[FScholPer]

add reqirements resolve meta layer

Frank Scholter Peres frank.scholter_peres@mercedes-benz.com, Mercedes-Benz Tech Innovation GmbH
Provider Information

• Assisted by devin

[Copilot]

Copilot review overview

🔵 Needs a closer look

The documented and fallback requirements_projection command is not executable after the documented installation path.

Review effort: Lite
Findings: None

What changed in this PR

Adds deterministic Sphinx-Needs requirement projection and local MCP-based requirement lookup.

Changes:

• Adds projection, lookup, validation, and typed-link handling.
• Registers the resolve_requirements MCP tool.
• Updates packaging, documentation, instructions, and tests.

File	Description
packages/​context-discipline/​tests/​test_requirements_projection.py	Tests projection behavior
packages/​context-discipline/​tests/​test_requirements_lookup.py	Tests lookup behavior
packages/​context-discipline/​tests/​test_context_discipline_mcp.py	Tests MCP integration
packages/​context-discipline/​src/​requirements_projection.py	Builds deterministic requirement graphs
packages/​context-discipline/​src/​requirements_lookup.py	Resolves requirement IDs and links
packages/​context-discipline/​src/​context_discipline_mcp.py	Registers and dispatches the MCP tool
packages/​context-discipline/​README.md	Documents usage
packages/​context-discipline/​pyproject.toml	Updates package configuration
packages/​context-discipline/​mcp.yml	Declares the new tool
packages/​context-discipline/​.apm/​instructions/​working-memory-discipline.instructions.md	Adds resolution guidance

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[aryansingh0012]

packages/context-discipline/src/requirements_projection.py
_load_input() (lines 75–98):

Validates and unpacks needs from the JSON structure (line 94-97)
Returns only (document, digest) — the unpacked needs variable is discarded

build_requirements() (lines 145–251):

Calls _load_input() to get the document (line 151)
Then re-derives needs from scratch on lines 152–154:

The same navigation logic (document["versions"] → single version → version["needs"]) and type validation (isinstance(versions, dict), isinstance(needs, dict)) are performed twice. If one path needs to change later (e.g., to support multiple versions), there's a risk of drift.

Better approach: Have _load_input() return the needs dict directly, alongside the document and digest:

[FScholPer]

@odra can you also please check?

[odra]

@FScholPer it needs a rebase