Review asked why this is a plugin and why it exports so much. The design note
this branch shipped answered the second question itself — "a pure resolution
function plus an installation function" — and the code drifted to seventeen
exports and a Cordis plugin nobody approved or mounted.
The plugin is gone. Transport policy has one answer per process: nothing to
swap, and no scope narrower than the process to give one. Its `Config` was also
the only supplier of a configuration branch, so resolution now reads the
environment and nothing else — `mode`, the config-sourced fields, and the
`config` policy source were unreachable the moment the plugin left.
Four exports nothing outside the package used are internal again, and
`currentProxyPolicy` answers with the direct policy instead of `undefined`, so
`DIRECT_POLICY` no longer needs a public face. Nine functions remain, one per
way a caller can need the policy; two is not reachable with six consumer seams.
The package moves to `util/`. The note claimed a dependency on `undici`
disqualified it from that group; the charter governs harness dependencies, not
external ones, and the process note that says so predates this branch. The real
blocker was the harness dependency: resolution needed one method of
`LaunchEnvironmentSnapshot`, so it names a structural `EnvLookup` and the
launcher passes its snapshot unchanged. `net/` is dissolved.
Dropping the group's line from the repository layout also returns `AGENTS.md`
to its original ceiling, so the raise the merge needed is reverted.
Review follow-up: the per-step gate recomputed view(previous) on every
changed apply — a property read for identity-stable views, but a fresh
throwaway object per change for computing views. The registry now keeps a
WeakMap from state object to raw view: the previous state's view was
cached when that state was current, so each distinct state's view computes
exactly once (gate and snapshot share the memo) and the quiet path
allocates nothing. Unlike the earlier lastView record, an entry is keyed
by the state itself — the view of that exact state by the pure-view
contract — so no stamping discipline exists to get wrong. Primitive
states bypass the WeakMap and compute directly.
Review consensus: for the only declaring unit the token function was
literally the view function, so the extra wire member bought nothing —
default raw-view comparison produces the identical Object.is on
state.turns at the identical cost (one property read per side). The
per-step gate stays: it holds the no-stored-baseline property, and its
measured overhead (~ns per changed state) is four orders of magnitude
below the push path it guards.
Review follow-up (imccyu): comparing view(previous) recomputes the view on
every quiet change. The wire block now takes an optional viewKey(state)
declaring the cheap comparison token; the drive compares tokens across the
previous and next states and calls view only for an actual push. Default
stays the raw view output. turnOutline declares viewKey: state => state.turns,
making the identity-stable-turns convention an explicit contract; the
registration erasure forwards viewKey (dropping it silently reverted the
gate to the fallback, caught by the new view-call-counting test).
Review suggestion (imccyu): the drive holds both the previous and next
state, so the identity gate can compute view(previous) and view(next) in
the driving step and compare them directly. The stored lastView cell field
and its stamp-on-every-change rule are deleted; with no dedup memory,
nothing can go stale across listener generations by construction, and a
rebuilt cell no longer pushes an unchanged view on its first live event.
Costs one extra pure view() call per changed state.
The dsh 0.1.2-alpha.2 release bumped every workspace after this branch
forked, so the new package failed the workspace version constraint in the
merge tree. The bounded preview() also grew identical enough to its
deliberate host/client mirror to trip clone detection; the client copy now
carries a jscpd ignore region naming the wire-boundary rationale.
preview() now slices a single text block to limit * 2 before joining and
normalizing, so one multi-megabyte block no longer pays a full-string pass;
the host projection and the client turn-navigation helper stay mirrored.
outlineEntry keeps dropping entries with damaged turn/seq (marks cannot
exist or jump without them) but degrades malformed prompt/response
previews to empty strings so the turn stays navigable.
The change feed stamped lastView only when a listener was subscribed, so a
value change during a listener-free window (HMR swap) froze the baseline
and a later transition back to the old value was silently deduplicated.
The baseline now advances on every changed state, heard or not; broadcast
still only happens with listeners. Docs, catalog, and the feature note
follow the corrected semantics.
