mesh-knowledge, claude-code: a long record is reviewed within a bound sized to it, and a proposal waiting on its author leaves the module healthy (hq issue 508)
#222
Merged
mesh-adminmerged 1 commits from fix/508-a-long-record-is-reviewed into main2026-10-11 18:41:55 +00:00
Fixes novox/hq issue 508 (a long proposal cannot be reviewed within the judgement's 25 seconds, and a proposal left unreviewed turns the knowledge base unhealthy and blocks its own rollouts).
1. A long record can be reviewed
Chosen: a bound sized to the question, collected by ticket. Every judgement mesh-knowledge asks carries within = 25 s + 2 s per KiB of question, at most 120 s.
The tool runner ends every bus call at 30 s (mesh-tools node-tools bus.RequestTimeout and launch.CallTimeout), so a bound past 25 s cannot ride on one synchronous call. claude_code_judge therefore takes within (at most 120 s). A judgement not done in 20 s answers pending with a ticket. It runs on to its own bound, and the new tool claude_code_judged (ticket) collects it from the same node, waiting up to 20 s per call. Without within, the judgement behaves exactly as before, so the issue tracker's triage is unchanged.
Why not clause by clause: each judgement pays the agent's start-up time, which is most of the 14 to 25 s that short records take today. A dozen clause-sized judgements would each risk the same 25 s and take minutes together. They would still need a whole-record judgement to check whether the reasoning and links agree with the clauses.
Rollout order: an older claude-code refuses within. mesh-knowledge then asks again without it, so a short record is still reviewed whichever module rolls first.
2. A record waiting on its author does not make the module unhealthy
Health is unhealthy only when the database cannot serve, the records cannot be read, or a rule that judges the whole record set fails (issue 504's case, accepted with nothing citing it, stays unhealthy). A proposal out of review tries is named in the healthy answer's why and told to the operator, as #219 does. Because health no longer raises it, a notice that could not be sent is now marked untold and sent again at the next step.
Tests (written first; they fail on main)
claude-code: TestALongJudgementIsCollectedByItsTicket, TestALongJudgementIsBoundedAndSaysWhyItFailed, TestTheToolsTakeWithinAndTicket. On main they do not compile, because Judge.Wait, Collect and judgedTool do not exist there.
mesh-knowledge: TestALongRecordsReviewGetsTheLargerBound, TestAPendingJudgementIsCollectedByItsTicket, TestAnAgentModuleThatTakesNoWithinIsAskedWithoutIt. On main they do not compile, because judgementOver does not exist there.
mesh-knowledge: TestARecordThatCouldNotBeReviewedLeavesTheModuleHealthyAndIsTold replaces the test that asserted unhealthy. On main it fails: unhealthy after try 3: 1 proposed record(s) could not be reviewed in 3 tries ....
go vet ./... and go test -race ./... -count=1 pass in both modules. mesh-knowledge was run with initdb off PATH, so its 24 postgres-backed tests ran on the memory store and their postgres half was skipped (NOT TESTED ON POSTGRES). The words, agenthome and paths checks pass.
The manifests change too: claude-code lists claude_code_judged, and mesh-knowledge invokes it.
modules/mesh-delivery is untouched (#221 is open there).
Fixes novox/hq issue 508 (a long proposal cannot be reviewed within the judgement's 25 seconds, and a proposal left unreviewed turns the knowledge base unhealthy and blocks its own rollouts).
## 1. A long record can be reviewed
**Chosen: a bound sized to the question, collected by ticket.** Every judgement mesh-knowledge asks carries `within` = 25 s + 2 s per KiB of question, at most 120 s.
The tool runner ends every bus call at 30 s (mesh-tools node-tools `bus.RequestTimeout` and `launch.CallTimeout`), so a bound past 25 s cannot ride on one synchronous call. `claude_code_judge` therefore takes `within` (at most 120 s). A judgement not done in 20 s answers `pending` with a `ticket`. It runs on to its own bound, and the new tool `claude_code_judged` (`ticket`) collects it from the same node, waiting up to 20 s per call. Without `within`, the judgement behaves exactly as before, so the issue tracker's triage is unchanged.
Why not clause by clause: each judgement pays the agent's start-up time, which is most of the 14 to 25 s that short records take today. A dozen clause-sized judgements would each risk the same 25 s and take minutes together. They would still need a whole-record judgement to check whether the reasoning and links agree with the clauses.
Rollout order: an older claude-code refuses `within`. mesh-knowledge then asks again without it, so a short record is still reviewed whichever module rolls first.
