<?xml version="1.0" encoding="utf-8"?>
<entry xmlns="http://www.w3.org/2005/Atom" xml:base="https://docs.spec-kitty.ai/adr/4.x/2026-10-01-2-mission-status-read-api-and-dashboard-extraction/">
  <id>https://docs.spec-kitty.ai/adr/4.x/2026-10-01-2-mission-status-read-api-and-dashboard-extraction/</id>
  <title>ADR: a mission status read API replaces the CLI-bundled dashboard (extract and replace)</title>
  <updated>2026-10-30T00:00:00.000Z</updated>
  <author><name>Spec Kitty</name></author>
  <link rel="alternate" type="text/html" href="https://docs.spec-kitty.ai/adr/4.x/2026-10-01-2-mission-status-read-api-and-dashboard-extraction/"/>
  <summary>Remove the bundled dashboard; keep one read-only mission status API (overview + detail) consumed by external UIs (#645). Accepted 2026-10-12 as the open-core producer surface.</summary>
  <content type="html">&lt;p&gt;&lt;strong&gt;Status:&lt;/strong&gt; Accepted (2026-10-12). D-1 and D-2 recorded a direction the operator stated on
2026-10-01 and were &lt;strong&gt;Accepted&lt;/strong&gt; then. D-8 was &lt;strong&gt;amended and accepted&lt;/strong&gt; later on 2026-10-01:
the operator moved deletion to the front (see D-8). D-3 to D-7 began as architect amendments
awaiting ratification; the Amendment 2026-10-03 settled the Java-service shape of D-4/D-5/D-7,
and the &lt;strong&gt;Amendment 2026-10-12&lt;/strong&gt; below ratifies the whole record: it frames the read API as
the Build Kitty / open-core &lt;strong&gt;producer&lt;/strong&gt; surface, keeps it in-repo now and extracts it when
its contract freezes, and keeps &lt;code dir=&quot;auto&quot;&gt;contracts/mission-status/&lt;/code&gt; and the conformance fixtures
canonical beside the Python reducer. The record moves from Proposed to Accepted with that
amendment. Amended by &lt;a href=&quot;/adr/4.x/2026-10-29-1-the-events-domain-and-its-java-services-live-in-spec-kitty-events/&quot;&gt;ADR 2026-10-29-1&lt;/a&gt; (the events services live in spec-kitty-events); see Amendment 2026-10-29.
&lt;strong&gt;Amendment 2026-10-03 (Accepted):&lt;/strong&gt; D-4 and D-7 are amended for the agreed Java service
design. D-5 records that the CLI’s own read verbs stay native Python behind one read seam
for now. See “Amendment 2026-10-03” below.
&lt;strong&gt;Amendment 2026-10-12 (Accepted):&lt;/strong&gt; the read API is open-core (producer side), not a Team
Kitty surface; in-repo now, extracted when frozen. See “Amendment 2026-10-12” below.
&lt;strong&gt;Amendment 2026-10-30 (Accepted):&lt;/strong&gt; factual wording amendment (#6604): D-5’s “Lamport”
ordering phrases describe the superseded record. The status path is ordered by wall clock
&lt;code dir=&quot;auto&quot;&gt;(at, event_id)&lt;/code&gt; with the approved-departure lift, per ADR 2026-02-09-3 Amendment
2026-10-22. See “Amendment 2026-10-30” below.&lt;/p&gt;
&lt;p&gt;The inline ‘(Proposed)’ labels on D-3..D-7 are historical; all are Accepted as of 2026-10-12.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Date:&lt;/strong&gt; 2026-10-01 (amended 2026-10-03, 2026-10-12)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Deciders:&lt;/strong&gt; Stijn Dejongh (operator). Analysis by &lt;code dir=&quot;auto&quot;&gt;architect-alphonso&lt;/code&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Technical Story:&lt;/strong&gt; epic &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/645&quot;&gt;#645&lt;/a&gt;
(Stable Application API Surface), its children #956 (registry and cache), #957 (mission and
WP resources, &lt;code dir=&quot;auto&quot;&gt;WorkPackageAssignment&lt;/code&gt;), #2789 (&lt;code dir=&quot;auto&quot;&gt;--since&lt;/code&gt; cursor reads) and #460 (transport);
dashboard defects #4520, #4767, #4768 and #4769; roadmap
&lt;a href=&quot;/plans/4-0-0-milestone-roadmap/&quot;&gt;direction update 2026-10-01&lt;/a&gt;.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id=&quot;context-and-problem-statement&quot;&gt;Context and Problem Statement&lt;/h2&gt;
&lt;p&gt;The CLI ships a local web dashboard in &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/dashboard/&lt;/code&gt;: about 9.4k lines of
Python, JavaScript, CSS and HTML. &lt;code dir=&quot;auto&quot;&gt;spec-kitty dashboard&lt;/code&gt; starts it as a detached daemon. The
daemon records itself in &lt;code dir=&quot;auto&quot;&gt;.kittify/.dashboard&lt;/code&gt;, scans for a free port and kills stale
processes. The browser polls &lt;code dir=&quot;auto&quot;&gt;/api/features&lt;/code&gt; once a second
(&lt;code dir=&quot;auto&quot;&gt;static/dashboard/dashboard.js:1749&lt;/code&gt;). Each poll rescans every mission and re-reduces each
&lt;code dir=&quot;auto&quot;&gt;status.events.jsonl&lt;/code&gt;, with no cache (&lt;code dir=&quot;auto&quot;&gt;scanner.py:895&lt;/code&gt;, #4520: 9–10 s per scan on this
repository’s 436 missions). The daemon has open trust defects: it SIGKILLs a recycled PID
(#4767), leaks its shutdown token through &lt;code dir=&quot;auto&quot;&gt;/api/health&lt;/code&gt; (#4768), and does no Host or Origin
check, so DNS rebinding can read the whole API (#4769).&lt;/p&gt;
&lt;p&gt;The bigger problem is architectural. Mission and work package (WP) state reaches consumers
along several independent paths, and nothing checks that they agree (#645, “Why”):&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The dashboard computes mission state in its presentation layer. &lt;code dir=&quot;auto&quot;&gt;_derive_mission_status&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;_derive_next_action&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;build_mission_registry&lt;/code&gt; live in &lt;code dir=&quot;auto&quot;&gt;dashboard/scanner.py:219&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;:249&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;:565&lt;/code&gt;. &lt;code dir=&quot;auto&quot;&gt;dashboard --json&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;cli/commands/dashboard.py:44&lt;/code&gt;) serves the same
