`dsh` shipped two config trees that were 43 rows the same: apps/cli/cordis.yml
composed web as 74 flat rows, while the TUI booted examples/tui-agent/cordis.yml
whose single `@deepseek-ai/dsh-tui-demo` row mounted twelve plugins behind a
twenty-key pass-through Config. Neither file was what its location claimed —
apps/cli hardcoded the "example" as the product default and the "demo" bundle
was the application — and every capability change had to be made twice.
- apps/cli/base.cordis.yml holds the 43 shared rows; tui.cordis.yml and
web.cordis.yml are patch lists stating only what differs per surface
- overlays apply as SIBLING patch lists at one include level, because include
patches never cross an include boundary. Precedence: base < surface <
(--config | personal ~/.dsh/config.yaml) < launcher flag/profile patches
- `--config` now applies an overlay INSTEAD OF the personal one, so a demo or
test tree never inherits the user's route; new `--config-replace` boots a file
as the entire tree (the old `--config` behaviour). Both survive /resume
- vendor/include: index each `insert`ed row as it is added so a later patch can
configure or disable it. Upstream built the id index once before the patch
loop, leaving every surface-only row — the whole TUI front door — silently
unpatchable from user config. Logged as local modification 8
- session identity moves to dsh-agent-loop's CONFIGURED_AGENT_IDENTITIES_KEY;
dsh-tui's MAIN_SESSION_ID_KEY is deleted (only the bundle read it)
- delete examples/tui-agent, examples/cordis-agent, packages/examples/tui-demo;
TUI tests → apps/cli/tests, cordis e2e → packages/cordis/tool-cordis/tests,
examples/code-mode survives as an overlay leaf
- `dsh web` gains --config, threaded into AppCLIEntry as an extra overlay
Three latent defects surfaced and are fixed here: the TUI captured the optional
sessionQuery service once at construction and could permanently disable /resume
when it won the mount race; the session-store root silently reverted to a
project-local ./.sessions; --config-replace was dropped by the resume handoff.
Verified by booting each tree through the real Loader (TUI 55 entries, web 75,
zero unsettled) rather than reading YAML. All eight terminal snapshots replay
byte-identically; 14/14 PTY smoke, 112/112 snapshots, 25/25 doc-sync, hygiene
and lint clean.
Two comment-only corrections from the final review round.
The `isCompactCheckpoint` redundancy note appealed to the mandatory-marker
invariant to explain something the enclosing branch now states outright:
after e029ffb88 both call sites are literally inside an
`isReplacementSurfaceEvent` branch, so the appeal became retained
reasoning. Say the call sites test it directly.
`appendPreCompactionLog` implied its tool-result content is what survives
compaction, but `bash`'s presenter is static, so the string never reaches
a fixture — the fixtures pin that the shadowed step's card survives.
Neutral text plus a note on where the card body comes from, so a later
reader does not "fix" a fixture to make the sentence true. Verified by the
absence of fixture drift from changing the string.
The replay loop spelled the classification as
`isSurfaceEligibleType(event.type) && !isAppendSurfaceEvent(event)` while
the live listener used `isReplacementSurfaceEvent(event)`. Under the
mandatory-marker invariant these select the same set: the only divergence
is a surface-eligible event carrying no marker, which no active Session
can hold — `Session.append` and the seed construction loop both reject it
through the same `planSurfaceEvent`.
Using the same predicate on both paths makes "both paths classify a
surface event by the same marker" structural rather than argued, and
retires the TUI's last `isSurfaceEligibleType` use.
No behavior change: the TUI suite and all 113 snapshot fixtures pass
unchanged.
Review follow-ups on b9b2e593f, all documentation precision.
The redundancy note on `isCompactCheckpoint` leaned on a reading its call
site does not state: `index.ts` reaches it for surface-eligible non-append
events, which is the same set as replacements only because the marker is
mandatory. Say that instead.
