G's grants already carry a state's max-mib onto its bucket (Bucket.MaxBytes, §469); main's second copy of that cap in
BucketStreamConfig and Composed is dropped for G's, and main's tests (the declared cap row, the held bucket raised)
are kept. The checker keeps G's version 0.1 (§368) with main's compositions max-mib 256.
Builds: design 47 G §469, §368; A-277 (since mesh-delivery #206 the grants part carries max-mib)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The assigner's asks.go, tools.go, seat_tools_test.go and assign/wants.go,
the builder's builds.go and the indexer's candidate_values.go, and the
import blocks of three tests the step runner's move reordered.
Builds: delivery review round 1, the gofmt finding
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
All of mesh-build-go's buildstep Go module leaves mesh-catalog for
modules/mesh-delivery-sdk: the step run, fetch and release and the
regenerate-and-compare step as the package steprun, the CPU share as
steprun/buildcpu, the conformance kit as steprun/conformance, the seat
generator as seatgen with cmd/seatgen, and cmd/mesh-pack. The three
build modules and the builder's and checker's tests import the SDK and
keep no copy; the SDK requires nothing of mesh-catalog. Code the build
modules share is the delivery domain's SDK, not one module's package.
It carries the build-module round's two new files of that package
(dropdetected.go 247c422f, storedbuilds.go 904b8aed).
Builds: design 47 G §464, §680, §327, §420, §495
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
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 checker's fixed names of the build modules (buildmodules.go in its program, rules and check) go: the
Examiner finds each build module by the language its seat's acting rule builds (buildrun.BuildSeats over
the seat list's definitions and the holders the indexer's modules name by their manifests' claims), the facts
carry the language where a level counts by it, and the program's manifest reads the steps and toolchains of
the seats and hears their toolchain-reported. The tools read the steps of each holder and follow the seat
list as the view does. Timers stay (A-118).
Builds: design 47 G A-82, A-118, §204
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
The seat's bodies hold TakeInTaken (branch, source, entries with repository and pull) and TakeInPending (branch); revision's
answer gains none-yet and its revision is optional; meshcatalogclient reads the field, never the error's words. The holders'
and readers' uses of the answer follow (a pointer revision).
Builds: design 47 G §476, §493, §530, §454, §648
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
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
The forge read named the forge module's tools (gitea_get_file, and four more never
called), which ADR 0656 §1 and the operator's (d) forbid: whichever module holds the
seat git answers. An absent file is the seat's absent, an error its error, and a call
that reaches nobody stays not heard, so the check says it could not run.
The compare captures no test reads any more go with the commit trees read nothing called.
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
Builds: design 47 G §252, §312, §317
A check that failed for a part of the mesh not heard was judged again at
every change of the nodes. It now says could not run, naming the part,
and is kept in the checker's own state could-not-run; only the running of
that part asks it again, once per heads and part at one fingerprint
(PartRunning). Not asked again within an hour it is the finding
check-could-not-run, which clears when the heads are checked again,
replaced, or their pull request closes. checker_status reads the state.
The node-engine seat's running event is wired once its definition lands.
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
Builds: design 47 G §253, §315
The checker built from G's heads makes no tag-wanted and waits on no published
of a tagged version: tagwanted.go and tags.go go (the verdict sayer stays, in
verdictsay.go), with st.tagged, groupVersions, Case.Tagged, TagAsks, the
PublishedHeard wiring, the group path's tagged Node package and Go packages,
the manifest's two published consumes and the tag-wanted event of asks.
Builds: design 47 G §528, §373, §321
Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
A large node's request is 1.2 MB, over the bus's 1 MiB message limit, so the compose ask failed once keeping no longer did.
The checker packs the request (a JSON string "gzip+base64:..." in the part's `any` request, no contract change) and
buildstep.ComposeJob.RequestBytes reads either form. A kept request of another version is refused naming it, and listing
no longer reads pieces. An older mesh-build-go refuses a packed ask loudly (it holds no request object): the check fails
naming it, so the catalog change is delivered with this one as a group.
Quick fix (chosen as the recovery): pack the request, ~290 KB for the largest node. Thorough fix (milestone G): the request
named by digest and served on a seat verb, as RunInput is.
