Carries PR #84 (its head `78ca24d` is an ancestor of this one), so this single merge delivers both rounds.
CB-590 collapses the CB-307 reply schedule and the CB-588 ticket schedule into one per lead, which closes the double-injection race. The review round found that this also collapsed the two reminder caps into one shared budget — so a busy reply stream could exhaust the cap and a ticket arriving afterwards would never be nudged at all. That was a regression, not a pre-existing wart: before this PR the two sources had independent counters.
The follow-up keeps the single schedule and gives each source its own budget. `decide()` returns INJECT while either source has pending work under its own cap, and STOP only when neither does. `stopOrRestart` and its snapshot-diff race logic are untouched.
Verified by the lead: `mvn -f bridged/pom.xml clean install` unpiped, exit code captured — Tests run: 814, Failures: 0, Errors: 0, Skipped: 0 — BUILD SUCCESS. `ReplyPushLoopTest`: 41 tests.
Known and deliberately out of scope: an item arriving during the 15s backoff is already inside the tick's "before" snapshot, so at-cap work can be abandoned rather than nudged once. That shape predates this PR in the ticket-only loop and is filed as #87 (CB-598).
Carries PR #83 (its head is an ancestor of this one), so this single merge delivers both rounds.
CB-527 bounds prefetch. CB-528 makes a publish wait for its broker confirm, so a failed publish is never reported as success, and correlates a `mandatory` Return back to the right publish.
The review round on #83 found one real race: `failPendingPublishesOnRecovery` swept the pending maps without holding `publishChannelLock`, so a publish issued on the already-recovered channel could be failed by the sweep — the exact inversion of what CB-528 exists to prevent, and undetectable by dedup because `MessageService.reply` mints a fresh msgId per call. Fixed by taking the same lock. `close()` now fails in-flight publishes promptly instead of letting them time out after 10s.
Verified by the lead, not taken on the worker's word:
- `mvn -f bridged/pom.xml clean install` unpiped: Tests run: 809, Failures: 0, Errors: 0, Skipped: 0 — BUILD SUCCESS
- `mvn test -Pcontract -Dtest=AmqpReplyInboxContractTest` against a real broker: Tests run: 8 — BUILD SUCCESS
That second run mattered: #83's own new tests are contract-tagged, so the default suite and CI prove nothing about them.
decide(lead, reminderCount) shared one counter across the reply and
ticket sources after PR #84 collapsed both onto a single per-lead
schedule. A reply stream that used up the whole budget could then
make decide() STOP even for a ticket that had never been nudged and
had coalesced onto the same still-active schedule — stranding it with
no live schedule left, since stopOrRestart's racedIn check does not
save work that was already present in the "before" snapshot.
decide() now tracks a per-source count (replyReminderCount,
ticketReminderCount) and returns INJECT while either source is still
under its own cap, STOP only when both are exhausted. Still exactly
one schedule per lead; stopOrRestart's snapshot-diff logic is
untouched.
failPendingPublishesOnRecovery walked and cleared pendingBySeq/pendingByMsgId
without holding publishChannelLock, so a publish() that registered while the
sweep was still iterating could be failed even though it published on the
already-recovered channel — a successful publish reported as failed, and
since MessageService.reply mints a fresh msgId per retry, dedup can't catch
the resulting duplicate. Guard the sweep with publishChannelLock: publish()
only holds it for the seq/map-put/basicPublish, so the sweep can only ever
wait for an in-flight basicPublish to return, never a broker round trip.
Also make close() fail in-flight publishes immediately with a clear message
instead of leaving them to idle out the 10s confirm timeout, and record the
(currently unreachable) msgId-uniqueness assumption pendingByMsgId relies on.
AmqpReplyInboxRecoveryRaceTest drives the sweep and a real publish() against
each other directly (no live broker reconnect) using Proxy-backed fake AMQP
channels and a large in-flight backlog to make the race window observable;
confirmed it fails without the guard (reply "fresh" wrongly failed as
"connection recovered mid-publish") and passes with it.
