatively from earlier waves to later ones.

         Wave 1: [T1.1] [T1.2] [T1.3]     Global Survey (3 parallel)
              ↓ Context P1
         Wave 2: [T2.1] [T2.2] [T2.3]     Theory (3 parallel)
              ↓ Context P1+P2
         Wave 3: [T3.1] [T3.2] [T3.3]     Algorithm (3 parallel)
              ↓ Context P1+P2+P3
         Wave 4: [T4.1] [T4.2] [T4.3]     Module (3 parallel)
              ↓ Context P1-P4
         Wave 5: [T5.1] [T5.2] [T5.3]     Local (3 parallel)
              ↓ Context P1-P5
         Wave 6: [T6.1] [T6.2] [T6.3]     Integration (3 parallel)

2. Wave Definitions

Wave 1: Global Survey

Property Value
Phase Name Global Survey
Track Count 3
Dependencies None (entry wave)
Context Input Project codebase only
Context Output Governing equations, architecture map, validation KPIs
Max Parallelism 3

Tracks:

ID Role Dimension Scope
T1.1 Problem_Domain_Analyst domain_modeling Full project
T1.2 Software_Architect architecture_analysis Full project
T1.3 Validation_Strategist validation_design Full project

Wave 2: Theoretical Foundations

Property Value
Phase Name Theoretical Foundations
Track Count 3
Dependencies Wave 1
Context Input Context Package P1
Context Output LaTeX formulas, convergence theorems, complexity bounds
Max Parallelism 3

Tracks:

ID Role Dimension Deps context_from
T2.1 Mathematician formula_derivation T1.1 T1.1
T2.2 Convergence_Analyst convergence_analysis T1.1 T1.1
T2.3 Complexity_Analyst complexity_analysis T1.1;T1.2 T1.1;T1.2

Wave 3: Algorithm Design & Stability

Property Value
Phase Name Algorithm Design
Track Count 3
Dependencies Wave 2
Context Input Context Package P1+P2
Context Output Pseudocode, stability conditions, performance model
Max Parallelism 3

Tracks:

ID Role Dimension Deps context_from
T3.1 Algorithm_Designer method_selection T2.1 T2.1;T2.2;T2.3
T3.2 Stability_Analyst stability_analysis T2.1;T2.2 T2.1;T2.2
T3.3 Performance_Modeler performance_modeling T2.3 T1.2;T2.3

Wave 4: Module Implementation

Property Value
Phase Name Module Implementation
Track Count 3
Dependencies Wave 3
Context Input Context Package P1-P3
Context Output Code-algorithm mapping, data structure decisions, API contracts
Max Parallelism 3

Tracks:

ID Role Dimension Deps context_from
T4.1 Module_Implementer implementation_analysis T3.1 T1.2;T3.1
T4.2 Data_Structure_Designer data_structure_review T3.1;T3.3 T2.3;T3.1;T3.3
T4.3 Interface_Analyst interface_analysis T3.1 T1.2;T3.1

Wave 5: Local Function-Level

Property Value
Phase Name Local Function-Level
Track Count 3
Dependencies Wave 4
Context Input Context Package P1-P4
Context Output Optimization recommendations, edge case catalog, precision risk matrix
Max Parallelism 3

Tracks:

ID Role Dimension Deps context_from
T5.1 Code_Optimizer optimization T4.1 T3.3;T4.1
T5.2 Edge_Case_Analyst edge_case_analysis T4.1 T1.1;T3.2;T4.1
T5.3 Precision_Auditor precision_audit T4.1;T4.2 T3.2;T4.1;T4.2

Wave 6: Integration & QA

Property Value
Phase Name Integration & QA
Track Count 3
Dependencies Wave 5
Context Input Context Package P1-P5 (ALL cumulative)
Context Output Final test plan, benchmark report, QA assessment
Max Parallelism 3

Tracks:

ID Role Dimension Deps context_from
T6.1 Integration_Tester integration_testing T5.1;T5.2 T1.3;T5.1;T5.2
T6.2 Benchmark_Engineer benchmarking T5.1 T1.3;T3.3;T5.1
T6.3 QA_Auditor quality_assurance T5.1;T5.2;T5.3 T1.1;T2.1;T3.1;T4.1;T5.1;T5.2;T5.3

3. Context Flow Map

Directed Context (prev_context column)

T1.1 ──► T2.1, T2.2, T2.3
T1.2 ──► T2.3, T3.3, T4.1, T4.3
T1.3 ──► T6.1, T6.2
T2.1 ──► T3.1, T3.2
T2.2 ──► T3.1, T3.2
T2.3 ──► T3.1, T3.3, T4.2
T3.1 ──► T4.1, T4.2, T4.3
T3.2 ──► T5.2, T5.3
T3.3 ──► T4.2, T5.1, T6.2
T4.1 ──► T5.1, T5.2, T5.3
T4.2 ──► T5.3
T5.1 ──► T6.1, T6.2
T5.2 ──► T6.1
T5.3 ──► T6.3

Broadcast Context (discoveries.ndjson)

All agents read/append to the same discoveries.ndjson. Key discovery types flow across waves:

Wave 1: governing_equation, architecture_pattern ──► all subsequent waves
Wave 2: convergence_property ──► Wave 3-6
Wave 3: stability_issue, numerical_method ──► Wave 4-6
Wave 4: (implementation findings) ──► Wave 5-6
Wave 5: precision_risk, performance_bottleneck ──► Wave 6

4. Perspective Reuse Matrix

How each wave's output is reused by later waves:

Source P2 Reuse P3 Reuse P4 Reuse P5 Reuse P6 Reuse
P1 Equations Formalize → LaTeX Constrain methods Code-eq mapping Singularity sources Correctness baseline
P1 Architecture Constrain discretization Parallel strategy Module boundaries Hotspot location Integration scope
P1 Validation - Benchmark selection Test criteria Edge case sources Final validation
P2 Formulas - Parameter constraints Loop termination Precision requirements Theory verification
P2 Convergence - Mesh refinement Iteration control Error tolerance Rate verification
P2 Complexity - Performance baseline Data structure choice Optimization targets Perf comparison
P3 Pseudocode - - Impl reference Line-by-line audit Regression baseline
P3 Stability - - Precision selection Cancellation detection Numerical verification
P3 Performance - - Memory layout Vectorization targets Benchmark targets
P4 Modules - - - Function-level focus Module test plan
P5 Optimization - - - - Performance tests
P5 Edge Cases - - - - Regression tests
P5 Precision - - - - Numerical tests

5. Diamond Properties

Property Value
Total Waves 6
Total Tasks 18 (3 per wave)
Max Parallelism per Wave 3
Widest Context Fan-in T6.3 (receives from 7 tasks)
Deepest Dependency Chain T1.1 → T2.1 → T3.1 → T4.1 → T5.1 → T6.1 (depth 6)
Context Accumulation Cumulative (each wave adds to previous context)
Topology Type Staged Parallel with Diamond convergence at Wave 6