The Agent Note now owns three facts it was leaving to a future reader.
`rebuildTranscript` materializes a component per append-origin event and
runs on mount, color-scheme change, and every reasoning toggle — work
compaction used to bound for exactly the long sessions it serves, so the
cost now tracks session length rather than the surface. `Consequences`
names `surface-replayed-compaction` as the durable evidence for the
live/replay equivalence claim, so the two fixtures that must move together
are findable from the Note rather than the PR thread. `Deferred` records
that a page can now carry a checkpoint whose `surfaceOp.start` fell out of
the window: pagination no longer cuts on the checkpoint's provenance
group, `FoldAdapter` pads with a non-surface sentinel, and `nodes()`
degrades to `degradedSeqs()` — which is already close to the transcript
projection A2 should build deliberately.
Also corrects the definite-assignment comment in the snapshot scenario:
the assertion rests on the awaited setup invoking `beforeMount`, not on
that call being synchronous.
Review follow-ups on the append-origin transcript projection.
The live/replay equivalence claim was stated unconditionally but does not
cover `tool/call`: only replay re-derives call pairing, because a call
event carries no `surfaceOp` of its own and inherits transcript
membership from the `assistant/message` that advertised it — which the
live listener has necessarily just rendered. Narrow the claim in the TUI
README and Agent Note, and record at `rebuildTranscript` why the filter
is replay-only rather than a missing live branch.
Add `surface-replayed-compaction`: the three existing fixtures all come
from the live path, leaving the resume case the bug report leads with
pinned only by a unit test. The new checkpoint mounts with the
replacement already stored and records byte-identical to
`surface-after-compaction-wide`, so the two fixtures now pin the
equivalence they assert. The shared fixture appends move into
`appendPreCompactionLog` / `appendCompactionCheckpoint`.
`MESSAGE_TYPES` is not "human message event types" — it includes
`assistant/message`. Say what the code distinguishes (append-origin
conversation messages vs. model-only replacement copies) at the const,
the `paginate` and `session.history` JSDoc, the apiproxy README, and the
Agent Note.
Also: spell the replace shape as `Extract<SurfaceOp, { op: 'replace' }>`
for symmetry with the module's two other uses; document why
`isCompactCheckpoint` keeps a replacement check that is redundant at both
call sites; say that Ctrl+R toggles reasoning, which rebuilds the
transcript; and qualify "the sole source of derived history" as derived
*model* history now that the transcript is the other projection.
The terminal and history pagination both treated the model-visible surface as
the human transcript. A landed compaction replacement therefore erased the
conversation it summarized — messages the reader had already seen — and a
model-only replacement copy consumed a page's `maxMessages` quota, which could
also split a compaction's provenance from the replacement citing it.
`dsh-session` now exports the marker split `isAppendSurfaceEvent` /
`isReplacementSurfaceEvent`. The terminal replays append-origin surface events,
keeps a shadowed step's tool cards paired through its append-origin assistant
message, and renders one dim marker where a compaction landed; the checkpoint is
recognized through the compaction seam's `isCompactCheckpointSource` contract,
not the shape of the replacement. `session.history` counts only append-origin
human messages. Everything model-facing keeps reading `session.surface`.
CommandService.execute now returns a CommandExecution — the normalized
result plus the commandId minted for its command/run/command/done records —
and the wire admission value carries commandId exactly when matched, so the
issuing client can correlate its RPC acknowledgment with the flow node the
lifecycle events produce. apiproxy api/schema/handler, the connection
fixture, and the TUI/plan/goal consumers follow the new shape.
Extract model-command, questions, and resume sub-machines from the
~1600-line createTuiChat closure into src/chat/ factories that take
explicit dependency bundles (shared ChatChannelDeps/ChannelNotice).
Reorganize src/ so chat/ holds all chat-channel concerns (former input
and session/ files move under it); xml-tool-output moves to components/;
TuiRuntime/TuiResumeHost move to runtime.ts. index.ts drops 2067->~1530
lines. Behavior identical: 167 tests and all TUI snapshots pass unchanged.