Adds scripts/bridged-launchd-wrapper.sh so the launchd-run daemon still gets
WORKER_GITEA_TOKEN/AI_GATEWAY_TOKEN by execing through a login shell (launchd
never sources secrets.sh itself). bridged now logs at startup which required
token env vars (derived from each profile's tokenEnv/gitTokenEnv, not a
hand-written list) resolved or are MISSING, by name only. Fills in the real
paths in deploy/dev.ltms.bridged.plist for this host and points it at the
wrapper. scripts/redeploy-bridged.sh now detects a loaded launchd agent and
uses launchctl unload/load instead of a raw kill+nohup, because a bare
SIGTERM exits this JVM at 143 (measured) which KeepAlive.SuccessfulExit=false
reads as a crash and would race the script's own restart; --check reports
installed/loaded state and stays read-only.
Both reply-queued and ticket-terminal nudges could independently decide
to inject into the same lead pane in the same window, since they ran as
two separate schedules keyed differently (worker target vs. lead) that
never checked each other. Replace both with a single per-lead schedule
(activeLeads) that drains pending reply targets and pending tickets
together, sends at most one combined nudge per tick, and shares one
reminder cap across both sources — so two injections into the same pane
can no longer overlap, and work queued while the lead is busy is never
lost, only deferred.
CB-527: basicQos(prefetch) on the consume channel before basicConsume, configurable
via broker.prefetch (default 32), so an undrained inbox backlog stays on the broker
instead of growing the JVM heap without limit.
CB-528: publish moves to its own confirm-mode channel with mandatory=true and a
return listener, so an unroutable or unconfirmed reply now throws instead of
vanishing silently. The confirm callback checks the per-message returned flag
(set by the return listener, which the broker always fires before the matching
confirm) so an acked-but-returned publish is still reported as a failure. The ack
path stays on its own channel/lock and never waits on a publish confirm.
CLAUDE.md told every member 'You mount only the bridge MCP' and told the lead
'a worker mounts only the bridge MCP and cannot run your other tooling'. Both
were false for Claude Code members.
Measured by spawning one member per backend and asking each what it actually has:
opencode (gx) 11 bridge tools only claim TRUE
claude-code (local) 11 bridge + 45 gitea + 2 context7 claim FALSE
Source is ~/.claude.json user-scope mcpServers; --mcp-config adds to that scope
rather than replacing it, so the worktree parity overlay (which correctly
neutralises .mcp.json and opencode.json) cannot see or stop it.
The forge tools are mounted but not usable: CB-592's blocked sentinel means
get_me and list_issues both fail with 'invalid username, password or token'.
That is defence in depth working in a path it was not designed for, so the text
now says a mounted tool is not a working tool rather than pretending the tools
are absent.
Block propagated byte-identically to wiki 7-Use-Cases.md (wiki 1d95e3f).
systems/vms moved the LLM route timeout again, 1800s -> 86400s (24h), after
the silent-truncation risk was discussed. They tried request: 0s first: it
removes the total-duration timer, but on an AIGatewayRoute the idle timeout
is derived from the request timeout, so 0s also removed any bound on a
stalled connection.
At 86400s our own MessageService.ASYNC_TIMEOUT_MS (30 min) binds first, so a
runaway request now ends as a clean FAILED ticket we raised instead of a
silently truncated 200. While the gateway sat at 1800s the two numbers were
equal and did not nest.
`local` runs on /anthropic and `gx` on /v1, both weight 100; `local-direct`
stays weight 0 as the escape hatch.
systems/vms fixed both blockers, and each was re-checked from this side rather
than taken on trust:
listener buffer 32 KiB -> 32 Mi ours: 1.2 MB body -> 200 (was 413)
LLM route timeout 60s -> 1800s ours: 101s stream -> 200,
message_stop present, 4000/4000
Neither was deliberate: 32 KiB was Envoy Gateway's default
per_connection_buffer_limit_bytes, and 60s was Envoy AI Gateway's own default.
