BEAM program analysis via Souffle Datalog.
Argus extracts facts from BEAM bytecode and feeds them to Souffle Datalog rules for whole-program, multi-module analysis. The pipeline is:
.beam files → normalize → emit facts → Souffle rules → resultsQuick start
# Detect supervision-tree anti-patterns in a project.
Argus.analyze([MyApp.Supervisor, MyApp.Worker], :supervision)
# Find ETS tables created without heir protection.
Argus.analyze([MyApp.Cache], :ets)
# Run custom Datalog rules.
Argus.analyze([MyApp.Worker], {:custom, "path/to/rules.dl"})
# Run every built-in analysis and get structured findings.
Argus.run_analyses([MyApp.Supervisor, MyApp.Worker])Architecture
Layer 1 (generic) walks every BEAM instruction and emits base facts about instructions, registers, control flow, and calls. Layer 2 (domain extractors) produces higher-level semantic facts by interpreting OTP patterns, supervision trees, and other BEAM-specific constructs.
Both layers feed into Souffle, which evaluates Datalog rules and returns derived relations as results.
Summary
Functions
Runs an analysis against the given modules.
Runs analyses and returns structured findings.
Functions
@spec analyze([atom() | String.t()], Argus.Analysis.analysis(), keyword()) :: {:ok, Argus.Analysis.result()} | {:error, term()}
Runs an analysis against the given modules.
See Argus.Analysis.run/3 for details.
@spec run_analyses( [atom() | String.t()], keyword() ) :: {:ok, Argus.Findings.t()} | {:error, term()}
Runs analyses and returns structured findings.
Extracts facts from the given modules (atoms or .beam paths) once,
evaluates each selected analysis against them, and converts every result
row into a finding with severity, prose, and code anchors.
See Argus.Findings.run/2 for options and the degradation contract.
{:ok, findings} = Argus.run_analyses([MyApp.Cache], analyses: [:ets])
Enum.each(findings.findings, &IO.puts(&1.title))