UCR Modernization

Behavioral Test Capture from Retiring SMEs

Playwright (regression suite)FastAPIReactVitePostgres 16 + pgvectorVertex AIClaude Sonnet 4.6FHIR-aligned domain modelCloud RunIdentity-Aware ProxyWorkload Identity Federation

704

SME Knowledge Sessions

208

Behavioral Tests (v2)

32 / day

Questions Answered

Overview

DCHS staff spend their time on people, not data entry. The modernization is gated on first proving behavioral equivalence — every workflow Rikki documented over 36 years becomes an executable test the modernized UCR must pass before it ships. Knowledge transfer is happening live: the open-questions hub lets SMEs answer the platform team async, and answers automatically update the behavioral specs. 208 Playwright tests under qa_items_v2 prompt enforce sequential workflow + UCR hover-menu nav discipline (iterated from v1 after live QAT runs surfaced two structural issues). 703 historical sessions (ERDs, ETL specs, Jama stories, runbooks, meeting notes) ingested into the knowledge platform — queryable as structured rows with full provenance. Rikki's retirement-risk timer is unwinding: 32 tribal-knowledge questions answered in a single round, open count 44 → 27 — each answer locks a specific behavioral spec the modernized app must preserve. WS6 reconciliation themes (legacy-vs-modernized gaps) are now direct DB queries; WS6 code-vs-intent reconciliation SME-signed for downstream agents. Cloud Run + Cloud SQL + Identity-Aware Proxy + Workload Identity Federation infrastructure ready — all Terraform, awaiting GCP IAM grant.

Impact & Results

704

SME Knowledge Sessions

ERDs, ETL, Jama stories, runbooks, meetings — full provenance

208

Behavioral Tests Ready

Playwright suite (v2) at 98.1% compile-clean

32 / day

Tribal Questions Closed

via the SME open-questions hub

SME-Confirmed Sessions

authoritative sources the rebuild must honor

Years of Tribal Knowledge

captured before Rikki's retirement

Behavioral

Acceptance Contract

modernized UCR must pass 208/208 before ship

Key Features

Behavioral acceptance contract — 208 Playwright tests (98.1% compile-clean) become the criteria the modernized UCR must pass before shipping

Historical UCR corpus ingested — 703 sessions across ERDs, ETL specs, Jama stories, runbooks, meeting notes — queryable as structured rows with full provenance

Rikki retirement-risk timer unwinding — 32 tribal-knowledge questions answered in a single round; each answer locks a specific behavioral spec

WS6 code-vs-intent reconciliation SME-signed — authoritative for downstream agents (CODEX queries, TESTGEN v1 generation)

Open-questions hub — SMEs answer the platform team async; answers automatically update the behavioral specs they touch

Cloud Run + Cloud SQL + IAP + Workload Identity Federation deployment infrastructure ready (Terraform, awaiting IAM grant)

FHIR-aligned domain model under the hood — designed for human-services data semantics

Mock UI screens complete — 7 interactive prototypes for stakeholder review

Next Project

Content Automation (Linkit)