## 2. A record waiting on its author does not make the module unhealthy
`Health` is unhealthy only when the database cannot serve, the records cannot be read, or a rule that judges the whole record set fails (issue 504's case, accepted with nothing citing it, stays unhealthy). A proposal out of review tries is named in the healthy answer's why and told to the operator, as #219 does. Because health no longer raises it, a notice that could not be sent is now marked untold and sent again at the next step.
## Tests (written first; they fail on main)
- claude-code: `TestALongJudgementIsCollectedByItsTicket`, `TestALongJudgementIsBoundedAndSaysWhyItFailed`, `TestTheToolsTakeWithinAndTicket`. On main they do not compile, because `Judge.Wait`, `Collect` and `judgedTool` do not exist there.
- mesh-knowledge: `TestALongRecordsReviewGetsTheLargerBound`, `TestAPendingJudgementIsCollectedByItsTicket`, `TestAnAgentModuleThatTakesNoWithinIsAskedWithoutIt`. On main they do not compile, because `judgementOver` does not exist there.
- mesh-knowledge: `TestARecordThatCouldNotBeReviewedLeavesTheModuleHealthyAndIsTold` replaces the test that asserted unhealthy. On main it fails: `unhealthy after try 3: 1 proposed record(s) could not be reviewed in 3 tries ...`.
`go vet ./...` and `go test -race ./... -count=1` pass in both modules. mesh-knowledge was run with initdb off PATH, so its 24 postgres-backed tests ran on the memory store and their postgres half was skipped (`NOT TESTED ON POSTGRES`). The words, agenthome and paths checks pass.
The manifests change too: claude-code lists `claude_code_judged`, and mesh-knowledge invokes it.
modules/mesh-delivery is untouched (#221 is open there).
mesh/merge-gate pass: builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (…
A judgement over one bus call's 30 seconds is collected by its ticket from
claude-code, so an eleven-clause record can be reviewed; a proposal out of
review tries is told to the operator, not raised as the module's fault, so
it no longer holds the module's own rollouts.
Deliverynovox/mesh-catalog@fcab25604e18 — delivered since 2026-10-11T18:49:55Z
Delivery plan — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
tier 0: claude-code, mesh-knowledge
ace: receives claude-code
g14: receives claude-code
novox: receives claude-code, mesh-knowledge
shanks: receives claude-code
claude-code keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before claude-code is judged on its first machine
Machines
claude-code on ace: passed — healthy 3 times over 2m14s
claude-code on novox: passed — carried by release-1791744295592500422
mesh-knowledge on novox: passed — healthy 3 times over 2m7s
claude-code on the rest running it: sent
mesh-knowledge on the rest running it: sent
Transitions
2026-10-11 18:33 (new) → proposed (announced): novox/mesh-catalog#222's head announced
2026-10-11 18:36 proposed → checked (checked): the verdict names this commit
2026-10-11 18:36 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
2026-10-11 18:43 ready → published (merged): on the trunk its modules follow: merged there, and its walk opened — or nothing for a walk to move
2026-10-11 18:43 published → delivering (go): its walk started: let go by this owner in its group's order, by a person, or on the controller's own path — not a failed walk that carried a later commit for it, after which it waits for a walk of its own commit
2026-10-11 18:49 delivering → delivered (done): its walk is done, carrying its own commit: every machine of its deploy plan passed, or was left as its policy says
The commit's note under refs/notes/mesh-plan keeps every transition: git log --notes=mesh-plan.
<!-- mesh-delivery:view -->
**Delivery** `novox/mesh-catalog@fcab25604e18` — **delivered** since 2026-10-11T18:49:55Z
**Delivery plan** — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
- tier 0: claude-code, mesh-knowledge
- ace: receives claude-code
- g14: receives claude-code
- novox: receives claude-code, mesh-knowledge
- shanks: receives claude-code
- claude-code keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before claude-code is judged on its first machine
**Machines**
- claude-code on ace: passed — healthy 3 times over 2m14s
- claude-code on novox: passed — carried by release-1791744295592500422
- mesh-knowledge on novox: passed — healthy 3 times over 2m7s
- claude-code on the rest running it: sent
- mesh-knowledge on the rest running it: sent
**Transitions**
- 2026-10-11 18:33 (new) → proposed (announced): novox/mesh-catalog#222's head announced
- 2026-10-11 18:36 proposed → checked (checked): the verdict names this commit
- 2026-10-11 18:36 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
- 2026-10-11 18:43 ready → published (merged): on the trunk its modules follow: merged there, and its walk opened — or nothing for a walk to move
- 2026-10-11 18:43 published → delivering (go): its walk started: let go by this owner in its group's order, by a person, or on the controller's own path — not a failed walk that carried a later commit for it, after which it waits for a walk of its own commit
- 2026-10-11 18:49 delivering → delivered (done): its walk is done, carrying its own commit: every machine of its deploy plan passed, or was left as its policy says
The commit's note under `refs/notes/mesh-plan` keeps every transition: `git log --notes=mesh-plan`.
mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: claude-code, mesh-knowledge)
mesh/repo-check: PASS — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
Delivery plan — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
tier 0: claude-code, mesh-knowledge
ace: receives claude-code
g14: receives claude-code
novox: receives claude-code, mesh-knowledge
shanks: receives claude-code
claude-code keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before claude-code is judged on its first machine
read by no module's build: modules/claude-code/README.md, modules/claude-code/cmd/claude-code/judge_test.go, modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/judge_bound_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/manifest_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_bounds_test.go
Run by the build seat on ace as build-1791743583993801974 (builds --log build-1791743583993801974).
--- the gate: claude-code, mesh-knowledge
note: claude-code uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
claude-code: ok, 45 tool(s), keeps state servers, holdings, config, proposals, reads claude-licence-manager.bindings, keeps agent-home (valuable); WARNING: settings ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339)
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
long-running resources without health: 0
files asking for a setting without saying whether it is trusted: 1
2 manifest(s) checked. Judged against the seats this binary carries; a claim on one of the mesh's own seats is judged fully at registration, and a seat declared by a module not given here reads as unknown. Identities judged on a 6-character machine name, against the bounds of the providers given here
judging 2 module(s) changed against 4 machine(s), as the snapshot of 2026-10-11T18:32:22Z says they run
composing every machine as the mesh is
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
composing every machine with the change
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
merge gate: PASS — every machine composes with the change as it did without (4 of 4 compose)
judged by controller development build against the facts of 2026-10-11T18:32:22Z
modules the change moves:
- claude-code (changed)
- mesh-knowledge (changed)
machines:
- a machine (f35): composes
- holds mesh-dns-resolver (idt): composes
- holds git, holds mesh-artifact-store, holds mesh-broker, holds mesh-catalog, holds mesh-delivery, holds mesh-dns-resolver, holds mesh-store, holds mesh-vault, holds npm-package-registry, the control node, the hub (ygcyr): composes
- a machine (yvbpsm): composes
a merge would rebuild 2 module(s) in 1 tier(s): claude-code, mesh-knowledge
notes:
- modules/claude-code/README.md, modules/claude-code/cmd/claude-code/judge_test.go, modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/judge_bound_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/manifest_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/review_bounds_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok git.novox.be/novox/mesh-catalog/checks/words 0.002s
words: 12 retired words and 70 daemons' lines, none in the 41934 strings of 678 files
ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.010s
agent-home: no archive and no file written into under the agent's home
ok git.novox.be/novox/mesh-catalog/checks/paths 0.006s
paths: no module names another's state on its machine (3 known, listed to go)
long-running resources without health: 69, as kept
testing modules/claude-code
ok claude-code/cmd/claude-code 3.301s
testing modules/mesh-knowledge
ok mesh-knowledge/cmd/mesh-knowledge 91.578s
--- its merge-check-typescript.sh (typescript toolchain)
**Merge check** at `fcab2560`
- `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: claude-code, mesh-knowledge)
- `mesh/repo-check`: **PASS** — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
**Delivery plan** — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
- tier 0: claude-code, mesh-knowledge
- ace: receives claude-code
- g14: receives claude-code
- novox: receives claude-code, mesh-knowledge
- shanks: receives claude-code
- claude-code keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before claude-code is judged on its first machine
- read by no module's build: modules/claude-code/README.md, modules/claude-code/cmd/claude-code/judge_test.go, modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/judge_bound_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/manifest_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_bounds_test.go
Run by the build seat on ace as `build-1791743583993801974` (`builds --log build-1791743583993801974`).
