The Models page drops the generic schema renderer and the visible
environment-variable field: each editor is a curated per-family card whose
primary input is one write-only API key stored under a derived
<ROUTE>_API_KEY reference (recorded as apiKeyEnv in the pi-ai profile), an
unkeyed whole-section provider opens as its setup card, and the collapsed
customized-settings fold carries baseURL/reasoningEffort (deepseek) or
reasoning (pi-ai). dsh-client-schema-form reduces to the schema/draft model
layer (no React).
apps/cli/cordis.yml gains settings-local, credentials-local, and the bare
dormant llm-pi-ai row (manifest deps added for the resolver contract);
llm-deepseek drops its !!js apiKey inline for per-request credential
resolution. Both adapters tag apiKeyEnv role('credential-ref') so the
form mounts the credential control. The web e2e scaffold isolates a
harness home per run — an in-process boot must never touch the
developer's real ~/.dsh — and the new models-settings scenario pins the
whole loop through the shipped app: dormant directory as add vocabulary,
schema-driven editor apply landing in settings.yaml, the route
registering live (topology frame), and a write-only key landing in the
temp .env with the configured badge converging. A hermetic test-owned
reference name keeps a developer's real provider keys from flipping the
badge. schema-form joins the platform module table (seed + externals)
so client bundles share one instance.
The Models settings section joins llm.providers (the configurable
directory with live state), settings.describe (schemas, layered redacted
values, secret slots), and credentials.describe (value-free badges) into
provider rows with one editor card at a time. The editor renders the
provider's profile subtree through dsh-client-schema-form; the
credential-ref role mounts a control that shows configured/source state
and stores keys write-only through credentials.set. Apply without
removals merges a minimal patch (stored secrets outside it survive);
apply after a reset — and row deletion — replace the user section so
removals land. The client runtime bridges the three new host frames to
typed ctx events (settings/credentials/models changed), the page
refetches on any of them once loaded, and ui-model's per-session picker
directories reload on models/changed so a settings-born route appears in
open pickers without a reopen.
@deepseek-ai/dsh-client-schema-form rehydrates the wire's serialized
schemastery envelope (new Schema(json)) and edits a draft user section
against it: presence-in-draft marks a field overridden with a per-field
reset, inherited values render as placeholders, role('secret') slots are
write-only with configured-state placeholders from the wire's secrets
list, dict adds take a union-typed sKey as their vocabulary, and any
node the renderer cannot faithfully edit falls back to a read-only view
instead of silently disappearing. renderField(context) is the role hook
the Models page will use for the credential-ref control; validateDraft
runs the same rehydrated validator the host uses, so the browser and
host judge one schema.
Eight compiler-locked methods: settings.describe/update/replace serve
redacted layered namespace views (secrets structurally absent from every
layer, write-only in the update direction) and fold seam refusals into
settings-rejected; credentials.describe/set/unset expose value-free views
with credential-rejected on shadowed writes; llm.providers merges the
configurable directory with live routes and llm.models claims the
host-scoped catalog reservation through the buildModelCatalog extraction
session.models now shares. Three HostFrame invalidations bridge the seam
events (host/settings-changed, host/credentials-changed,
host/models-changed), and the connection route generalizes the native-
dialog check into a privileged-method set covering all four writes. The
fixture and both fake clients grow the same face.
New credentials data-structure page (type-equiv manifested), group README,
rewritten llm-deepseek/llm-pi-ai READMEs (dynamic configuration, dict
profiles, credential chain), capability-seams/service-role registration,
Agent Note (bilingual), demo compositions mounting settings-local +
credentials-local with no inline key plumbing, installSettingsSection
consumer helper on the settings seam (deduplicating both adapters' wiring),
jscpd symmetry markers for the provider twins, runtime-closure additions for
python/sdk-runtime, and doc-budget ceilings AGENTS.md 1750→1755 /
packages/README.md 850→865 for the structural one-line group rows.
llm-deepseek gains a Loader+Include composition spec proving external
settings.yaml/.env edits reach the very next request, and a real-API e2e
where only a credentials-local document holds the key. The headless example
pins the first-run missing-credential UX as a keyless stream-json snapshot
(new credentials.cordis.snapshot.yml scenario); runLoaderSmoke learns
expectedExitCode so a designed failure surface can be pinned instead of
masked.
providers becomes a dict keyed by provider route, so the composition base
and the llm-pi-ai settings section merge per provider and the route set is
structural; the pre-release array shape and per-profile provider field fail
loud with migration directions. The adapter reads a profiles thunk once per
operation and resolves the credential per stream call (literal apiKey, then
apiKeyEnv through ctx.credentials with an ambient env fallback, then pi-ai's
provider-native discovery), so key, endpoint, and knob changes reach the
next request without restarts. Route-set or captured-retry-policy changes
re-register the same adapter instance in one synchronous section; an invalid
settings snapshot keeps the last good profiles.
