Digital Passport Model Assessment Workbench

A Python toolkit and thin CLI for SHACL-based assessment of digital passport data models, pairwise comparison of composed solutions, and SHACL-only use-case coverage analysis.

This README is primarily a repository and development entry point. The user-facing conceptual and usage reference is at the published documentation site. The repository-source docs remain available in docs/.

What this repository contains

• A publishable Python package with a CLI entry point: dpawb
• A thin stdio MCP server entry point: dpawb-mcp
• Contract-aligned input schemas for composition profiles, use cases, and alignments
• Built-in vocabularies and templates exposed through discovery commands
• Synthetic local fixtures used only for tests and validation
• A progression-based examples tree for human and AI-agent tutorials

Tool Shape

The tool is designed primarily as an agent-usable analytical toolkit with a thin CLI. The core command surface is:

• assess
• coverage
• compare
• prioritize
• recommend-composition
• schema
• vocabulary
• template
• capabilities
• summarize

AI-agent use

The intended primary integration mode is an AI agent orchestrating the analytical pipeline through the package API or the CLI. In practice, this means:

• deterministic file-based inputs
• JSON outputs suitable for agent parsing and chaining
• explicit analytical steps instead of a chat-oriented interface

The CLI remains useful for direct human invocation, but the main product shape is an analytical engine that can be called by agent skills, workflow runners, or tool adapters.

The package also ships a thin stdio MCP server over the same deterministic command surface:

• assess
• coverage
• compare
• prioritize
• recommend-composition
• schema
• vocabulary
• template
• capabilities
• summarize

The MCP surface stays intentionally thin:

• keep dpawb as the core Python package
• expose the same operations through dpawb-mcp
• ship the MCP runtime as both a local Python entry point and a GitHub Container Registry image
• publish registry metadata from the GitHub mirror

The MCP runtime and publication shape are documented in the published MCP server page.

The public Python API is documented in the published API reference.

MCP Access

The MCP server is published with this identity:

• registry name: io.github.CE-RISE-software/dpawb
• OCI image pattern: ghcr.io/ce-rise-software/dpawb-mcp:<release-version>
• transport: stdio
• official registry base: https://registry.modelcontextprotocol.io/

The server is discoverable in the official MCP Registry under the name io.github.CE-RISE-software/dpawb.

Registry discovery page:

• https://registry.modelcontextprotocol.io/?q=dpawb

Local client configuration example:

{
  "mcpServers": {
    "dpawb": {
      "command": "dpawb-mcp"
    }
  }
}

OCI-oriented registry metadata is declared in server.json. The publish workflow fills in the release version and image tag when publishing.

Install

python -m venv .venv
. .venv/bin/activate
pip install -e .

Published package name:

pip install dpawb

The installed MCP server command is:

dpawb-mcp

If you are working in a restricted environment, the package is configured to build with setuptools so editable installs do not depend on fetching an extra build backend.

If build isolation or wheel support is unavailable locally, use:

python -m pip install --no-build-isolation -e .

In constrained environments where editable installation is blocked by local Python packaging tooling, the repository still supports a repo-native execution path:

./scripts/run-local.sh capabilities
./scripts/test-local.sh
make smoke
make test
make validate

Minimal local MCP runtime command:

python -m dpawb.mcp_server

Example commands

dpawb assess --profile fixtures/profiles/synthetic_evolution_latest.yaml
dpawb coverage --profile fixtures/profiles/synthetic_evolution_latest.yaml --use-case fixtures/use_cases/product_identity_lookup.yaml
dpawb compare --left left_assessment.json --right right_assessment.json
dpawb recommend-composition --left left_assessment.json --right right_assessment.json --comparison comparison_result.json
dpawb summarize --result comparison_result.json --format markdown
dpawb capabilities

Local Release Check

Before wiring PyPI CI/CD, run the local packaging check in an environment with wheel available:

make release-check

This builds a wheel and sdist, installs the wheel into a clean temporary environment, and runs installed CLI smoke checks.