```
--- the gate: claude-code, mesh-knowledge
note: claude-code uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
claude-code: ok, 45 tool(s), keeps state servers, holdings, config, proposals, reads claude-licence-manager.bindings, keeps agent-home (valuable); WARNING: settings ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339)
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
long-running resources without health: 0
files asking for a setting without saying whether it is trusted: 1
2 manifest(s) checked. Judged against the seats this binary carries; a claim on one of the mesh's own seats is judged fully at registration, and a seat declared by a module not given here reads as unknown. Identities judged on a 6-character machine name, against the bounds of the providers given here
judging 2 module(s) changed against 4 machine(s), as the snapshot of 2026-10-11T18:32:22Z says they run
composing every machine as the mesh is
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
composing every machine with the change
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
merge gate: PASS — every machine composes with the change as it did without (4 of 4 compose)
judged by controller development build against the facts of 2026-10-11T18:32:22Z
modules the change moves:
- claude-code (changed)
- mesh-knowledge (changed)
machines:
- a machine (f35): composes
- holds mesh-dns-resolver (idt): composes
- holds git, holds mesh-artifact-store, holds mesh-broker, holds mesh-catalog, holds mesh-delivery, holds mesh-dns-resolver, holds mesh-store, holds mesh-vault, holds npm-package-registry, the control node, the hub (ygcyr): composes
- a machine (yvbpsm): composes
a merge would rebuild 2 module(s) in 1 tier(s): claude-code, mesh-knowledge
notes:
- modules/claude-code/README.md, modules/claude-code/cmd/claude-code/judge_test.go, modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/judge_bound_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/manifest_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/review_bounds_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok git.novox.be/novox/mesh-catalog/checks/words 0.002s
words: 12 retired words and 70 daemons' lines, none in the 41934 strings of 678 files
ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.010s
agent-home: no archive and no file written into under the agent's home
ok git.novox.be/novox/mesh-catalog/checks/paths 0.006s
paths: no module names another's state on its machine (3 known, listed to go)
long-running resources without health: 69, as kept
testing modules/claude-code
ok claude-code/cmd/claude-code 3.301s
testing modules/mesh-knowledge
ok mesh-knowledge/cmd/mesh-knowledge 91.578s
--- its merge-check-typescript.sh (typescript toolchain)
```
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Fixes novox/hq issue 508 (a long proposal cannot be reviewed within the judgement's 25 seconds, and a proposal left unreviewed turns the knowledge base unhealthy and blocks its own rollouts).
1. A long record can be reviewed
Chosen: a bound sized to the question, collected by ticket. Every judgement mesh-knowledge asks carries
within= 25 s + 2 s per KiB of question, at most 120 s.The tool runner ends every bus call at 30 s (mesh-tools node-tools
bus.RequestTimeoutandlaunch.CallTimeout), so a bound past 25 s cannot ride on one synchronous call.claude_code_judgetherefore takeswithin(at most 120 s). A judgement not done in 20 s answerspendingwith aticket. It runs on to its own bound, and the new toolclaude_code_judged(ticket) collects it from the same node, waiting up to 20 s per call. Withoutwithin, the judgement behaves exactly as before, so the issue tracker's triage is unchanged.Why not clause by clause: each judgement pays the agent's start-up time, which is most of the 14 to 25 s that short records take today. A dozen clause-sized judgements would each risk the same 25 s and take minutes together. They would still need a whole-record judgement to check whether the reasoning and links agree with the clauses.
Rollout order: an older claude-code refuses
within. mesh-knowledge then asks again without it, so a short record is still reviewed whichever module rolls first.2. A record waiting on its author does not make the module unhealthy
Healthis unhealthy only when the database cannot serve, the records cannot be read, or a rule that judges the whole record set fails (issue 504's case, accepted with nothing citing it, stays unhealthy). A proposal out of review tries is named in the healthy answer's why and told to the operator, as #219 does. Because health no longer raises it, a notice that could not be sent is now marked untold and sent again at the next step.Tests (written first; they fail on main)
TestALongJudgementIsCollectedByItsTicket,TestALongJudgementIsBoundedAndSaysWhyItFailed,TestTheToolsTakeWithinAndTicket. On main they do not compile, becauseJudge.Wait,CollectandjudgedTooldo not exist there.TestALongRecordsReviewGetsTheLargerBound,TestAPendingJudgementIsCollectedByItsTicket,TestAnAgentModuleThatTakesNoWithinIsAskedWithoutIt. On main they do not compile, becausejudgementOverdoes not exist there.TestARecordThatCouldNotBeReviewedLeavesTheModuleHealthyAndIsToldreplaces the test that asserted unhealthy. On main it fails:unhealthy after try 3: 1 proposed record(s) could not be reviewed in 3 tries ....go vet ./...andgo test -race ./... -count=1pass in both modules. mesh-knowledge was run with initdb off PATH, so its 24 postgres-backed tests ran on the memory store and their postgres half was skipped (NOT TESTED ON POSTGRES). The words, agenthome and paths checks pass.The manifests change too: claude-code lists
claude_code_judged, and mesh-knowledge invokes it.modules/mesh-delivery is untouched (#221 is open there).
Delivery
novox/mesh-catalog@fcab25604e18— delivered since 2026-10-11T18:49:55ZDelivery plan — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
Machines
Transitions
The commit's note under
refs/notes/mesh-plankeeps every transition:git log --notes=mesh-plan.Merge check at
fcab2560mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: claude-code, mesh-knowledge)mesh/repo-check: PASS — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)Delivery plan — builds claude-code, mesh-knowledge → ace, g14, novox, shanks; no bus step
Run by the build seat on ace as
build-1791743583993801974(builds --log build-1791743583993801974).