registry.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;orchestrator-api mission-state&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;list-ready&lt;/code&gt; reduce through &lt;code dir=&quot;auto&quot;&gt;reduce(read_events(...))&lt;/code&gt;
(&lt;code dir=&quot;auto&quot;&gt;orchestrator_api/commands.py:1165&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;:1226&lt;/code&gt;). That function is the wall-clock
&lt;code dir=&quot;auto&quot;&gt;reduce_parsed&lt;/code&gt; reducer (&lt;code dir=&quot;auto&quot;&gt;status/reducer.py:175-183&lt;/code&gt;). The dashboard reads through
&lt;code dir=&quot;auto&quot;&gt;materialize_snapshot&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;status/reducer.py:367&lt;/code&gt;), which is the Lamport &lt;code dir=&quot;auto&quot;&gt;reduce_shared_state&lt;/code&gt;
reduction. Two external read surfaces therefore use the two different reducers that
CLAUDE.md’s “reducer duality” note describes. &lt;em&gt;Inferred:&lt;/em&gt; they can disagree when events
are concurrent (#4941).&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;agent tasks status --json&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;cli/commands/agent/tasks.py:1419&lt;/code&gt;) builds a third board.&lt;/li&gt;
&lt;li&gt;The living architecture docs already record the gap: “Dashboard reads filesystem directly
rather than through Event Store interface”
(&lt;a href=&quot;/architecture/04_implementation_mapping/&quot;&gt;implementation mapping&lt;/a&gt;,
Divergence Notes 4).&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;On 2026-10-01 the operator proposed two things. First, remove the dashboard but keep a read
path to mission status. Second, expose that read path at two granularities: a high-level
status built from the ledger and events, and a detail view with WP content, status and
assignment. An external dashboard can then pick the granularity it needs.&lt;/p&gt;
&lt;h2 id=&quot;decision-drivers&quot;&gt;Decision Drivers&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Single canonical authority&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;DIRECTIVE_044&lt;/code&gt;): one read surface, backed by one service
layer, which every consumer reads through (#645, “Intended effect”).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Bounded contexts&lt;/strong&gt; (&lt;code dir=&quot;auto&quot;&gt;DIRECTIVE_031&lt;/code&gt;): status reads belong to the status domain, not to a
UI. Governance and glossary views belong to their own contexts.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The adapter-seam rule&lt;/strong&gt; (roadmap direction update, rule 2): the CLI core does not talk to
external systems. A UI is a consumer, not part of the core.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Local-first, read-only:&lt;/strong&gt; the read path must work offline, with hosted drain off, and
must never write.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Modulith first:&lt;/strong&gt; keep the module in-process today, with a boundary that can be
extracted later into another package or repository.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;considered-options&quot;&gt;Considered Options&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Keep and harden the bundled dashboard.&lt;/strong&gt; Fix #4767–#4769, add the #956 cache and adopt
FastAPI (#460).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Remove the dashboard and let consumers read files and &lt;code dir=&quot;auto&quot;&gt;events tail&lt;/code&gt;.&lt;/strong&gt; Ship no read API.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Remove the dashboard and keep a mission status read API with two granularities,
in-process plus CLI &lt;code dir=&quot;auto&quot;&gt;--json&lt;/code&gt;&lt;/strong&gt; (chosen).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Remove the dashboard and ship a mission status HTTP service in the CLI.&lt;/strong&gt; This would be
a slimmer server that keeps the daemon.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;decision-outcome&quot;&gt;Decision Outcome&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Chosen option:&lt;/strong&gt; Option 3. It meets every driver. It keeps the operator’s two granularities
and moves the logic the dashboard holds today behind the status boundary. It also drops the
daemon and the poll-and-rescan cost from the CLI.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-1 (Accepted): remove the CLI-bundled dashboard.&lt;/strong&gt; The server, daemon, static UI,
&lt;code dir=&quot;auto&quot;&gt;spec-kitty dashboard&lt;/code&gt; command, &lt;code dir=&quot;auto&quot;&gt;/spec-kitty.dashboard&lt;/code&gt; slash command and the
&lt;code dir=&quot;auto&quot;&gt;spk-admin-dashboard&lt;/code&gt; skill are removed. A replacement UI is built in its own repository
as an ordinary consumer. The original plan was extract-and-replace: move the domain logic
in &lt;code dir=&quot;auto&quot;&gt;dashboard/scanner.py&lt;/code&gt; into the read API and rehome every route before deleting
anything. The amended D-8 deletes first instead; the read API and the route rehoming then
build the replacement read path from the remaining read models, and the deleted code
stays recoverable from git history.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-2 (Accepted): keep a read path, the Mission Status Read API, at two granularities.&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Overview&lt;/strong&gt; (low detail, many missions): for each mission, its identity (&lt;code dir=&quot;auto&quot;&gt;mission_id&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;mid8&lt;/code&gt;, slug, friendly name, display number), lane counts, weighted progress, a derived
lifecycle status, the latest event, and a change cursor.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Detail&lt;/strong&gt; (one mission): each WP’s authored plan and resolved runtime state, kept
separate. That covers lane, assignment (agent, profile, role, model), dependencies and
readiness, review evidence, and the WP’s event history. The WP prompt body is returned
only on request.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-3 (Proposed): each granularity maps onto a read model that already exists. The API
