A-163: the asker in `accept.<queue>.<asker>` is the asking seat's name (mesh-pipeline-runner,
mesh-catalog, mesh-wants, mesh-builds), never a nickname: the builder asks as mesh-builds (was
"builder"), the indexer's validation as mesh-catalog (was "indexer"), the assigner's runs as
mesh-wants (was "assigner"), the runner's steps on `accept.check.mesh-pipeline-runner` and
`accept.delivery.mesh-pipeline-runner` (were `accept.check.step` and `accept.delivery.step`, where
the holder read `step` as the asker). An asker naming none (the checker) asks on main's subject
(A-184). A build seat's queues are delivery, check and recheck alone (A-170, A-184): buildstep's
AskKinds holds no step, a holder takes a step from check or delivery, and the grants part makes a
worker per queue, never per kind. A holder keys a named ask's step `<asker>.<node>.<id>` and main's
`<node>.<id>` (§481, §483); readers watch their own name's keys. The composition step's step-progress
names the one caller mesh-wants' definition lists for step, read from the seat list (A-294).
Builds: design 47 G §481, §483, §543, §600, A-163, A-170, A-184, A-248, A-294
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder's unpublished count held every package without a reference and named the Node
build module for all of them, so a Go package built by a build reserving none (another
watched branch, a manifest with no MAJOR.MINOR, a candidate) left its key never done and
was sent to the wrong seat. A reserving build's package is published under its version or
its build fails, so the count now holds images alone.
Builds: design 47 G A-289, §371
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The ask carried the reserved version (§573), but the build modules' shared run gave each
artifact's build only the artifact's own version, which no asker sets, so no build got
MESH_VERSION and a Go package was packed as a candidate and never published. The run now
gives every build the ask's version, and the builder names the registry seat each artifact
of a versioned build is published through, so a Go package is published under its
reserved version on go-package-registry.
Builds: design 47 G §370, §573, §458; design 51 *Publishing*
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
A module whose only artifact is a Go package (mesh-wireguard's go-overlay) never had a
done build: PlanOf refused every package of language go. The Go build module already packs
and publishes a Go package; the builder now asks it.
Builds: design 47 G §370; design 51 *Publishing*
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The asks were made at a fixed time while the channel client checked their expiry against the real clock, so from
2026-11-12 every ask "expires before it is asked". The client now takes the same fixed clock.
Builds: design 47 G §568
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder says wait-held for a key queued behind a package, the package as its part and on (package:<ref>), on its
stream waits, and wait-ended published once the key is claimed; a toolchain wait never ends by that, and a package wait
never by a toolchain's end. Files: internal/build/waits.go (Package, On, WaitPublished), internal/build/run.go (Hold at
the queued result, End at the claim), cmd/mesh-delivery-builder/waits.go (Part in the body, the End rule).
Builds: design 47 G A-192, §499
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder's build.ended is said for a failed build as well (outcome, error whole);
the assigner's rule-1 hold reads "the build of M at F failed: <error>" from it and
"is being built" while none is heard. The assigner's reads of the builder's builds
bucket stay (see the report: a restarted assigner holds no ends heard).
Builds: design 47 G §575, §568, A-162 (survey AS23)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The delivery repository's contracts and SDK live under modules/ as
mesh-delivery-contracts and mesh-delivery-sdk, their Go module paths
renamed to match the folders (the TypeScript names already did), and
every importer in mesh-delivery, mesh-catalog, mesh-host, mesh-tools
and mesh-sdk follows in the same group: a design as it should be, one
path per module, checked by the group's check building every importer.
Builds: design 47 G §427, §88
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
mesh/check checking 7f53665bebb5 on bf36f8444562 | novox/mesh-delivery#205 in group fix/47-G-reader-release-2
builder_build_again's answer (BuildAgainResult.Asks: "the ask ids put") was settled when the ask's id was
noted in the key, before its put: the answer could come back naming an ask not yet on its stream (the check's
flake in TestAnAskTheKeyNamesAndNobodyPutIsPutAtStart, "after 0 asks", about 4 runs in 100 under load), and a
put the stream refused after the note was answered as asked, the key's failure unsaid. The watch is now told
after the put; a refused put ends the look and the answer says the key's failure.
Builds: lead answer A-148 (hq G-20, 3ce1dcf: the fix rides the release), issue 1126 note 72 (builder_build_again answers the asks put), ADR 0668 §1 (the note and the put are one act)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The watchman asks the seat that places modules who holds each build seat (seatlist.HolderOf) and watches that
holder's steps; its manifest reads the seats' steps and invokes the seat's assignments. The builder's stuck
reader looks a step up by its seat, the tests find the holders on their test seat list, the watchman's
integration bed answers who holds a seat. Its manifest tests take the manifest's requires.