The adapter takes an options thunk and a per-stream credential resolver
instead of frozen construction facts: base URL, catalog, defaults, idle
budget, and the API key re-resolve at each operation, so a settings or
credential change reaches the very next request while in-flight streams
keep the facts they started with. resolveAdapterOptions is the one explicit
resolve step (entry config fails loud at load; a live snapshot failing a
beyond-schema bound keeps the last good options). The plugin layers its
entry config under the optional llm-deepseek settings section and resolves
keys literal-first through ctx.credentials with an ambient env fallback; a
missing key now registers the route, warns, and fails each request with
actionable MISSING_CREDENTIAL instead of failing plugin load. The
registration-captured retry policy re-registers the route in place when it
changes.
Live environment wins read-only (shadowed writes reject instead of
appearing to succeed); the file is the writable source with byte-preserving
line edits, a quoting ladder dotenv reads back verbatim, atomic 0600
writes, wholesale snapshot replacement on reload, and write-drain
teardown.
writeFileAtomic: exclusive-create random-suffix temp + rename carrying the
caller-stated mode; settings-local persistSection now consumes it. The
credentials-local store shares it next.
Conflicts were the parallel-edit set: packages/README group table (kept
master's session-title wording, re-added the settings row condensed),
doc-budgets AGENTS.md ceiling (took master's 1750), and regenerated
graph/catalog docs plus re-recorded translation pairs. packages/README.md
ceiling 845 -> 850: the table legitimately gained one row for the new
settings group and the row is already minimal.
Two-package capability family mirroring session-persistence/:
- dsh-settings: abstract Settings service — namespace registry with
caller-fiber effect registrations, layered resolution (schema defaults
< composition base < user document), schemastery validation,
per-namespace deep-equal commit detection, and the settings/updated
event. Boot/registration validation fails loud; provider publishes
keep last-good per namespace.
- dsh-settings-local: settings.yaml/.json provider — resolveSpec
defaulting to $DSH_HOME/settings.yaml, chokidar hot reload,
content-equality self-write suppression, atomic 0600 tmp+rename
writes, comment-preserving YAML namespace patching.
Consumers register inside ctx.inject(['settings'], …), so every
composition works unchanged without a mounted provider. Real Loader +
Include composition test proves cordis.yml boot and external-edit hot
propagation; HMR disposal test proves registry cleanup. Both packages
hold per-file 100% coverage.
Doc budgets rise 1705→1710 (AGENTS.md) and 835→845 (packages/README.md):
one structural line per file for the new package group.
Agent Note: .agents/notes/implemented/architecture/2026-07-28-user-settings-seam.md
Union resolutions throughout — the fixture serves both the goal and plan
projection units (catalog gains /goal beside /plan; the retired
goal-fixture sample command yields to the real goal mirror), the mux
baseline spec expects all four unit frames, and the tsconfig paths /
Model Experience allowlist carry both domains' outlets.
A jsdom slot test runtime for feature specs: a real Cordis Context, the
production SlotsService and web-react renderer, and typed session/workspace
doubles (TestSessions implements ISessions with FixtureSession sessions;
TestWorkspaces implements IWorkspaces), so the compiler flags fixture drift
when a production face changes. Fixtures feed plain data: list rows,
conversation snapshots, and ISession-typed behavior stubs; provide-bundle
materialization runs the shared SessionProvideChannel.
DOM snapshot support: declare()/renderSlot() mount a single slot inside a
data-slot wrapper for local .snap capture, and a snapshot serializer folds
CSS-module class hashes to their semantic locals and collapses svg internals
to a content fingerprint. The typed provide() constrains declared-service
fakes to Partial of the service's outward face.
Plan mode is entered through /plan only — the select control is retired.
The conversation.input.plan seat (now right of the access-mode control)
renders a read-only Plan chip while the projection's effective target is
plan mode; its hover x executes /plan off, and the chip follows the folded
target (appears on /plan immediately, disappears on /plan off) with frames
correcting either way. While plan mode is targeted the composer textarea's
placeholder switches to the plan-task wording — InputBar reads the same
projection through the standard-kit useProjection (the TodoDock posture:
a type-only key merge, no domain service edge), and owner placeholders
still win.
Static: the cache package.json files array matches the workspace
constraint shape, the unused dsh-storage-json devDependency is dropped
(tests run on the memory backend), and docs/module-graph.md is
regenerated for the new package edge.
Coverage: two unreachable branches deleted rather than tested —
coldSnapshot's floor-0 tail reuse (a baseSeq-0 restore never throws and
an unrelated record still carries a usable watermark) and flushSoft's
non-mandatory clean-skip (throttle triggers only fire dirty). New tests
close the real gaps: write() on a never-dirty session and the non-JSON
unit-state rejection, plugin disposal clearing armed interval timers,
cachedSnapshot's all-version-mismatched and cwd-identity arms, the
zero-units empty-log cut, the coordinator seek-hook ladder (suffix /
not-found / plain failure / abort-reason relay), and the superseded-
retirement race proving forget()'s exact-entry guard.
SessionSummary grows an optional projections column (whole value per key,
same passthrough posture as the history-tail block): attached rows cut the
live registry watermark cache; cold rows view the persisted projection
cache's stored rows via the new registry viewCheckpoint face (version-
matching keys only, zero I/O) — the RFC's motivating scenario, every
session's title across a listing without loading one event log. The column
is fail-soft and absence-coded: no registry, no cache row, or a throwing
read serve the row without the column, never breaking the listing.
