ThunderGraph Model Documentation

ThunderGraph Model is an executable systems modeling library for Python.

User

• Install
  • Requirements
  • Install for development (repo workflow)
  • Install as a library
  • Verify install
• Overview
  • Start here
  • Concrete concept examples
  • API reference
• Mental Model
  • Mini walkthrough (same symbol, three layers)
• Quickstart (Concrete Example)
  • Install
  • Recommended: instantiate → evaluate
  • Advanced: explicit compile_graph + Evaluator
  • Composable requirements next to Part
  • What happened (both paths)
  • Next examples
• Concept: Requirements — parameter / attribute / constraint (DEFAULT)
  • The default pattern: package-level parameter / attribute / constraint
  • Leaf model.requirement(...) vs composable Requirement
  • Package-level value surface
  • Advanced (rare): leaf-level requirement_input / requirement_attribute / requirement_accept_expr
  • Summary: which pattern to use
• Concept: External Compute (Concrete Binding)
  • Example
  • Input and output contract
  • Rule of thumb
• FAQ
  • When should I use ConfiguredModel.evaluate vs compile_graph + Evaluator?
  • Why do I need RunContext?
  • Why are inputs bound by stable id in the explicit pipeline?
  • I changed code in a notebook and behavior is weird.
  • When should I use Requirement (composable requirement package)?
  • What is the difference between requirement_input and requirement_attribute?
  • Where do I start if I am new?
  • The graph compiled but evaluate failed or results look wrong

Developer

• Developer architecture
  • Subsystems
  • Pipeline sequences
  • Invariants (mental checklist for contributors)
  • Composable requirement packages
  • Extension points
• Repository map
  • Layout
  • Suggested reading order (new contributor)
• Extension playbook
  • Supported extension surfaces
  • What is not a stable extension API
  • Practical workflow
  • References
• Testing
  • Layout
  • Adding a regression
  • Notebooks
  • Coverage
  • References
• Contributing
  • Before you open a PR
  • Checklist (public behavior changes)
  • Docs build
  • References
• Releasing
  • Version
  • What ships
  • Documentation
  • References
• NumPy-style docstrings for tg_model (Phase 0 contract)
  • Standard
  • What counts as “public”
  • Required sections by symbol kind
  • Style rules
  • PR checklist (touched public API)
  • Pilot reference

API Reference

• API Reference
  • tg_model
  • tg_model.model
  • tg_model.execution
  • tg_model.integrations
  • tg_model.analysis

Phase 2 Drafts (working pages)

• What ThunderGraph Model Is
  • What It Is Not
  • Core Promise
  • Reader Mental Model
  • Primary Entry Points
  • Why This Matters
• Execution Pipeline (Compile -> Instantiate -> Graph -> Evaluate)
  • 1) Compile types
  • 2) Instantiate a configured topology
  • 2a) Default run path: ConfiguredModel.evaluate (recommended)
  • 3) Compile dependency graph
  • 4) Validate graph (optional but recommended)
  • 5) Evaluate one run (explicit pipeline)
  • Requirement acceptance path
  • Behavioral path
  • One sentence summary
• Glossary
  • Element
  • System
  • Part
  • Requirement
  • ModelDefinitionContext
  • Ref / PartRef / AttributeRef / RequirementRef
  • compile (type compile)
  • ConfiguredModel
  • ValueSlot
  • RunContext
  • DependencyGraph
  • Evaluator
  • RunResult
  • Constraint
  • Requirement acceptance
  • allocate
  • requirement_input
  • requirement_accept_expr
  • ExternalComputeBinding
  • BehaviorTrace
  • Scenario
  • Stable ID
• Repository Map (thundergraph-model)
  • Core package
  • Documentation
  • Examples and notebooks
  • Tests
  • Practical navigation order