ReqLedger

ReqLedger is the Ledgerwerk record owner for requirements and acceptance criteria.

It stores durable requirement records as readable Markdown files with TOML metadata. It does not run tests, generate Gherkin, parse Gherkin, discover pytest tests, or own task/architecture records.

Downstream tools such as SpecMason can consume ReqLedger exports to create and validate behavior specs, pytest mappings, coverage reports, and evidence reports.

Scope boundary

Tool	Owns
reqledger	requirements and acceptance criteria
specmason	behavior specs, mappings, coverage, and evidence
taskledger	tasks and task workflow
archledger	architecture records and ADRs
planledger	implementation plans
wikimason	wiki content

ReqLedger is normative: accepted requirement records define intended product behavior. Source code and tests describe observed behavior but never silently rewrite requirements.

Requirements live as readable files

requirements/
  README.md
  manifest.json
  records/
    req-0001.req.md
  reports/
    reqledger/
      review.md
      review.json
reqledger.toml

Each requirement is a Markdown file with TOML front matter delimited by +++:

+++
schema_version = 1
id = "REQ-0001"
title = "Reject invalid login passwords"
kind = "functional"
status = "accepted"
priority = "must"
tags = ["auth", "login"]
source = "manual"
created = "2026-06-16"
updated = "2026-06-16"

[[criteria]]
id = "AC-0001"
statement = "Login is rejected when an invalid password is submitted."
verification = "behavior"
status = "accepted"
tags = ["auth"]
+++

# REQ-0001: Reject invalid login passwords

## Intent

The product must reject invalid passwords for registered users.

Quickstart

pip install reqledger
reqledger init
reqledger new "Reject invalid login passwords" \
  --kind functional \
  --priority must \
  --criterion "Login is rejected when an invalid password is submitted."
reqledger validate
reqledger index
reqledger review
reqledger export --format json

Commands

reqledger init [--force] [--json]
reqledger new "Title" [--kind] [--priority] [--tag TAG]... [--criterion STMT]... [--status] [--source] [--json]
reqledger list [--status STATUS] [--tag TAG] [--json]
reqledger show REQ-0001 [--json]
reqledger validate [--json]
reqledger index [--json]
reqledger link REQ-0001 --task ID --arch ID --spec PATH --evidence PATH [--json]
reqledger review [--json]
reqledger export --format json [--output PATH] [--json]

Global options: reqledger --version, reqledger --config PATH <command>.

Validation is fail-closed and returns exit code 1 on validation errors, 2 on usage/config errors, and 0 when all records are valid.

Configuration

ReqLedger looks for config in this order (first wins):

1. explicit --config PATH,
2. reqledger.toml discovered upward from the current directory,
3. .reqledger.toml discovered upward,
4. built-in defaults.

When both reqledger.toml and .reqledger.toml exist in the same directory, reqledger.toml wins. Paths declared in config resolve relative to the config file's directory; with no config they resolve relative to the current working directory.

Default config:

schema_version = 1

[paths]
root = "requirements"
records_dir = "requirements/records"
manifest = "requirements/manifest.json"
reports_dir = "requirements/reports"
reports_state_dir = "requirements/reports/reqledger"

[ids]
requirement_prefix = "REQ"
criterion_prefix = "AC"
width = 4

[review]
draft_stale_days = 90

Manifest and export

reqledger index writes a deterministic requirements/manifest.json:

{
  "schema_version": 1,
  "tool": "reqledger",
  "requirements": [
    {
      "id": "REQ-0001",
      "title": "Reject invalid login passwords",
      "path": "requirements/records/req-0001.req.md",
      "kind": "functional",
      "status": "accepted",
      "priority": "must",
      "tags": ["auth", "login"],
      "source": "manual",
      "source_refs": [],
      "criteria": [
        {
          "id": "AC-0001",
          "statement": "Login is rejected when an invalid password is submitted.",
          "verification": "behavior",
          "status": "accepted",
          "tags": ["auth"]
        }
      ],
      "refs": {
        "tasks": [],
        "architecture": [],
        "specs": [],
        "evidence": []
      }
    }
  ]
}

The manifest is derived state; Markdown records remain the source of truth. reqledger export --format json emits the same shape for downstream tools.

Design constraints

• Prefer readable files over hidden state.
• Keep deterministic output.
• Keep dependencies minimal (typer, ledgercore, tomli for Python < 3.11).
• Preserve user-authored Markdown bodies.
• Do not use title matching for identity; IDs are the identity.
• Validation is fail-closed.
• External links are references, not ownership transfers.
• ReqLedger does not parse or generate Gherkin, discover or run tests, or import SpecMason at runtime.

Development

pip install -e ".[dev]"
pytest -q

ReqLedger reuses shared Ledgerwerk primitives from ledgercore (ID allocation, config discovery, deterministic JSON, atomic writes) and implements the TOML +++ front matter and requirements domain itself.

License

Apache-2.0.