Sprint demos

End-to-end demonstrations recorded as build-tagged Go tests so they are reproducible, asserted, and easy to embed in sprint reviews. Each demo prints a structured transcript that doubles as the acceptance record for its sprint.

Running

make demo-sprint1        # Sprint 1: identity & authentication
make demo-sprint2        # Sprint 2: RBAC
make demo-sprint3        # Sprint 3: provenance + transparency log
make demo-sprint4        # Sprint 4: backup + restore + retention
make demo-sprint5        # Sprint 5: health probes + breakers + degradation

The build tag (-tags=demo) keeps these out of the regular go test runs so CI stays fast.

Recording an asciinema cast

For visual sprint reviews:

asciinema rec docs/demos/sprint-01.cast -- bash -lc "nix develop --command make demo-sprint1"
asciinema upload docs/demos/sprint-01.cast    # optional, when sharing externally

The .cast files are JSON and play back deterministically; commit them next to the transcript so reviewers can pick either format.

Transcripts

File	Sprint	Scope
sprint-01-transcript.txt	1	Register → login (pwd+TOTP) → JWT → API call → JWKS verify → revoke
sprint-02-transcript.txt	2	4 baseline roles matrix → in-process mTLS-cn → hot-reload → analyst recipe → malformed-keeps-prior
sprint-03-transcript.txt	3	sign event → JWKS verify → Merkle append → STH → inclusion proof → tamper-detect → consistency proof
sprint-04-transcript.txt	4	migrate → seed → backup (age + sidecar) → tamper → restore with journal STH → retention sweep
sprint-05-transcript.txt	5	health probes → graceful degradation → backoff+retry → circuit breaker trip/recover → ctx-cancel exclusion

Anatomy of a demo test

Each internal/identity/demo/*_test.go (or future internal/*/demo/*_test.go) follows the same shape:

• //go:build demo so it stays out of CI.
• Banner header (╔══...╗) and per-step banners (STEP N — …) so the transcript reads top-to-bottom like a narrative.
• must(t, err, "…") to fail loudly if any step regresses.
• A final Acceptance checklist mirroring the ADR's acceptance criteria so reviewers can tick boxes mechanically.

When sprint demos regress, the test fails and CI (eventually) refuses the regression. Until CI runs them, run them locally as part of sprint close.