New package on the domain data form: one session_projcache record per
session (key → {stateVersion, observedSeq, state}), landing beside
workspace.json under the shipped json backend. Write policy: two mandatory
points (turn/end + session disposal) with count/interval throttling between
them (both Config fields required — flush cadence is a deployment choice);
every background write is fail-soft (log + stay stale, self-heal on the
next write or cold read). coldSnapshot(id) runs the read ladder — cached
rows + persistence readFrom from the registry's anchored restore floor +
registry restore + fail-soft write-back — detecting crash-repair-shrunk
logs via the one-below anchor and degrading to a single full re-read.
Mounted in apps/cli/cordis.yml (writeEveryEvents 200 / writeIntervalMs
5000).
goal (GoalService + the 'goal' projection unit), goal-session (automatic
continuation), command-goal (/goal — lands in the materialized effective
catalog and the slash menu with zero client changes), and ui-goal (the
GoalBar dock surface). apps/cli deps updated accordingly.
New client plugin package. The GoalBar component migrates verbatim from the
original PR (phase labels, inline edit form, resume-when-paused, pending
lockout, inline errors); its data path is rebuilt for the projection model:
- GoalDock adapter reads useProjection('goal') from the standard kit — no
controller, no store, no refresh chain, no event listener (the original
PR's Session goal fields, write-revision fence, and coalesced refetch all
dissolve into the value store's seq rule).
- The inject face carries only the three mutation verbs; each reads the CAS
ref from the session's current projected value at call time and maps the
settled RPC result inline (callbacks from inject, live state from
useProjection).
- Seats at conversation.input.dock order 1 (the slot contract names GoalBar
as a resident of this strip), below TodoPanel/QueueDock, tucked against
the composer; hero phase renders nothing for free.
- IconSparkle16 is package-local (sole consumer) — ui-primitives keeps only
the shared action icons.
- goalbar.spec migrates with GoalSnapshot props (activation is process-local
and absent from the projection; the strip renders durable phase only).
dsh-goal/types becomes the client-safe pure outlet (tool-todo dual-outlet
shape): GoalId/GoalRef/GoalPhase/GoalBlockReason/GoalSnapshot, the new
GoalProjection payload, and the SessionProjectionMap 'goal' key merge — zero
host imports. Host-coupled vocabulary (GoalView/activation, change metas,
message source, folds, GoalError codes, the scoped goal/changed event) moves
to src/domain.ts, re-exported from the package root. ./client re-exports the
outlet for client aggregates.
GoalService registers the projection unit under ctx.inject(['sessionProjections']):
applyGoalProjection is a projection-grade last-wins fold — plain-JSON state,
same-reference return on non-goal or malformed events (a throwing apply
would tear down the registry drive; strict validation stays with the write
side and foldGoal). Activation is process-local and deliberately absent from
the projection value.
Rewrite ui-plan as a pure browser surface plugin. The control occupies the
composer's named plan seat (declared empty by ui-conversation); reads render
the host-computed plan projection through the standard-kit useProjection
(absent key = capability absence, hides the control), writes execute /plan
or /plan off through command.execute. The node half becomes the empty
roster apply: plan behavior (command, policy, projection unit) is owned by
dsh-plan-mode, already composed on the web roster with its policy in
cordis.yml. The superseded RPC-backed setPlanMode face, the WEB_PLAN_SECTION
duplicate, and the node-plugin spec are removed; the roster row moves from
the retired CLIENT_PACKAGES table to the cordis.yml dshClient roster.
plan-mode contributes the RFC's complete double-event example: command/run
records named plan set the wanted target (off -> false, else true),
plan/mode commits and clears it, and view derives {active, pending}.
Pending is thereby a pure replay quantity recovered from the log alone.
The plan key merges into SessionProjectionMap from src/types.ts, served
through the new ./types and ./client outlets (session-title template);
compositions without the registry are unaffected.
The master merge left the generated catalogs/module graph stale, the
python runtime closure missing dsh-session-projection (now reached
through session-title and tool-todo), and apiproxy holding a dead
session-title dependency (the bespoke title frame is retired). The four
files the merge pushed under the per-file coverage floor (commands
executor + invariant, projection registry drive tails, TUI) join the
existing TODO(gui) deferral block per the GUI-lane policy; the remaining
coverage-run failures reproduce identically on pure origin/master
(environment-bound suites: sdk process exit, TUI PTY timing, workflow
worker timing, title loader slow-boot) and are not this branch's debt.
commandId crosses three boundaries (session log, wire admission response,
client flow pairing), so per the branded-id rule it becomes
Branded<'CommandId'>, declared in a new pure @deepseek-ai/dsh-commands/brand
outlet (the dsh-llm/brand shape: type + constructor, no Context merges, so
wire and client programs can name it without loading the host plugin). The
event payloads, CommandExecution, and the executor mint carry the brand; the
wire schema gains commandIdSchema as the domain's single brand-cast point
(the approvals precedent); CommandNode and the fixture's fabrication cast
follow type-only.