Adding an algorithm

MemDiver auto-discovers algorithms via pkgutil.walk_packages on import. Adding one is a single file.

1. Choose a mode

• known_key → requires context.secrets (ground-truth keylog). Place in algorithms/known_key/.

• unknown_key → no ground truth. Place in algorithms/unknown_key/.

2. Subclass BaseAlgorithm

# algorithms/unknown_key/my_algo.py
from algorithms.base import AlgorithmResult, AnalysisContext, BaseAlgorithm, Match
from core.constants import UNKNOWN_KEY


class MyAlgoAlgorithm(BaseAlgorithm):
    name = "my_algo"
    description = "One-line summary shown in the UI algorithm picker."
    mode = UNKNOWN_KEY

    def run(self, dump_data: bytes, context: AnalysisContext) -> AlgorithmResult:
        threshold = context.extra.get("my_threshold", 0.5)
        matches: list[Match] = []
        # ... analyze dump_data, append Match instances ...
        confidence = 1.0 if matches else 0.0
        return AlgorithmResult(
            algorithm_name=self.name,
            confidence=confidence,
            matches=matches,
            metadata={"threshold": threshold},
        )

No registration call is needed — the next get_registry() call discovers the class via pkgutil.walk_packages + issubclass(BaseAlgorithm).

3. Surface in the UI

Add "my_algo" to frontend/src/stores/app-store.ts ANALYSIS_ALGORITHMS. Optionally extend frontend/src/utils/algorithm-availability.ts if the algorithm has mode-specific requirements.

4. Test

Write a unit test under tests/test_my_algo.py following the pattern of existing tests/test_*.py. Include a fixture dump from tests/fixtures/generate_*.py so the test is self-contained.