Skip to content

Evidence and protocol CLI

Language: English · 简体中文

The strategy-pipeline command provides two entry points for research evidence:

  • research-protocol creates a protocol-manifest template or evaluates an existing JSON/YAML manifest.
  • afml-evidence generates sizing and strategy-risk evidence, with optional HRP returns evidence. It can also merge the generated evidence into a protocol manifest.

Initialize or evaluate a protocol

Create a template for the chosen protocol level:

strategy-pipeline research-protocol --level exploratory --init-manifest protocol.yml

To evaluate a manifest, pass --manifest instead. The command writes a report to research_protocol_report.json by default; --output changes that path. Strict evaluation is enabled by default and returns a non-zero exit code when the report does not pass. Use --no-strict when a failing report should still return exit code 0.

Generate AFML evidence

Point the command at a run directory:

strategy-pipeline afml-evidence --run-dir artifacts/runs/example

Optional arguments set the target Sharpe ratio, evaluation period, bootstrap sample count, random seed, and HRP returns input. Pass --manifest to merge the generated evidence into an existing protocol manifest. The command writes back to that manifest unless --manifest-output specifies a separate path. JSON manifests remain JSON; other supported manifest files are written as YAML.

The CLI lives in strategy-pipeline. Evidence calculations are implemented by portfolio-backtester; protocol manifest loading and evaluation are implemented by quant-market-research. This repository provides argument parsing, file I/O, and command orchestration around those components.