Builds: design 47 G A-82, A-126
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder's own lookup of the seat holder and read of the router's record (routerBucket, recordVia in
cmd/mesh-delivery-builder/retake.go) are deleted: the ask is published and the router's decided heard
through operatorask.Asker, and each warrant is held to its ask and acted on once by the operator-channel
client's Take.
Builds: design 47 G A-101 (one operator-channel asker in mesh-sdk), §700, A-74
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
Builder.Candidate, ForgetCandidate, considerCandidate, trunkHead, TrunkMoved, the cands and via
maps, the package-member wait that read via.Members (memberBuilds, memberEnded, MemberBoundEnded,
the member bound key) and the committed build output modules/mesh-delivery-builder/mesh-delivery-builder
are removed. Nothing fed Candidate in the process: the builder follows no candidate and learns a
run's builds from the build step ask. Their tests go with them.
Builds: design 47 G A-84 (the builder follows no candidate), §194, §348, §473
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The follower closes Ready before it tells its entries, so the sweep at ready could close a building key
whose entry was not taken in yet and the entry then asked again (four asks where two). The sweep now
takes the fingerprints of the halves' pictures as current.
Builds: design 47 G §348, §479 (the builder reads the catalog half per watched branch), issue 1660's sweep
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The seat mesh-builds' accepted kind step (cherry-picked contract 0dc8c723 of feat/47-G-recover-runner) names an entry by
fingerprint; the builder reads it through the catalog half or the seat mesh-catalog's entry, builds it through its one claim, plan
and lanes (a build per fingerprint, no pull request number) and says the end as step-ended.<node> under the ask's id, naming each
entry as asked. A build already done is passed at once; the end of a build wakes its waiting step through the log entry.
Builds: design 47 G §230, §194, §473, A-84, A-91, A-97, A-99
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder finds a build module as the indexer's seed does: the build seat whose acting rule
`build` has an `on` matching the artifact's language or kind:image, and that seat's holder, read
from the seat list's definitions and the claims of the catalog half's manifests. Task.Module, every
lane key, the unpublished modules, the toolchain sources, the ends watched, the sayer and
builder_stuck (through `<seat>.steps`) derive from it; the floor's sources are the holders of the
seats git and mesh-catalog. buildrun.GoBuildModule, NodeBuildModule, LanguageOf,
buildstep.BuildModuleOf use, build.ImageBuildModule and StuckBuildModules are gone from the builder
(the checker keeps its own names in two small files until its part reads them from the list).
Builds: design 47 G §7, §428, §704, §204, A-82
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
When a registry seat says published, the builder says the seat mesh-builds' package-published once per artifact the reservation
newly marks: the artifact reference, version and digest published, the watched branch and the source revision kept on the
reservation from the entry the build came from (A-57), and the dependencies of kind 1 (resolved) and 3 (toolchain) the step result
named (§466, §673), their parts the artifact references the entry's edges name. The reservation keeps the dependencies from the
build's end until the publication.
Builds: design 47 G §52, §334, §349, §371, §466, §673, A-57
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
An entry the builder takes from the catalog half carries its origin: the watched branch, the source revision of the catalog revision it
came from and whether the branch is the first of the module's registration. A build of another branch reserves no version and is
built and delivered by its fingerprint; a reservation of the first branch keeps the branch and the source revision for the
publication.
Builds: design 47 G §371, A-57
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
builder_build reads the builds and asked states, and the tools bundle follows the catalog half itself (registrations for the
watched branches, revision per branch) for `newest`; the manifest names the tool and the two seat verbs.
Builds: design 47 G §589, A-87
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder adopts meshcatalogclient's branch follower on its views reader (A-5, A-15): the modules, their manifests (kept from the
entry that holds them, by its digest), their packages and each package's name in its ecosystem are read from the catalog half's
entries, never the indexer's modules, manifests, candidates and packages views; an entry of a branch other than the module's first
builds its fingerprint and supersedes only its own branch's older one; the start's sweep runs once every followed half's first view
is in; the builder's tools bundle reads package names through the same follower. A run's candidate is no input of the builder
(A-84): its read of the indexer's candidates view goes, and the entry point for a run's build ask is not built (no line gives the
ask's form). The seats' events the builder's own handler takes are named per seat so the follower's events have one taker.
Builds: design 47 G §348, §479, §614, A-5, A-15, A-84
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder's asker on the seat operator-channel (level approve, about the key, state asked), the teller's three bus
adapters (publish, the router's record through the seat's holder on the seat list, decided heard through the hub's
events handler); the manifest uses the seat and consumes decided.