composes these models and never re-derives them.&lt;/strong&gt;&lt;/p&gt;




















&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Granularity&lt;/th&gt;&lt;th&gt;Reads&lt;/th&gt;&lt;th&gt;Existing read model&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;Overview&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;meta.json&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;status.events.jsonl&lt;/code&gt; only, no WP files&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;materialize_snapshot&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;status/reducer.py:367&lt;/code&gt;) gives lane summary; &lt;code dir=&quot;auto&quot;&gt;compute_weighted_progress&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;status/progress.py:119&lt;/code&gt;); identity resolution; &lt;code dir=&quot;auto&quot;&gt;TailCursor&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;status/tail_reader.py:60&lt;/code&gt;) gives the change cursor&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;Detail&lt;/td&gt;&lt;td&gt;plus &lt;code dir=&quot;auto&quot;&gt;tasks/WP*.md&lt;/code&gt; frontmatter and body&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;reconstruct_wp_view&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;status/wp_view.py:256&lt;/code&gt;), which never conflates authored and resolved state; the dependency graph and &lt;code dir=&quot;auto&quot;&gt;dependency_readiness_for_wp&lt;/code&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;p&gt;Every read goes through the coordination-aware status read surface: the
&lt;code dir=&quot;auto&quot;&gt;_read_path_resolver&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;MissionStatus.load&lt;/code&gt; path. Callers never rebuild paths. The
contract carries a schema version, as #2789 requires. Its types must not reuse the name
&lt;code dir=&quot;auto&quot;&gt;MissionStatus&lt;/code&gt;, because the aggregate root at &lt;code dir=&quot;auto&quot;&gt;status/aggregate.py:168&lt;/code&gt; already owns it.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-4 (Proposed): the first form is an in-process query module plus CLI &lt;code dir=&quot;auto&quot;&gt;--json&lt;/code&gt; verbs.
It is not an HTTP endpoint.&lt;/strong&gt; The API has one typed, versioned query function per
granularity. Each function has one CLI &lt;code dir=&quot;auto&quot;&gt;--json&lt;/code&gt; verb that a non-Python consumer can call
as a subprocess. The CLI wheel ships no server, daemon, port or PID file. Choosing a
transport such as HTTP, OpenAPI or a change feed belongs to the consumer repository
(#460 moves with it). A consumer running a server imports the module or shells out to
the verb. Change detection comes from polling the overview cursor and then
&lt;code dir=&quot;auto&quot;&gt;events tail --json&lt;/code&gt; (#3858). It does not come from the in-process status fan-out, which
runs only while a CLI command runs.
&lt;em&gt;Amended 2026-10-03:&lt;/em&gt; the external form is now a separate read-only service. The
in-process module remains the CLI’s own read path, behind one seam. See
“Amendment 2026-10-03”.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-5 (Proposed): fold existing surfaces into the API instead of duplicating them.&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The overview verb replaces the deleted &lt;code dir=&quot;auto&quot;&gt;dashboard --json&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;agent tasks status --json&lt;/code&gt; renders from the detail query.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;orchestrator-api mission-state&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;list-ready&lt;/code&gt; keep their verbs and envelopes (the
contract is owned by #5231) but compute through the read API. This also moves them onto
the Lamport reducer.&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;events tail --json&lt;/code&gt; stays as the change feed, and the overview exposes its cursor.&lt;/li&gt;
&lt;li&gt;The TypedDicts in the deleted &lt;code dir=&quot;auto&quot;&gt;dashboard/api_types.py&lt;/code&gt; (recover from git history) seed the contract types, renamed to the
Mission canon (#957 shapes): &lt;code dir=&quot;auto&quot;&gt;/api/features&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;FeatureItem&lt;/code&gt; become missions.&lt;/li&gt;
&lt;li&gt;An architectural gate fails any display or external consumer that reduces status itself,
using the shrink-only allowlist pattern (#645 acceptance). Domain-internal reducers such
as consolidation, migrations and gates are out of scope for that gate.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-6 (Proposed): the API is read-only and independent of drain.&lt;/strong&gt; It never calls the
writing &lt;code dir=&quot;auto&quot;&gt;materialize()&lt;/code&gt;, never refreshes derived views, never takes a lock, and returns the
same answer whether hosted drain is on or off. It is a pull over the committed ledger. The
produce/drain decoupling is push-side and gets its own ADR (roadmap next step 3). Neither
decision blocks the other.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-7 (Proposed): placement in the layer chain.&lt;/strong&gt; The module lives in the &lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt;
layer, next to the status domain it reads. The read models it composes are there already
(&lt;code dir=&quot;auto&quot;&gt;status/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;missions/_read_path_resolver.py&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;mission_metadata&lt;/code&gt;). Placing it in
&lt;code dir=&quot;auto&quot;&gt;runtime&lt;/code&gt; or &lt;code dir=&quot;auto&quot;&gt;mission_runtime&lt;/code&gt; would add edges to their shrink-only outbound ledgers
(&lt;code dir=&quot;auto&quot;&gt;tests/architectural/test_layer_rules.py&lt;/code&gt;). The contract types depend only on the
standard library, so they can be extracted later into a client package without bringing
the CLI along.
