Show a recruiter or interviewer that this repo is a disciplined internal-product artifact, not a toy notebook and not fake enterprise theater.
For clone/install/test/artifact inspection steps, use the shorter reviewer path in docs/reviewer_guide.md.
make install PYTHON=python3.12
make runOptional:
make api
.venv/bin/python cli.py list-demo-casesStart with the scope language:
- deterministic
- bundled synthetic demo data; free-form input is not screened
- administrative readiness only
- no clinical judgment
- no approval prediction
Point out:
- bundled synthetic regression suite status
- policy drift status
- rulebook governance with reviewed vs active snapshots
- supported procedure registry with category, rule family, and provenance labels
This is a strong signal that the repo cares about reliability and governance, not just output formatting.
Use MRI-01-complete.
Call out:
- evidence mapping
- requirement-level reasoning
- audit trace
- deterministic letter drafting
Use MRI-08-edge-below-threshold.
Call out the difference between:
- present but below threshold ->
NOT_READY - missing or unclear -> not this case
- documented but outside configured result categories ->
NEEDS_REVIEW, not a threshold failure
Use CPAP-02-borderline.
Call out:
CANNOT_DETERMINEis deliberate- the system refuses to infer missing documentation
- this is safer than pretending certainty
Example:
.venv/bin/python cli.py evaluate --demo-case CPAP-02-borderline
curl http://127.0.0.1:8000/supported-proceduresThat makes the artifact feel like a compact internal product instead of a UI-only demo.
Use MRI-CERV-01-ready.
Call out:
- the engine was not rewritten to add it
- the same deterministic extraction contract was reused
- the procedure registry now surfaces rule family and provenance metadata
- this is depth, not platform theater
Use MRI-KNEE-01-ready.
Call out:
- this is a different clinical domain but still a narrow administrative contract
- only one new extractor field was added
- prior imaging is now a documented threshold, not just a missingness check
- the repo still avoids medical-necessity scoring or approval prediction
Use:
.venv/bin/python cli.py rulebook-diff --from-release 2026-04-09-reviewed-v0.4 --to-release 2026-07-17-active-v0.6Call out:
- a human can see exactly what changed between reviewed and active rulebooks
- drift monitoring remains separate from promotion
- this is a compact governance story without pretending to be a platform
- "This checks administrative readiness, not approval likelihood."
- "The system is narrow on purpose so every output can be defended."
- "Missing data produces refusal, not fake confidence."
- "The same deterministic workflow powers the UI, API, CLI, and exported artifacts."
- "The third pass added a non-spine pathway, a versioned rulebook, and golden acceptance checks without widening the product claim."