A jump clicked while the Load-earlier pull was in flight fell through the
settle effect's nearest-turn fallback and landed on the wrong row. The
effect now keeps the pending jump (busy pulse stays) while loadingOlder is
true and a retry tick re-issues loadThrough when the pull settles.
Also repins the long-interactions e2e rail block to the outline-rail
semantics (fixed mark pitch, load-and-jump labels) and covers the
loadThrough forwarding paths in apply-inject and the client-runtime stub.
loadThrough now assigns its low-water target only when it owns the loop
(retargeting stays inside the running-jump branch): a call refused while
a plain load-earlier pull holds the pager no longer leaves jumpTargetSeq
behind to drag a later jump all the way to the head. The loop also
carries the doOpen stale-pass guard so a mid-flight resync stops it
instead of paging the new stream generation toward the old target.
Edge-branch coverage for the preview reading bound, repeated and empty
drafts, draftless turn ends, orphan-draft clearing, and same-response
recommits (the CI per-file gate exercises them); the jump-settle spec
now asserts only the busy lifecycle after settlement — jsdom's zero
geometry made the rAF active-turn resync timing-dependent under
coverage instrumentation, and the landing position contract lives in
the browser e2e.
Outline entries gain the turn's final text-bearing assistant preview:
each assistant message overwrites a state draft and turn/end commits
the survivor, matching the loaded rail's findLast semantic; the bare-
array wire keeps its identity across draft changes, so pushes stay at
three per turn. Preview budgets shrink to the rail card's clamps — one
50-character prompt line, up to three 120-character response lines,
ellipsis on clip — on loaded and unloaded turns alike (stateVersion 2
discards v1 cache rows).
The feed previously fired on every changed state reference of a
client-visible unit; it now also compares the raw view output against
the last delivered one and stays quiet when Object.is-identical, so a
unit can buffer working fields in state behind an identity-stable
projection. Units whose views build fresh objects per call are
unaffected.
Clicking an unloaded rail mark from the pinned tail raced the pinned
scroll snap: the first prepend's compensation fires a non-reader scroll
delivery, the snap called toBottom, and toBottom cancels a pending jump
— so the jump silently stayed at the tail while history loaded. The
click now drops atBottom itself (jumping into history is leaving the
live tail), pinned by a jsdom regression and re-verified live: a
118-turn session lands on turn 1 in ~250ms from click.
READMEs record the rail's outline merge and the loadThrough paging
verb in both languages; the feature Agent Note owns the decision,
alternatives (sparse windows, minSeq wire bound, outline RPC, height
estimation), and coverage map. Regenerates the client/API catalogs the
widened faces feed.
A mid-paging landing keeps the jump armed with the target row as its
paging anchor, so later chunks and the load-earlier button's unmount
cannot drift the landing; the loader's completion runs one final
correction unless the reader already scrolled off the target. Adds the
browser contract: full outline ladder, keyboard jump on an unloaded
mark, landing geometry, and rail fades.
Marks keep a fixed 10px pitch instead of compressing into the frame:
overflow scrolls inside a hidden-scrollbar scroller with gradient fades
over each still-scrollable end, the preview compensates the rail scroll,
and the active mark keeps itself centred while the pointer is off the
rail. Pointer-to-mark mapping now works in ladder coordinates.
The rail now merges the turnOutline projection with loaded-window items
(view-layer only; loaded wins, outline fills mid-turn prompt previews),
renders unloaded turns as dimmer marks, and clicking one holds the
reader's place, pages history through the turn's seq, and lands on its
row after the commit — no height estimation. Settlement repages once
per head movement, then falls back to the nearest rendered turn.
Session.loadThrough(seq) loops the existing prepend pager (200-message
pages) until the window covers the target seq, with a shared low-water
retarget for repeated calls, a no-progress guard against empty pages
still claiming history, and loadOlder's fail-soft error posture. Busy
state rides the existing loadingOlder snapshot bit.
New turnOutline projection unit serving every started turn's number,
turn/start seq, and bounded first-prompt preview through the
session-projection seam, mounted in the web-app bundle for the chat
turn rail. Entries stay strictly increasing; previews mirror the rail's
loaded-turn preview budget.
Yichen Jiang