&lt;em&gt;Amended 2026-10-03:&lt;/em&gt; the in-process module and its seam stay here. The external service
lives in its own JVM module in this repository, outside the Python wheel. See
“Amendment 2026-10-03”.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;D-8 (Amended and accepted 2026-10-01): order of work. Delete first.&lt;/strong&gt; The operator
reversed the original order, which put deletion last, after the read API, the reader
re-pointing and the route rehoming. Reason: the dashboard carries open P1 security
defects (#4767, #4769) and its replacement does not gate 4.0.0 GA. If the new UI does not
arrive in time, the code is restored from git.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Delete &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/dashboard/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;spec-kitty dashboard&lt;/code&gt;, the slash command and the
skill, with an upgrade migration that removes the installed command files and the
dashboard’s runtime files, touching only paths it can prove are package-owned (version
marker or managed path; charter, “User Customization Preservation”), and stops a
dashboard server an older CLI left running when its command line proves it (#5530).&lt;/li&gt;
&lt;li&gt;Build the Mission Status Read API and its contract tests (#5528).&lt;/li&gt;
&lt;li&gt;Re-point the orchestrator-api reads and &lt;code dir=&quot;auto&quot;&gt;tasks status&lt;/code&gt; to the API (#5532).&lt;/li&gt;
&lt;li&gt;Assign read surfaces to the former artifact and governance routes (#5533).&lt;/li&gt;
&lt;li&gt;Land the bypass gate.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;code dir=&quot;auto&quot;&gt;verify-setup --diagnostics&lt;/code&gt; imported &lt;code dir=&quot;auto&quot;&gt;dashboard.diagnostics&lt;/code&gt;; that module moved to
&lt;code dir=&quot;auto&quot;&gt;specify_cli/diagnostics/project.py&lt;/code&gt; without its dashboard health probe. The docs site’s
glossary page assets moved to &lt;code dir=&quot;auto&quot;&gt;scripts/docs/glossary_page/&lt;/code&gt;. There is no deprecation
release. None of this gates 4.0.0 GA.&lt;/p&gt;
&lt;p&gt;The original order (Proposed, superseded): build the API, re-point readers and dashboard
handlers, land the bypass gate, deprecate &lt;code dir=&quot;auto&quot;&gt;spec-kitty dashboard&lt;/code&gt; for one release, run the
removal migration, then delete the package.&lt;/p&gt;
&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Where the deleted dashboard’s routes go&lt;/strong&gt; (former &lt;code dir=&quot;auto&quot;&gt;dashboard/handlers/router.py&lt;/code&gt;). With
deletion first, these are the homes the follow-ups build; until then the routes have no
replacement:&lt;/p&gt;





























&lt;table&gt;&lt;thead&gt;&lt;tr&gt;&lt;th&gt;Route&lt;/th&gt;&lt;th&gt;New home&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;/api/features&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/kanban/&amp;#x3C;id&gt;&lt;/code&gt;&lt;/td&gt;&lt;td&gt;This API: overview and detail&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;/api/research/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/contracts/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/checklists/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/artifact/&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/dossier/&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Mission artifact reads. Planning artifacts are files, not events. A follow-up assigns their read surface (mission dossier or artifact placement); this ADR does not&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;/api/charter&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/charter-lint&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/glossary-health&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/glossary-terms&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/glossary&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Governance and glossary contexts (#954, #955). Not this API&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;/api/diagnostics&lt;/code&gt;&lt;/td&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;verify-setup --diagnostics&lt;/code&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td&gt;&lt;code dir=&quot;auto&quot;&gt;/api/health&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/shutdown&lt;/code&gt;&lt;/td&gt;&lt;td&gt;Removed with the daemon&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt;
&lt;h3 id=&quot;consequences&quot;&gt;Consequences&lt;/h3&gt;
&lt;h4 id=&quot;positive&quot;&gt;Positive&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;One read authority replaces four ad-hoc readers that use two different reducers.&lt;/li&gt;
&lt;li&gt;The CLI loses a daemon, its PID-file and kill-sweep defects (#4767–#4769), and the per-poll
rescan (#4520). Those issues close when the dashboard is deleted, not by hardening it.&lt;/li&gt;
&lt;li&gt;External UIs, MCP adapters and SDKs get a documented, versioned contract and can choose a
granularity.&lt;/li&gt;
&lt;li&gt;The boundary is ready for extraction: contract types without dependencies, in-process today.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;negative&quot;&gt;Negative&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;Until a replacement UI ships, users have no browser kanban. &lt;code dir=&quot;auto&quot;&gt;agent tasks status&lt;/code&gt; is the
interim view.&lt;/li&gt;
&lt;li&gt;Removal ships without a deprecation release; an upgrade migration cleans the 17 agent
surfaces.&lt;/li&gt;
&lt;li&gt;Until the read API lands, external readers have &lt;code dir=&quot;auto&quot;&gt;agent tasks status --json&lt;/code&gt;,