Claude-Session: https://claude.ai/code/session_01NcbgfFfQ4Z9UE5jK32PmE5
mesh-host #219's check failed: the request kept for g14 was 300127 bytes compressed, over the 262144 the state takes.
The key of the check and node now holds a pointer (digest, size, pieces) and each piece of the gzip bytes is a value of
its own under pieces/<digest>/<i>, at most 128 KiB; checker_composition_requests follows the pointer and checks the digest.
Quick fix: a larger cap or a better compression: still a limit, met again as the mesh grows. Thorough fix (chosen, and
bounded only by the bucket): pieces. Not an object store: the checker is a foundation part, whose data stays in the old
sections (ADR 0619 §2), and the tool runner serves no old-section object read to its tools. No manifest change.
Claude-Session: https://claude.ai/code/session_01NcbgfFfQ4Z9UE5jK32PmE5
mesh/check checking 4735fc0b5fdb on 8ad2bf6951d5 | novox/mesh-delivery#151 in group feat/registry-milestone
Conflicts in the checker's judge.go, builds.go, composition_test.go and try_candidates.go:
the group's contract types and kept composition path stay, #164's behaviour is taken whole.
WaitsOnPreview reads the contracts' meshcatalogclient.NoneNoPreview instead of a copied
constant; awaitBuilds names its failures failedBuilds, beside the builds package.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
mesh/check failed: novox: not composed on novox: mesh-delivery-checker: frombus: mesh-delivery-checker is wanted at src… | 3f9acf022325 on b719dc0ecf27
The indexer writes a pull request's candidate first from its pull message alone, every module's
fingerprint "no merge preview is computed", and the fingerprints only with the preview, on the same
heads. A check that began on that first write kept its case: it waited on no build, said nothing,
and the engines, reading the previewed candidate, refused the module as not yet built (mesh-catalog
#624, mesh-delivery #163). Such a check now takes the previewed candidate before it judges the
manifests, waits on the builds and composes, as a whole try already does.
A build not ended within the build-wait-bound is named and the change is not composed, never an
engine's refusal of a fingerprint it has no done build of read as the change's.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
mesh/check checking 918d21b7dac5 on 6bc232eb78b9 | novox/mesh-delivery#151 in group feat/registry-milestone
The request sent to the composer is kept gzip-compressed in the state compositions (ttl 3 days) under the check's key and the node, and the tool checker_composition_requests reads one by check and node or lists a check's. A request that cannot be kept fails the node's composition, naming why.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
mesh/check failed (3): ace: the check of novox/mesh-catalog#612 failed: the unit level: the candidate's mes…; and 2 more | 932e7ccf0255 on bab05c66b3cb
G's own check is made by the running B0 checker, which hands no digests, so the
program answers every head definition: 342 KB, over the 256 KiB a run's answer may
be. Without digests the descriptions of verbs, events and their schemas are left
out (171 KB for the group's heads); neither the manifest validator nor the composer
reads one, and the validator's verdicts over the group are the same either way.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The whole heads' list is 342 KB for this group, over the 256 KiB a run's answer
may be. The checker hands the program its reader's seat-to-digest list (5.6 KB)
and lays the answer, the new or changed definitions, over its own list.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The parked composition-at-running path is folded as the lead's ruling Q3: one
path, so no check composes with the running engines' own code (ADR 0656 §5).
The sandbox request no longer drops a change's manifest verdict: it carries the
check's judge's, or none, which the composer reads as accepted (ADR 0578 §3).
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
B0's shape wins: the running checker asks and records facts, the checking program
decides. The registry's types win: the rules and the program read the generated
checker contract and the domain SDK, and no mesh-sdk delivery import is left.