The 60s bounded GENERATION as well as prompt size — a tiny prompt with a long
answer returned 504 at 60.05s.
Verified with real workloads, not liveness probes. A `local` member read this
document and CLAUDE.md in full — 48,344 bytes of file content, comfortably past
the old 32,768 ceiling — and answered four questions correctly, including the
document's length (said ~456, actual 455). A `gx` member did the same. The
trivial 3-question probe is what hid the 32 KiB ceiling for an afternoon, so it
no longer counts as proof here.
§7.2 is new and is the part that matters later. One risk is ACCEPTED, not
solved: on a mid-response timeout over chunked HTTP/1.1, Envoy ends the chunked
encoding cleanly instead of resetting, so a truncated answer arrives as HTTP 200
with no error and no terminator (envoyproxy/envoy#17186, acknowledged 2021,
never fixed; the Dec 2025 fix#42269 is HTTP/2 only and SSE here is HTTP/1.1).
Measured at the old 60s: 200, 61.07s, 2473 of 4000 emitted, message_stop 0,
error events 0, ending on a well-formed frame.
The recommended defence — reject a stream with no terminator — does NOT
transfer to us: Claude Code and opencode are third-party clients and we do not
own their SSE parsing. So this is acceptable because a request would have to run
1800s to trip it, not because we could detect it. If a member ever returns a
confident but truncated answer, suspect this before anything in our own code.
Also recorded, from the upstream bisection: ClientTrafficPolicy is honoured in
standalone `aigw run` but BackendTrafficPolicy is silently ignored, and nothing
external distinguishes them (envoyproxy/gateway#9513). Same silent-default shape
this repo keeps hitting.
bridged.yaml carries the same notes inline (gitignored, so not in this commit).
Refs: gitea #76
I wrote "Caddy request_body max_size and/or Envoy's own" and marked it
unverified. The Caddy half was wrong, and an unverified guess still points the
next reader at the wrong component.
Confirmed by the systems/vms side: Envoy Gateway defaults a listener's
per_connection_buffer_limit_bytes to 32768, and aigw buffers the WHOLE request
body before it can route on the model name. So that default is not a network
tuning knob — it is a hard ceiling on prompt size. From the live config_dump:
listener default/llm/http per_connection_buffer_limit_bytes: 32768
Nobody chose 32 KiB; it was inherited.
Both TLS edges are innocent, and the technique that showed it is better than
mine: both 413s carry x-llm-consumer, a header their auth proxy sets only AFTER
authenticating, so the body cleared both edges and the auth. On llm.vm, aigw
413s at 39 KB while the vLLM backend answers 200 at the same size. I found the
boundary; they found the component, by reading the failure's response headers.
Consequences recorded in the doc:
* DO NOT plan around 32 KiB. The intended ceiling is far higher, so sizing our
profiles to it would be designing around a bug.
* Their fix (ClientTrafficPolicy, bufferLimit: 8Mi) is written but NOT
deployed, pending their operator's approval. We do not re-test until they
confirm — a half-changed system gives a number neither side can trust.
* In standalone `aigw run` a SecurityPolicy is accepted and then silently
ignored, so "the config was accepted" proves nothing there. They will verify
by re-reading the live config_dump and sending a large request. Same
silent-default shape this repo keeps hitting, one layer down.
bridged.yaml carries the same correction (gitignored, so not in this commit).
Refs: gitea #76
Deployed U1-U2c, restarted, spawned both new profiles for real, then reverted.
llm.ltms.dev answers HTTP 413 above 32 KiB (32768 bytes), on BOTH surfaces:
/v1 32695 bytes -> 200 /anthropic 32095 bytes -> 200
/v1 32795 bytes -> 413 /anthropic 32855 bytes -> 413
That is far below one agent turn. It is an edge limit (Caddy request_body
max_size, and/or Envoy), so the fix is in systems/vms, not here.