&lt;code dir=&quot;auto&quot;&gt;orchestrator-api mission-state&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;events tail --json&lt;/code&gt;, which still use the two
reducers the Context section describes.&lt;/li&gt;
&lt;li&gt;Mission artifact and governance views lose their only UI until their own read surfaces land.&lt;/li&gt;
&lt;/ul&gt;
&lt;h4 id=&quot;neutral&quot;&gt;Neutral&lt;/h4&gt;
&lt;ul&gt;
&lt;li&gt;The &lt;code dir=&quot;auto&quot;&gt;orchestrator-api&lt;/code&gt; lifecycle verbs and their contract are unchanged.&lt;/li&gt;
&lt;li&gt;The landscape’s Dashboard container stays as a concept. Only its in-repo implementation
leaves (&lt;a href=&quot;/architecture/00_landscape/&quot;&gt;landscape&lt;/a&gt;).&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id=&quot;confirmation&quot;&gt;Confirmation&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;Contract tests pin the overview and detail shapes and their schema version.&lt;/li&gt;
&lt;li&gt;On the same fixture mission, the orchestrator-api reads, &lt;code dir=&quot;auto&quot;&gt;tasks status --json&lt;/code&gt; and the
overview and detail queries agree.&lt;/li&gt;
&lt;li&gt;The bypass gate stays green and its allowlist only shrinks.&lt;/li&gt;
&lt;li&gt;The overview of a mission reads no &lt;code dir=&quot;auto&quot;&gt;tasks/&lt;/code&gt; files (asserted by a test).&lt;/li&gt;
&lt;li&gt;&lt;code dir=&quot;auto&quot;&gt;src/specify_cli/dashboard/&lt;/code&gt; is gone and &lt;code dir=&quot;auto&quot;&gt;pyproject.toml&lt;/code&gt; ships no server dependency.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;amendment-2026-10-03-accepted-a-separate-java-read-service-the-cli-stays-native-behind-one-seam&quot;&gt;Amendment 2026-10-03 (Accepted): a separate Java read service; the CLI stays native behind one seam&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Deciders:&lt;/strong&gt; Stijn Dejongh (operator), 2026-10-03. The service design was agreed with the
UI contributors and recorded on
&lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/5528#issuecomment-5937187090&quot;&gt;#5528&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Why.&lt;/strong&gt; External consumers (the new dashboard in &lt;code dir=&quot;auto&quot;&gt;spec-kitty-mission-ui&lt;/code&gt; and Kitty Desktop)
need a browser-friendly transport with a change stream, and the UI team builds both sides.
Option 4 was rejected because it put a daemon &lt;em&gt;inside the CLI&lt;/em&gt;. A separately built and
separately released binary keeps the wheel server-free and keeps that trust model out of the
CLI core.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;D-4 (amended).&lt;/strong&gt; The external form of the Mission Status Read API is a detached,
read-only HTTP service: Java 25, Spring Boot 4 (Spring MVC on virtual threads),
distributed as a GraalVM native binary per OS. It serves REST resources for the two D-2
granularities (&lt;code dir=&quot;auto&quot;&gt;/api/v1/missions&lt;/code&gt;, &lt;code dir=&quot;auto&quot;&gt;/api/v1/missions/{id}&lt;/code&gt;), an SSE change stream on
&lt;code dir=&quot;auto&quot;&gt;/api/v1/events&lt;/code&gt;, and an OpenAPI 3.1 document. It binds to &lt;code dir=&quot;auto&quot;&gt;127.0.0.1&lt;/code&gt; only, checks Host
and Origin, and puts no secrets in responses (the #4767–#4769 lessons). It may serve a
pinned, checksummed dashboard bundle at &lt;code dir=&quot;auto&quot;&gt;/&lt;/code&gt;. The contract is written first
(&lt;code dir=&quot;auto&quot;&gt;contracts/mission-status/openapi.yaml&lt;/code&gt;, #5558); the service and the UI both build
against it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-5 (amended): the CLI stays native for now, behind one read seam that is ready for the
cut-over.&lt;/strong&gt;
&lt;ul&gt;
&lt;li&gt;The status module exposes a single read port for the D-2 granularities (overview and
detail). Every CLI command that reads mission status, including
&lt;code dir=&quot;auto&quot;&gt;agent tasks status --json&lt;/code&gt; and orchestrator-api &lt;code dir=&quot;auto&quot;&gt;mission-state&lt;/code&gt; / &lt;code dir=&quot;auto&quot;&gt;list-ready&lt;/code&gt;, calls
only that port. No command reduces events, resolves read paths or composes read models
itself.&lt;/li&gt;
&lt;li&gt;Today the port has one adapter: the native, in-process Python implementation (D-3). It
never calls the service, so the CLI keeps working offline with no JVM binary present.&lt;/li&gt;
&lt;li&gt;The service ports the Lamport reduction (&lt;code dir=&quot;auto&quot;&gt;status.reducer.materialize&lt;/code&gt;). A conformance
test runs both implementations over the same missions and fails on any difference, so
there is one &lt;em&gt;answer&lt;/em&gt; while there are two implementations.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Cut-over (later, separate decision).&lt;/strong&gt; Once the Java service has proven itself, a
second adapter that calls its API replaces the native one &lt;em&gt;inside the seam&lt;/em&gt;. The CLI
commands do not change and cannot tell the difference. The native implementation is
then retired.&lt;/li&gt;
&lt;li&gt;The D-5 bypass gate enforces the seam: a CLI command that reads status without going
through the port fails it.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-6 (unchanged).&lt;/strong&gt; Both implementations are read-only and independent of drain.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-7 (amended).&lt;/strong&gt; The service is its own build in this repository (for example
&lt;code dir=&quot;auto&quot;&gt;services/mission-status-api/&lt;/code&gt;), with a path-filtered CI workflow and its own release
artifact. It is not part of the Python wheel, so “the CLI wheel ships no server, daemon,
port or PID file” still holds. The read port and its native adapter stay in the
&lt;code dir=&quot;auto&quot;&gt;specify_cli&lt;/code&gt; status domain as D-7 originally placed them.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Consequences.&lt;/strong&gt; Until the cut-over there are two reducer implementations. That is
acceptable only because the conformance test is a gate, not a report. Tightening the status
module’s encapsulation is now prerequisite work for the cut-over, tracked with the reader
re-pointing (#5532). The JVM toolchain and the native release matrix are new CI cost,
confined to the path-filtered workflow. Still open on #5528: the lifecycle-status
vocabulary, the WP status chips, and the artifact reads (#5533).&lt;/p&gt;
&lt;h2 id=&quot;amendment-2026-10-12-accepted-open-core-producer-surface-in-repo-now-extracted-when-frozen&quot;&gt;Amendment 2026-10-12 (Accepted): open-core producer surface; in-repo now, extracted when frozen&lt;/h2&gt;
&lt;p&gt;&lt;strong&gt;Deciders:&lt;/strong&gt; Stijn Dejongh (operator), 2026-10-12. This amendment ratifies the record:
with it the ADR moves from Proposed to &lt;strong&gt;Accepted&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Why.&lt;/strong&gt; The read API’s place in the product needed naming. It is the stable entry point to
the mission ledger and event data that the free CLI already exposes — the &lt;strong&gt;producer&lt;/strong&gt; side
of the open-core line (internally, “Build Kitty”). It is not a hosted Team Kitty surface, and
it is not gated on hosted drain (D-6). Framing it as open-core producer settles two things
that D-3..D-7 left to “modulith first, extractable later”: &lt;em&gt;who owns it&lt;/em&gt; (the open-core CLI,
not the hosted product) and &lt;em&gt;where it may go&lt;/em&gt; (its own open-core surface, never a Team Kitty
repository).&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;D-9 (Accepted): the Mission Status Read API is the open-core producer surface.&lt;/strong&gt; It
reads the committed ledger and events and returns the two D-2 granularities to any
consumer. Team Kitty — the “Multiplayer Mode” UI/multiplayer surfaces (mission-ui, the pack
composer) — is a &lt;strong&gt;consumer&lt;/strong&gt; of this API across its contract
boundary, not part of it. The boundary is the contract, and the contract is public.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-10 (Accepted): in-repo now, extracted when frozen.&lt;/strong&gt; The Java service of the amended
D-4/D-7 (&lt;code dir=&quot;auto&quot;&gt;services/mission-status-api/&lt;/code&gt;, epic #5528; #5529 the overview resource, #957 the
detail resource, #5558 the contract) &lt;strong&gt;starts in-repo as a modular component&lt;/strong&gt; — its own
build, path-filtered CI and release artifact, outside the Python wheel. It is &lt;strong&gt;extracted
as needed, as an open-core surface&lt;/strong&gt; (its own repository or package), once its contract has
proven stable — &lt;strong&gt;never into a Team Kitty / hosted repository&lt;/strong&gt;. The native Python adapter
(D-3/D-5) stays the CLI’s offline read path until the amended-D-5 cut-over. &lt;em&gt;(Target state
when written: &lt;code dir=&quot;auto&quot;&gt;services/mission-status-api/&lt;/code&gt; did not exist on disk yet — this decision fixes
where it starts. Amended 2026-10-13: slice 1 of #5529 lands the service scaffold under
&lt;code dir=&quot;auto&quot;&gt;services/mission-status-api/&lt;/code&gt; with its own Gradle build and the path-filtered
&lt;code dir=&quot;auto&quot;&gt;mission-status-api.yml&lt;/code&gt; workflow. The tree is excluded from the wheel and the sdist, so
only &lt;code dir=&quot;auto&quot;&gt;contracts/mission-status/&lt;/code&gt; and the conformance fixtures are part of this repository’s
shipped surface; the decision itself is unchanged.)&lt;/em&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-11 (Accepted): the contract and fixtures stay canonical beside the Python reducer.&lt;/strong&gt;
&lt;code dir=&quot;auto&quot;&gt;contracts/mission-status/&lt;/code&gt; (the OpenAPI document, #5558) and the conformance fixtures
remain the canonical source of record &lt;strong&gt;in this repository, beside the Python reducer they
conform the Java service against&lt;/strong&gt; (the amended-D-5 conformance gate). An extraction moves
the service build, not the contract’s authority away from the reducer it is proven against.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;D-12 (Accepted): this surface is on the GA gate.&lt;/strong&gt; As the producer side of the open-core
line, the mission-status API backend is part of the 4.0.0 GA gate (with charter modular
extraction), alongside local integrity — see the
&lt;a href=&quot;/plans/4-0-0-milestone-roadmap/&quot;&gt;4.0.0 milestone roadmap&lt;/a&gt;. This does not change any
code, reducer or gate; it records the direction and the ownership.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Consequences.&lt;/strong&gt; The read API has a settled identity: open-core producer, in-repo today,
extractable as its own open-core surface when frozen, with the contract anchored to the
Python reducer. Consumer-side UI/multiplayer work (Team Kitty / Multiplayer Mode) may build
against the contract now (see &lt;a href=&quot;/adr/4.x/2026-10-06-1-team-kitty-surfaces-are-hidden-unless-drain-is-on/&quot;&gt;ADR 2026-10-06-1&lt;/a&gt;,
Amendment 2026-10-12). Nothing hosted — zeitgeist auth, the hosted backend, the SaaS surface