The registry's series, renames and unparsed seat definitions become facts with
their refusals as rules; the tagged versions ride on the case. The program reads
the heads' seat definitions through sdk/seatdefs and the running checker lays them
over its reader's list by seat, reading no file of the heads: the reader's list
is larger than a run input's bound.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
mesh/check checking 9c9ed7a1d8f1 on 2d201f781ced | novox/mesh-delivery#162 in group feat/b0-checking-program
The operator's answer: mesh-build-go reading the checker's state inputs broke ADR 0596 §2. The run ask's input now names
the verb seat:mesh-delivery-checker.input; the checker's tools serve it from the state inputs (Reader.Input, checked
against its digest), and the state is the checker's own again, read by no other module.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
ADR 0656 §5 has every check judged by parts built from the change's heads, and no record keeps the cycle rule in the
running checker. The facts carry each ordering edge's two modules and kind, never manifests: the whole catalogue's
graph is 50 edges, 3195 bytes.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
A level's digest, unmet packages, kept packages' names and test packages are read by no rule: a mesh-catalog check
step's two levels were 78682 bytes as facts and are 4506. A build node with no job run held a check four begin bounds;
it now holds it two.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The running checker asks the checking program twice on the build seat's job run: first the
heads' seat list, which the manifest validator's requests (change and trunk) and the
node-engine composer's carry; then the verdict of the facts it gathered. The program is built
at the group's mesh-delivery head, else at mesh-delivery's trunk, and its input is kept in the
checker's new state inputs as the exact bytes of its digest, read with checker_inputs. A run
ask no build node begins within 5 minutes is put once more, then fails the check naming the
job; a program that does not build or answer fails the check naming it, and the verdict
records who judged it (JudgedBy). A second step-seen hook on a context is added to the
first, so a try still sees its steps.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
Every asking function only records what it met; the one verdict is the rules' of the facts, so nothing in the
running checker judges beside the checking program.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The running checker records the manifest validator's cases, each fetch's programs and the engines' answers as facts,
and the rules package judges them, with the notes it gives a reader, so the checking program judges every rule the
running checker judged. The module graph's cycle stays with the asking: its facts are whole manifests, over a run's
input bound; G's program reads them from the heads.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The running checker records each failure where it is met and the node of each step it reads; the verdict is
rules.Judge of those facts, so the checking program, given the same facts, gives the same verdict. Its own
second outcome and its node set are gone.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The checking program judges a check from its facts (milestone 47/B0,
ADR 0644, ADR 0656): the rules package holds Facts v1, the level rule,
the failure words and the verdict's outcome, and the running checker
records each level answer as a fact and calls the rules. A row of the
every-failure tests judges the gathered facts and finds the running
checker's verdict.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
When the checker's seat list is not read, or a head's definitions are not
taken, the composer at a node-engine member's head was handed no list,
judged no machine, and the composition passed with warnings. It now
fails, naming why (ADR 0635 §3), and no composer is asked. The check's
readers of the members' heads are its manifest validator and that
composer (ADR 0636 §4 insight 1). The composer's Request.Seats contract
says both the judge and the composer read the list, and a composer
handed none judges no machine (ADR 0636 §1).
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The group try of 2026-11-06 10:07 composed with a composer that had no
seat list, so its composition failed on all four machines for paths and
holders the seats define. The checker's one list per check (the
reader's, the members' heads' laid over it, ADR 0636 §4) now goes to the
judge and the sandbox composer alike; a list the reader has not read
goes to neither, and the composer then judges no machine (ADR 0636 §1).
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The group's beside.go and pins_test.go already carry #161's fixes in their
group form (siblings from each sibling's go.mod); try_test.go keeps #161's
case that echoes the end through Put, in the group's event types.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The checker read any touched contracts module's definitions, skipped the
check that a definition names its own directory's seat, and copied the
indexer's tree reading. sdk/seatdefs now holds the reading and the rule of
which contracts module may define a seat; the indexer's keeping, its
revision seats and the checker's head seats all use it. Two heads defining
one seat fail the judging unless one hands it over to its domain's module.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
mesh/check being carried to d58b71724312 on 2d201f781ced | from 205a3b37501f
The end a run writes comes back on the key's watch, where Put stops the run's context and leaves the saying to the run, so the end was never said and the indexer kept the try's candidate. The group try's unit test now echoes the end on the watch while the run says it.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
The third failure of the kind (tries 1, 3, 4): a sibling a test builds, the node-engine's composer, needs every repository it requires at the zero version beside it. besideFor now reads each sibling's root go.mod through the forge and places those repositories as any sibling is placed; the catalogue's rule names only what its tests read themselves.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
jschoubben