Skip to main content

Registry

The adapter registry is how the core finds language adapters without importing them. There is a module-level singleton, adapter_registry, plus the AdapterRegistry class behind it.

from graphlens import adapter_registry, AdapterRegistry

Entry-point group​

The registry discovers adapters declared under the graphlens.adapters entry-point group:

[project.entry-points."graphlens.adapters"]
python = "graphlens_python:PythonAdapter"

Methods​

load(name: str) -> type[LanguageAdapter]​

Return the adapter class registered for name. In-memory registrations are checked first, then entry points. Raises AdapterNotFoundError if no adapter matches. Call the returned class to get an instance:

adapter = adapter_registry.load("python")()

available() -> list[str]​

Return the sorted names of all available adapters — both manually registered and discovered through entry points.

adapter_registry.available()      # ['python', 'typescript']

register(name: str, adapter_cls: type[LanguageAdapter]) -> None​

Register an adapter class in memory under name. Useful in tests, or to override a discovered adapter:

adapter_registry.register("python", MyPythonAdapter)

Typical use​

from pathlib import Path
from graphlens import adapter_registry

for name in adapter_registry.available():
    adapter = adapter_registry.load(name)()
    if adapter.can_handle(Path(".")):
        graph = adapter.analyze(Path("."))
        ...

This is exactly what the CLI's --lang auto does: enumerate available adapters and keep the ones whose can_handle() returns true for the project.