Structure

There are two distinct input areas in this repository:

• fixtures/ Synthetic, repository-local test data only. These files are used by tests, smoke checks, and CI validation.

• examples/01-source-ingestion/ Live-source example inputs intended for manual runs. These are the first step in the tutorial progression and are not part of CI validation.

The full examples tree is organized by analytical task, not by input data model:

• examples/01-source-ingestion/ Load and assess declared profiles.
• examples/02-structural-comparison/ Compare two profile assessment results.
• examples/03-reduced-use-case-comparison/ Run the first aligned use-case comparison.
• examples/04-extended-use-case-comparison/ Run a broader aligned use-case comparison.
• examples/05-composition-recommendation/ Recommend a combined profile from two assessed model sets.

Each example is usable by humans as a step-by-step command tutorial and by AI agents as a deterministic recipe over explicit files.

Source-ingestion examples

Example profiles for live SHACL sources are included at:

• examples/01-source-ingestion/profiles/battery_dpp_representation_live.yaml
• examples/01-source-ingestion/profiles/battery_product_identification_live.yaml
• examples/01-source-ingestion/profiles/dp_record_metadata_live.yaml
• examples/01-source-ingestion/profiles/traceability_and_life_cycle_events_live.yaml
• examples/01-source-ingestion/profiles/metadata_focused_composition_live.yaml
• examples/01-source-ingestion/profiles/metadata_and_traceability_live.yaml
• examples/02-structural-comparison/profiles/metadata_slice_left_live.yaml
• examples/02-structural-comparison/profiles/metadata_slice_right_live.yaml

If you want a single live source, the metadata-oriented example is the main starting point:

./scripts/run-local.sh assess --profile examples/01-source-ingestion/profiles/dp_record_metadata_live.yaml

If you want a composed profile, use:

./scripts/run-local.sh assess --profile examples/01-source-ingestion/profiles/metadata_and_traceability_live.yaml

You can also run the traceability-only example:

./scripts/run-local.sh assess --profile examples/01-source-ingestion/profiles/traceability_and_life_cycle_events_live.yaml

For manual coverage runs, example use cases are included at:

• examples/01-source-ingestion/use_cases/battery_dpp_representation.yaml
• examples/01-source-ingestion/use_cases/battery_product_identification.yaml
• examples/01-source-ingestion/use_cases/battery_passport_metadata_and_classification.yaml
• examples/01-source-ingestion/use_cases/record_identity_lookup.yaml
• examples/01-source-ingestion/use_cases/provenance_actor_lookup.yaml

Example:

./scripts/run-local.sh coverage \
  --profile examples/01-source-ingestion/profiles/dp_record_metadata_live.yaml \
  --use-case examples/01-source-ingestion/use_cases/record_identity_lookup.yaml

The main real comparison-driver use case is:

• examples/01-source-ingestion/use_cases/battery_dpp_representation.yaml

It intentionally stays narrow. It requires:

• passport identity
• battery identity
• passport version or revision
• one responsible actor
• one battery type or classification signal

and the joins needed to treat those as one battery-DPP representation slice.

The matching starting composition for that use case is:

• examples/01-source-ingestion/profiles/battery_dpp_representation_live.yaml

It currently composes:

• dp_record_metadata
• traceability_and_life_cycle_events

This is the current broader baseline for the battery-DPP comparison work.

For the first reduced real pass, the narrower identity-focused comparison slice is:

• canonical example folder: examples/03-reduced-use-case-comparison/
• use case: examples/03-reduced-use-case-comparison/use_cases/use_case.yaml
• left profile: examples/03-reduced-use-case-comparison/profiles/left_profile.yaml
• right profile: examples/03-reduced-use-case-comparison/profiles/right_profile.yaml

This reduced slice composes:

• dp_record_metadata
• product_profile
• traceability_and_life_cycle_events

and is the first validated product-identification comparison slice against the BatteryPass General Product Information model.

A second broader validated slice is also included: