API Reference¶

gen-dsp's Python API is organized into three packages:

gen_dsp.core -- Pipeline modules: parsing gen~ exports, building manifests, generating projects, and compiling
gen_dsp.platforms -- Platform registry and backend implementations
gen_dsp.graph -- Optional graph frontend (requires pydantic)

Data Flow¶

gen~ export dir -> parser.py -> ExportInfo -> manifest.py -> Manifest -> Platform.generate_project()

For the graph frontend:

Graph (Python/JSON) -> graph/compile/ -> C++ -> graph/adapter.py -> Manifest -> ProjectGenerator.from_graph()

Core Modules¶

New to the core API? Start with the Core API guide -- a usage-oriented walkthrough of the parse -> manifest -> generate -> build pipeline with examples. The table below links to the per-symbol reference.

Module	Description
`parser`	Regex-based parser for gen~ exports, produces `ExportInfo`
`manifest`	`Manifest` dataclass: front-end-agnostic IR with `ParamInfo`
`project`	`ProjectGenerator`: copies export files, runs template substitution
`builder`	`Builder` delegates to `Platform.build()`, returns `BuildResult`
`patcher`	Applies platform-specific fixes (e.g. `exp2f` on macOS)
`cache`	Resolves shared FetchContent cache directory
`midi`	MIDI mapping detection and compile-definition generation

Platform Registry¶

Module	Description
`platforms`	`Platform` base class, `PLATFORM_REGISTRY`, helper functions

Graph Frontend¶

Module	Description
`graph.models`	Pydantic models: `Graph`, `Param`, node types
`graph.compile`	`compile_graph()`: Graph to standalone C++
`graph.validate`	`validate_graph()`: connectivity and type checks
`graph.optimize`	`optimize_graph()`: dead-code elimination, constant folding
`graph.simulate`	`simulate()`: run graph in Python (requires numpy)
`graph.algebra`	`series()`, `parallel()`, `split()`, `merge()` combinators
`graph.adapter`	Bridge graph output to gen-dsp platform backends