Builds: design 47 G §568, A-74
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The reservation keeps the fingerprint of the catalog entry it was made from (versions.Source, A-76). The state sections
publish-queue (max-mib 4) and acted (max-mib 1); a consumer acting once per message of the node-engine seat's running and
manifest-reads-changed (why returned) by its stream sequence, the running of every subject read at the start with
lastPerSubject; one queue per holder sent one at a time as the registry seat's publish with the digest and no parts, the
seat and the package name read from the entry (A-73, A-76); builder_publish_queue and the new builder_publish_again, a
hand act. The old form goes: build.NewPublishAgain / ModePublishAgain, versions.Acts.PublishAgain and the
Again* decision, and the integration step that refused it.
Builds: design 47 G §588, §369, §695, A-73, A-76 (G-16, G-18, G-20)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
A candidate's module is fetched against each package member of its group by the digest of the member's build (fetch `builds`),
the member out of the fetched packages; a member not built holds it with wait-held naming the member, its bound a key of the
builder's bounds state with its own time to live (A-77), a failed member or one missing past the bound fails the step naming it.
Builds: design 47 G §373, §465 (A-77)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The delivery SDK's buildstep gains AskerBuilder, AskSubjectBy and ParseAskSubjectBy (an ask's subject names its
asker as the last token, main's form names none; ParseAskSubject reads both). The builder's put and its held-ask
reads use the builder's subject. Not built here: §543's refusal (the build modules'), the grants (grants part).
Builds: design 47 G §481, §484, §543, §600
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The join with CT-SD4 (6351e988) gives ParseStepKey four values (G §484, §649); the callers that read only the node or the id follow.
Builds: design 47 G §484, §649
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
An entry whose fingerprint names a toolchain version of its language other than
the agreed one is held in the state waits (one key <module>.<fingerprint>), says
wait-held and wait-ended (ran when the agreed toolchain names the version,
released by builder_release) on its stream waits, says wait-held again for every
key at its start; builder_waits and builder_release; its setting
toolchain-wait-minutes. Not built: the bound's finding toolchain-waits (its
scope is not in the design).
Builds: design 47 G §672, §267, §276, §633
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The builder agrees a language's toolchain from the three build seats'
toolchain-reported, kept in its state toolchains-heard (read back at its start),
forgets a node's on the node-engine seat's running with no fingerprint (stopped)
and never on a placement, says toolchain-agreed with its digests list only when
the agreed version changes; builder_toolchains and builder_toolchains_drop;
the manifest's reads of the build modules' toolchains are gone. The gate over
the assignments is now PlacedNodes, which only counts nodes.
Builds: design 47 G §492, §604, §546, §472, §574
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The state tag-asks and its bounds, builder_tag, builder_tag_asks and builder_untag, the checker's tag-wanted
consumer, the branch end's removal of tagged versions, the numbering of a check's build under a tag and the
event version.numbered (its contract in builds, its emit, its manifest line) are removed; a forge message still
wakes a floor. builder_publish_again refuses a held build, since nothing says a version numbered.
Builds: design 47 G §373, §321, §349, §413
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
A build the build node kept, which answers its digest and no reference, is done; its end carries
kept-on once for the build.
Builds: design 47 G §671, §379, §271, §526
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
Taken from f051de32 (the old numbering's refusal); the tests of D-B2 to D-B6 it sat on are left out.
Builds: design 47 G §55 (and its check line §56, test 70)
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The reservation keeps the repository and commit of the build and whether the module has a Go package; version-released writes
the tag through the seat git's verb tag before the key goes, and a tag the forge did not write leaves the release to be heard again.
Builds: design 47 G §370
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The builder keeps one key per module and MAJOR.MINOR in its state reserved, written by compare-and-set when a trunk
build is claimed; nothing is reserved, and the build says wait-held, before the catalog says versions-filled; the key goes
on version-released; a built reservation unpublished past ten minutes is the finding version-unreleased; builder_reserved,
builder_reserved_drop (hand act, recorded) and the seat mesh-builds' verb reserved.
Builds: design 47 G §369, §372, §52, §85 (reserved)
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The look arms a key of the state bounds with its own time to live; the bus's marker of its end raises call-unanswered
about the build key while it is still building and no look holds it, naming builder_build_again; the key's next written
end deletes the bound and clears the finding. A bound that cannot be written raises at once. No timer.
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
The put carries the ask's id as its message id, so it is made again through a deadline and no responders within the bound;
a put or a note that still meets one fails the key as a transport fault, which the builder does not retake by itself
(ADR 0664 §3) and the hand tool does. A key left building with no look holding it (its end never written: "claimed", no
ask) is taken as it stands by builder_build_again, from another builder too.
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
LogEnded copied the outcome, manifest, bases, artifacts and imports of the build's key but not its error or step, and the log entry had no field for them: a failed build read back as only "failed".
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
jschoubben