— is revived by this amendment. The produce/drain decoupling (D-6, roadmap next step) remains
its own decision.&lt;/p&gt;
&lt;h2 id=&quot;pros-and-cons-of-the-options&quot;&gt;Pros and Cons of the Options&lt;/h2&gt;
&lt;h3 id=&quot;option-1-keep-and-harden&quot;&gt;Option 1: keep and harden&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; no user-visible gap. &lt;strong&gt;Cons:&lt;/strong&gt; keeps a daemon and a UI in the CLI core, against
the adapter-seam rule. Needs security, cache and transport work (#4791, #956, #460) on a
surface the operator wants out. Does not fix the duplicate read paths.&lt;/p&gt;
&lt;h3 id=&quot;option-2-files-and-events-tail-only&quot;&gt;Option 2: files and &lt;code dir=&quot;auto&quot;&gt;events tail&lt;/code&gt; only&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; least code. &lt;strong&gt;Cons:&lt;/strong&gt; every consumer re-implements reduction, the read-path
resolution for the coordination topology, and the authored/resolved WP split. That is the
divergence #645 exists to end. It is a shadow path by construction.&lt;/p&gt;
&lt;h3 id=&quot;option-3-in-process-read-api-plus-cli---json-chosen&quot;&gt;Option 3: in-process read API plus CLI &lt;code dir=&quot;auto&quot;&gt;--json&lt;/code&gt; (chosen)&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; a single authority, no daemon, extractable later, and both granularities served
from existing read models. &lt;strong&gt;Cons:&lt;/strong&gt; consumers that are not Python pay a subprocess per
call. That cost is acceptable because change detection uses the cursor, not a rescan.&lt;/p&gt;
&lt;h3 id=&quot;option-4-http-service-in-the-cli&quot;&gt;Option 4: HTTP service in the CLI&lt;/h3&gt;
&lt;p&gt;&lt;strong&gt;Pros:&lt;/strong&gt; browser-friendly. &lt;strong&gt;Cons:&lt;/strong&gt; reintroduces the daemon lifecycle and the trust model
of #4767–#4769 in a new form. It also decides the transport for consumers who should own
it.&lt;/p&gt;
&lt;h2 id=&quot;open-questions-operator&quot;&gt;Open questions (operator)&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Name.&lt;/strong&gt; Resolved 2026-10-01: the operator ratified &lt;strong&gt;Mission Status Read API&lt;/strong&gt;.
“Runtime status” was rejected because it already names three things: &lt;code dir=&quot;auto&quot;&gt;src/runtime/&lt;/code&gt;
(the control loop), &lt;code dir=&quot;auto&quot;&gt;src/specify_cli/runtime/&lt;/code&gt; (agent assets) and WP runtime state
(&lt;code dir=&quot;auto&quot;&gt;DIRECTIVE_032&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Interim risk.&lt;/strong&gt; Resolved 2026-10-01: the dashboard is deleted first (amended D-8),
which closes #4767, #4768 and #4769.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Charter.&lt;/strong&gt; “Dashboard must support 100+ work packages without lag” (charter,
Performance and Scale) refers to the dashboard. It needs a charter amendment that
restates it as a read-API budget.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;more-information&quot;&gt;More Information&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href=&quot;/plans/domains/api-dashboard-domain-plan/&quot;&gt;API &amp;#x26; Dashboard domain plan&lt;/a&gt;: the
durable plan for #645 and #650, updated to point here.&lt;/li&gt;
&lt;li&gt;&lt;a href=&quot;/plans/4-0-0-milestone-roadmap/&quot;&gt;4.0.0 milestone roadmap&lt;/a&gt;: Strangler prep lane.&lt;/li&gt;
&lt;li&gt;&lt;a href=&quot;/architecture/status-model/&quot;&gt;Status model&lt;/a&gt;: the event log and reducers.&lt;/li&gt;
&lt;li&gt;&lt;a href=&quot;/adr/3.x/2026-09-26-3-hosted-interaction-opt-in/&quot;&gt;ADR 2026-09-26-3&lt;/a&gt;: drain posture. The
produce/drain amendment is a separate 4.x ADR.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;amendment-2026-10-29-accepted-the-events-services-live-in-spec-kitty-events&quot;&gt;Amendment 2026-10-29 (Accepted): the events services live in spec-kitty-events&lt;/h2&gt;
&lt;p&gt;&lt;a href=&quot;/adr/4.x/2026-10-29-1-the-events-domain-and-its-java-services-live-in-spec-kitty-events/&quot;&gt;ADR 2026-10-29-1&lt;/a&gt; records the operator and CEO ruling of 2026-10-29 that the events domain’s Java services and contracts move to &lt;code dir=&quot;auto&quot;&gt;spec-kitty-events&lt;/code&gt;, and that core consumes a released, pinned build.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;D-7 and D-9 to D-12 (the service is its own build in this repository, extracted when its contract freezes).&lt;/strong&gt; The service, its &lt;code dir=&quot;auto&quot;&gt;domain-model&lt;/code&gt; and &lt;code dir=&quot;auto&quot;&gt;event-log&lt;/code&gt; libraries and the &lt;code dir=&quot;auto&quot;&gt;mission-status&lt;/code&gt; contract move to &lt;code dir=&quot;auto&quot;&gt;spec-kitty-events&lt;/code&gt; with their history, and are released there on the &lt;code dir=&quot;auto&quot;&gt;service-v*&lt;/code&gt; line. Core pins the read service for CI conformance only. The CLI never launches it, so D-5 (CLI reads stay native Python) and D-7’s “the CLI wheel ships no server” hold unchanged. The contract and conformance fixtures are canonical in &lt;code dir=&quot;auto&quot;&gt;spec-kitty-events&lt;/code&gt;; core’s Python oracle tests run against the pinned contract bundle and fixtures.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Nothing else in this record changes.&lt;/p&gt;
&lt;h2 id=&quot;amendment-2026-10-30-accepted-the-ordering-is-wall-clock-not-lamport&quot;&gt;Amendment 2026-10-30 (Accepted): the ordering is wall clock, not Lamport&lt;/h2&gt;
&lt;p&gt;Factual amendment, tracking &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/6604&quot;&gt;#6604&lt;/a&gt;.
