use std::path::PathBuf;
use clap::Subcommand;
mod run;
mod score;
pub mod types;
#[derive(Subcommand, Debug, Clone)]
pub enum BenchmarkCommand {
/// Run benchmark suite and optionally compare against baseline
Run {
/// Run a smaller core benchmark subset
#[arg(long)]
quick: bool,
/// Override iteration count (default: full=3, quick=2)
#[arg(long)]
iterations: Option<u32>,
/// Baseline report path (default: latest benchmarks/v*.json if present)
#[arg(long)]
baseline: Option<PathBuf>,
/// Exit non-zero when regressions are detected
#[arg(long)]
fail_on_regression: bool,
/// Exit non-zero when transient upstream failures are detected
#[arg(long)]
fail_on_transient: bool,
/// Latency regression threshold percentage (default: 20)
#[arg(long, default_value = "20")]
latency_threshold_pct: f64,
/// Output-size regression threshold percentage (default: 10)
#[arg(long, default_value = "10")]
size_threshold_pct: f64,
/// Max allowed fail-fast latency for contract checks (default: 1500ms)
#[arg(long, default_value = "1500")]
max_fail_fast_ms: u64,
},
/// Run benchmark suite and persist as baseline JSON
SaveBaseline {
/// Run a smaller core benchmark subset
#[arg(long)]
quick: bool,
/// Override iteration count (default: full=3, quick=2)
#[arg(long)]
iterations: Option<u32>,
/// Output path (default: benchmarks/v<CARGO_PKG_VERSION>.json)
#[arg(long)]
output: Option<PathBuf>,
},
/// Score a PI JSONL agent session for BioMCP usage and coverage
ScoreSession {
/// Session JSONL path
session: PathBuf,
/// Expected command set file for coverage scoring
#[arg(long)]
expected: Option<PathBuf>,
/// Show a shorter markdown summary
#[arg(long)]
brief: bool,
},
}
pub async fn run(command: BenchmarkCommand, json_output: bool) -> anyhow::Result<String> {
match command {
BenchmarkCommand::Run {
quick,
iterations,
baseline,
fail_on_regression,
fail_on_transient,
latency_threshold_pct,
size_threshold_pct,
max_fail_fast_ms,
} => {
let opts = run::RunOptions {
quick,
iterations,
baseline,
fail_on_regression,
fail_on_transient,
latency_threshold_pct,
size_threshold_pct,
max_fail_fast_ms,
};
run::run_benchmark(opts, json_output).await
}
BenchmarkCommand::SaveBaseline {
quick,
iterations,
output,
} => {
let opts = run::SaveBaselineOptions {
quick,
iterations,
output,
};
run::save_baseline(opts, json_output).await
}
BenchmarkCommand::ScoreSession {
session,
expected,
brief,
} => {
let opts = score::ScoreSessionOptions {
session,
expected,
brief,
};
score::score_session(opts, json_output)
}
}
}
