Skip to content

PRODUCT.md

Latest commit

 

History

History
70 lines (51 loc) · 2.27 KB

PRODUCT.md

File metadata and controls

70 lines (51 loc) · 2.27 KB

PRODUCT

Product statement

pyCellModeller is a clean-slate reimplementation of CellModeller focused on a Torch-native simulation core.

For bootstrap v1, the product is intentionally narrow: a deterministic CPU-first engine with a minimal public API and native tutorial semantics.

Explicit backend and compatibility decision

Legacy PyOpenCL/OpenCL execution paths are intentionally retired for this rewrite.

  • v1 backend strategy: Torch-only
  • no CL/OpenCL naming in public APIs
  • no compatibility shims unless explicitly requested
  • no multi-backend architecture in bootstrap scope

Target users (current phase)

  • Developers establishing the simulation core architecture
  • Researchers validating deterministic stepping behavior
  • Contributors building future mechanics/biology/fields modules on a tested foundation

Current user value delivered

Users can currently:

  • install the package scaffold
  • import stable top-level API symbols
  • initialize a deterministic toy state
  • run deterministic CPU stepping
  • run unit/integration/regression tests for the implemented path

Current implemented API surface

  • pycellmodeller.Simulation
  • pycellmodeller.SimulationConfig
  • pycellmodeller.SimulationState

Supported flow:

  1. build config
  2. create simulation
  3. initialize example state
  4. step simulation deterministically

In-scope for current bootstrap

  • CPU-first deterministic behavior as reference
  • torch tensor state model
  • minimal stepping kernel (x <- x + v*dt)
  • documentation and tests that match actual implementation
  • milestone progression toward native tutorial examples

Not yet implemented (intentionally)

  • production mechanics models
  • extracellular field simulation
  • biological rule engine
  • persistence/export workflows
  • rich CLI workflows
  • visualization module (viz directory not present)
  • repository docs/ and docker/ directories (not present)

Success criteria for this stage

This stage is successful if:

  • documented API matches what imports and runs
  • deterministic CPU stepping remains stable under tests
  • architecture stays clean for future mechanics/fields/biology additions
  • a native pyCellModeller Tutorial 1 equivalent runs with semantic parity
  • success does not depend on executing legacy tutorial files unchanged