The decision does not change: D-3 to D-7, the D-5 read seam and the conformance gate all
stand exactly as accepted. What changes is the wording this record used for the ordering it
relies on, which no longer matches the code.&lt;/p&gt;
&lt;p&gt;The operator ruling of 2026-10-22 (recorded in &lt;a href=&quot;/adr/2.x/2026-02-09-3-event-log-merge-semantics/&quot;&gt;ADR 2026-02-09-3&lt;/a&gt;
Amendment 2026-10-22, and in the same day’s amendment to
&lt;a href=&quot;/adr/4.x/2026-10-14-1-event-native-core-one-canonical-event-writer/&quot;&gt;ADR 2026-10-14-1&lt;/a&gt;) settles
what the status path does for 4.0.0:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Wall clock.&lt;/strong&gt; Lane transitions are ordered by the writer’s &lt;code dir=&quot;auto&quot;&gt;at&lt;/code&gt;, then &lt;code dir=&quot;auto&quot;&gt;event_id&lt;/code&gt;. One
reducer applies that fold: &lt;code dir=&quot;auto&quot;&gt;status.reducer.reduce&lt;/code&gt; calls &lt;code dir=&quot;auto&quot;&gt;spec_kitty_events.diary.reduce_parsed&lt;/code&gt;
directly, and &lt;code dir=&quot;auto&quot;&gt;materialize_snapshot&lt;/code&gt; calls &lt;code dir=&quot;auto&quot;&gt;reduce_shared_state&lt;/code&gt;, a re-export of
&lt;code dir=&quot;auto&quot;&gt;diary.reduce&lt;/code&gt; (&lt;code dir=&quot;auto&quot;&gt;parse_diary&lt;/code&gt;, then &lt;code dir=&quot;auto&quot;&gt;reduce_parsed&lt;/code&gt;). There is one reducer and two
parsers, not a Lamport reducer beside a wall-clock one.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;From-lane continuity.&lt;/strong&gt; A transition applies only when its &lt;code dir=&quot;auto&quot;&gt;from_lane&lt;/code&gt; matches the WP’s
lane (spec-kitty-events#87; spec-kitty &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/4990&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;#4990&lt;/code&gt;&lt;/a&gt;,
closed 2026-09-25 in &lt;code dir=&quot;auto&quot;&gt;spec-kitty-events&lt;/code&gt; 10.4.0), so a stale approval stamped later cannot
override a committed rejection.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The approved-departure lift.&lt;/strong&gt; A non-seed departure from &lt;code dir=&quot;auto&quot;&gt;approved&lt;/code&gt; sorts after the
approval it followed, so a rejection minted on a clock behind its approval still applies;
the stored &lt;code dir=&quot;auto&quot;&gt;at&lt;/code&gt; is untouched (spec-kitty &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/4941&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;#4941&lt;/code&gt;&lt;/a&gt;,
fixed by &lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/pull/6261&quot;&gt;PR #6261&lt;/a&gt; in
&lt;code dir=&quot;auto&quot;&gt;spec-kitty-events&lt;/code&gt; 10.5.1).&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The residual, still open.&lt;/strong&gt; A &lt;strong&gt;non-approval&lt;/strong&gt; transition minted on a clock behind the
event it follows still sorts before that event. The lift covers departures from &lt;code dir=&quot;auto&quot;&gt;approved&lt;/code&gt;
only.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Lamport-primary ordering is a post-GA direction&lt;/strong&gt;
(&lt;a href=&quot;https://github.com/spec-kitty/spec-kitty/issues/6609&quot;&gt;&lt;code dir=&quot;auto&quot;&gt;#6609&lt;/code&gt;&lt;/a&gt;). It needs a clock on CLI
rows (spec-kitty-events#72) and its own ADR before it is built. This record does not adopt
it.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Which sentences above are superseded in wording (their decisions stand unchanged):&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The Context section’s “two different reducers” paragraph (the one describing
&lt;code dir=&quot;auto&quot;&gt;materialize_snapshot&lt;/code&gt; as “the Lamport &lt;code dir=&quot;auto&quot;&gt;reduce_shared_state&lt;/code&gt; reduction”) and D-5’s “moves
them onto the Lamport reducer” describe the superseded ordering. The read API’s point
still holds — external readers should reach one answer — but that answer is the wall-clock
fold, and the D-5 bypass gate is about the read seam, not about which clock the reducer
reads.&lt;/li&gt;
&lt;li&gt;D-5 (amended) “the service ports the Lamport reduction (&lt;code dir=&quot;auto&quot;&gt;status.reducer.materialize&lt;/code&gt;)” is
superseded: the service ports the same &lt;code dir=&quot;auto&quot;&gt;(at, event_id)&lt;/code&gt; fold, and its conformance test
compares that port against the Python fold. Gap G-4 in the
&lt;a href=&quot;/architecture/event-ledger-writers-and-readers/&quot;&gt;event-ledger writers and readers map&lt;/a&gt;
tracks the one part of that fold the Java port does not implement yet.&lt;/li&gt;
&lt;li&gt;“Until the cut-over there are two reducer implementations” now means two &lt;em&gt;implementations
of one algorithm&lt;/em&gt; (the Python adapter and the Java port), not two different algorithms.
The conformance gate is unchanged and remains the reason the pair is acceptable.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Nothing else in this record changes.&lt;/p&gt;</content>
</entry>
