Adds autocannon as a devDependency for the load-test tooling (T001), and the
shared schema migration T002 blocks: SLARun.version for optimistic
concurrency, plus two Postgres partial unique indexes
(assignments_one_current_per_ticket, escalation_events_ticket_rule_unique)
guarding against the assignment and escalation races research.md documents.
Applied directly to both the real dev DB and the throwaway test DB.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
User Story 3: initializes a real OpenTelemetry TracerProvider (previously
inert — getTracer() returned a no-op tracer with nothing ever exported).
Adds ticket.create, ai.escalation, and orchestration.assignment spans
covering both FR-006 cross-module paths, verified via a real, in-memory
test exporter that confirms actual trace/parent-span nesting, not mocked.
Also registers an AsyncLocalStorageContextManager
(@opentelemetry/context-async-hooks) — without one, OTel's context API is
a no-op that doesn't propagate across the await boundaries this feature's
own event-bus subscribers rely on for span nesting; caught by the first
version of the tracing integration test actually failing on real
parent/child assertions, not assumed.
Graceful degradation (FR-007) verified against a real, deliberately
unreachable OTLP endpoint: the SDK's own background export path (what
production actually exercises) never produces an unhandled rejection.
Starts on the 11 named business-health metrics: AI session
resolved/escalated outcomes (session.repository.ts, the single choke
point every branch in session.service.ts funnels through), human-vs-AI
resolution + resolution-time (a new TICKET_UPDATED/RESOLVED subscriber),
escalation rate (a new subscriber on ESCALATION_TRIGGERED, published
unconditionally since 008 but never previously consumed), and recurring
problems (tickets.service.ts's existing problem-creation call site).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
User Stories 1-2: every completed request (including 404s and early
replies from other hooks) now emits exactly one structured access-log
line, and every log line produced during that request's handling shares
its requestId/correlationId via a new AsyncLocalStorage-backed Pino mixin
— with zero changes to any existing log call site. The previously-dead
supporthub_http_request_duration_seconds histogram now actually receives
observations, so error rate and latency per route are computable from
/metrics alone.
Also bumps @opentelemetry/sdk-trace-base 1.x -> 2.x to align with the two
new tracing dependencies added in this same branch (exporter-trace-otlp-http,
resources) onto one consistent major version — npm had otherwise installed
two incompatible OTel core/resources majors side by side, which also
happened to resolve a moderate DoS advisory in @opentelemetry/core <2.8.0.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Replaces the no-op fastify.authenticate stub and the never-implemented
identity/auth login with real bcrypt password verification, JWT session
issuance/verification (reusing the existing JWT_SECRET), and a Redis-backed
revocation denylist for logout. Adds requireRole('ADMIN') to admin-only
configuration writes across 002-009 that previously relied on a decorator
that never actually checked anything. Adds self-identity (GET /auth/me,
re-validated against live account state) and admin-provisioned accounts
(POST /admin/users).
Making the auth check genuinely reject invalid/missing tokens exposed that
~18 pre-existing integration test files called already-gated routes with no
Authorization header (safe against the old no-op stub, broken against a real
one) — fixed via a shared tests/helpers/auth.ts (loginAs/authHeader) and a
file-by-file pass, plus two related SLA-run cleanup races exposed once admin
setup calls in those files' own beforeAll blocks started actually succeeding.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Phase 0 research resolves the business-calendar working-hours algorithm
(day-by-day walk via luxon, the first date/timezone dependency in this
codebase), the workingHours JSON shape, most-specific SLA-policy match
(reusing 005/006's resolution pattern), durable pause/resume (absolute
due-date shift, no in-memory state), and a repeatable-job breach-detection
design over per-run delayed jobs. Phase 1 adds data-model.md (one additive
refinement beyond doc06: SLARun.firstResponseBreachedAt), the admin/read
contract, and six quickstart scenarios including a genuine process-restart
boundary test for Constitution Principle VII.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Real Anthropic Claude integration per explicit product decision: a
ticket's AI session diagnoses the problem via a structured-output call,
applies a DB-configurable confidence-band policy (FR-005), and on
"proceed" reasons and acts through a small permission/risk-gated tool
system (FR-011/FR-012), optionally walking a matching runbook step by
step with the application — never the model — owning the step index
(FR-015/FR-016). Resolution requires real tool evidence, never customer
claims alone (FR-018) — verifyProductResolution is a documented
fail-closed placeholder mirroring the existing malware-scanner precedent,
since no real per-product operational signal exists yet.
AISupportSession.status mirrors onto Ticket.status through 003-ticketing's
existing AI_ANALYZING/AI_TROUBLESHOOTING/AI_VERIFYING/AI_RESOLVED/
HUMAN_ESCALATION state machine, discovered during planning to have been
built anticipating this exact feature. Two circular module dependencies
(escalation<->sessions, tools<->sessions) were designed around rather than
found as bugs: escalation is a pure summary formatter with no state
dependencies of its own, and tools stays a clean leaf module with zero
dependency on ai-support/sessions. Ticket creation enqueues the first
diagnosis turn via the existing queue infrastructure (off the hot path of
the inbound SaaS integration endpoint); a human actor changing ticket
status ends the AI session via the event-bus scaffold that existed in
this codebase but had never been wired to anything.
A real Prisma limitation was found and fixed before it reached tests:
compound-unique upsert rejects null for a nullable key column, so
AIConfidencePolicy uses find-then-update/create instead, same fix class
004 already used for the same underlying limitation.
Adds 9 unit tests (confidence-band, tool-policy-gate, runbook-step-
advance) and 6 integration test files, including the two constitution-
required standing E2E scenarios. AI-independent tests were run against
real Postgres/Redis/MinIO (88 passed, 0 failed across the full suite,
including every pre-existing 002/003/004 test). The AI-dependent tests
compile and skip cleanly via describe.skipIf but were not run against a
live model — no ANTHROPIC_API_KEY was available in this session; a real
key must be supplied before this feature can actually run.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