The running status glyph clipped its cosine throb to a blank column below
STATUS_FADE_MIN_OPACITY, so each breath read as bold->dim->disappear rather
than bold->dim->bold. Remove the cutoff and the constant; the trough now
renders as the dimmest gray, giving a symmetric breathe. Turn-boundary
appear/disappear is unchanged (fade envelope and non-truecolor visible
threshold).
Consolidates the personal dsh-tui customizations (module split into
components/session/extension, prompt template + running-glyph indicator,
copyable transcript, tool-card headers, timing placement, XML tool output,
status/footer rework) and ports upstream's model reasoning-effort selector
(Shift+Tab effort cycling, effort-aware /model, footer, and /status) onto
the personal module layout.
agent/disposed only cleared the status line, so an agent-loop-only reload that
disposed the agent while the TUI stayed mounted left the local disposed flag
false. Since retained agents accept deliveries after detachment, later input
drove a zombie agent/session. Set disposed on agent/disposed so dispatchMessage
reports it. Adds a regression that sends after disposal and asserts no delivery.
The reference-admission discard listener matched on followup()'s returned id,
but an agent/inbox/enqueue listener that synchronously cancels emits
agent/inbox/discard before followup() returns to assign that id. The match
then missed, leaking both the submit and discard listeners plus the attached
context per referenced prompt. Match on the content reference instead — the
same value send() carries onto the message, known before followup() runs, and
symmetric with the submit wrapper's content check.
The prior "ordinary allowed path" test passed only because the fake agent
returned a fixed 'stub' id that collided with the id the test constructed;
it now releases each wrapper through its own allowed admission, and a new
regression drives the synchronous-discard timing directly.
goal-session rides retry turns and survives admission failures. A
recovery policy closes a goal round's failed turn and reopens its
history under a retry trigger; the attempt now adopts that turn and
drops the failed turn's provisional reason, so the round settles from
the retry's own outcome instead of blocking an armed goal with
turn-error after a successful response. A downstream admission hook
that throws (rather than blocks) used to strand the queued reservation
forever; the listener now clears a still-turnless matching attempt on
the rejection path and reschedules the round.
agent-loop contains a persistently rejecting step close in the catch
path the same way the finally contains the turn close, so the
post-finally tail always publishes the terminal status — previously a
double veto escaped run(), leaving status at running while whenIdle()
resolved. The whenIdle catch arm is annotated as the backstop it now
is: every driver rejection path is contained today.
workspace-context folds an already-appended baseline from the session
log when the plugin is hot-remounted over a live session, instead of
injecting a duplicate from its fresh mount-local guard.
The TUI's reference-admission discard listener installs before
followup(): admission runs synchronously inside it on the common path,
so a listener installed afterwards missed its own cleanup and leaked
one callback per referenced prompt.
The TUI's referenced-prompt snapshot now rides the prompt's own
admission transaction instead of a pre-admission inject: while idle, a
one-shot prepended agent/prompt-submit wrapper appends the snapshot to
the allow decision's additionalContexts, so a blocking hook discards
the prompt and its attached context together instead of stranding the
snapshot in history for the next unrelated prompt. A prompt discarded
before admission releases the wrapper; steering keeps the inject path
since it bypasses admission and drains at the same boundary. The
session-reference snapshot adapter pinned the old context-before-prompt
order; the branch-wide order (prompt first, its contexts after) is now
asserted and the fixture re-recorded.
tool-tasks drops the last consumer of the removed thrown-disposed
contract: completion notices now inject unconditionally, which is
well-defined during owner teardown — the loop treats disposal like any
cancel, so the notice appends as durable idle context (persisted for
resume while the session is attached, dropped with the detached log
after). README pair and the owner-disposal tests state the new
delivery contract.
Tianyi Cui