The part worth recording is how it nearly passed. Two members, same message,
same moment: `local` finished in 66s, `gx` never finished at all. `local`
passed only because the probe was three trivial questions in a fresh session,
so the request fit under 32 KiB — the profile looked healthy and was a
landmine set to fire on the first turn that reads a file. So §7's checklist
was not wrong, it was too easy; it now says to use a file-reading task.
opencode's failure mode is worse than a crash: it catches the 413, compacts
its context, retries, and loops. Observed 10+ minutes BUSY with no reply. From
the lead's side that is indistinguishable from a slow worker. Reproduced
outside the bridge with the launcher's own generated config, which is how it
became a one-line error instead of a hang; §7.1 records that procedure.
Everything else about the migration checked out and is recorded so it is not
re-tested: token accepted on both surfaces, unauthenticated 401 (the Caddy
proxy does gate, whatever the gateway's own fail-open policy does),
/v1/models exactly ["deepseek-v4-flash"], the guard allowlist accepted
llm.ltms.dev, and the generated opencode provider block is correct with a real
llmk- key.
Also answers §3b's open question: reasoning survives BOTH surfaces —
/anthropic returns a real "type":"thinking" block and /v1 returns a populated
reasoning_content. The feared /v1 translation loss did not happen.
Config state (bridged.yaml is gitignored, so it is described rather than
committed): `local` back on http://gx00.gw:8000, `gx` kept at weight 0,
`local-direct` kept, llm.ltms.dev left in the guard allowlist. The file
carries these numbers and the exact two-key edit to switch back.
Verified after the revert with a task that reads two large files: correct on
all three questions. Daemon pid 66745, jar f1fd659423e6.
Refs: gitea #76
Live check on a member pane showed the CB-592 shadow did NOT take effect:
GITEA_ACCESS_TOKEN inside the pane was still the real admin token.
Measured cause. The overlay itself works — GITEA_TOKEN is injected the same
way, is exported by no shell file, and does reach the pane. The sentinel loses
one step later. A herdr pane runs a LOGIN shell, ~/.zprofile line 41 sources
${SHARED_ENV}/tools/secrets.sh, and that file does a plain unconditional
`export GITEA_ACCESS_TOKEN=...`. A login shell overwrites a value already in
the environment, so the real token is put back before the member starts.
Confirmed directly:
GITEA_ACCESS_TOKEN=cb592-sentinel zsh -lc ...
-> RESULT: sentinel was OVERWRITTEN by the login shell
This defeats any launcher-side overlay for any name secrets.sh exports. No
change in this repo can win it alone.
So this adds the half that does survive: BRIDGED_MEMBER=1, a name secrets.sh
never exports. It is a no-op until the operator guards the export:
[ -n "${BRIDGED_MEMBER:-}" ] || export GITEA_ACCESS_TOKEN=...
Setting it now costs nothing and makes that one line the whole remaining fix.
The sentinel stays: it is correct for any peer kind whose pane does not start
a login shell, and it keeps the intent explicit where every adapter passes.
Also corrects the javadoc and the test javadoc, which both claimed a
protection that was measured not to hold.
The other reported failure was my own bad test, not a regression. The probe
called /api/v1/user, which a minimal write:repository token cannot read. Same
token on the repo endpoint answers 200, so CB-302 is intact:
GITEA_ACCESS_TOKEN: /user=200 /repos/lms/claude-bridge=200
WORKER_GITEA_TOKEN: /user=403 /repos/lms/claude-bridge=200
Tests 805 -> 807. Both new tests proved to discriminate by reverting the
marker: everySpawnMarksThePaneAsAMember and
aProfileEnvEntryCannotClearTheMemberMarker both fail without it.
Refs: gitea #77
A live probe showed every spawned member carried the admin GITEA_ACCESS_TOKEN: 108
environment variables in a member's pane against 99 in the primary's. The operator's
rule is that only the leader and architects may use it; everyone else uses
WORKER_GITEA_TOKEN. We were not enforcing that at all.
The cause is invisible from inside the launcher. baseEnv builds a fresh map holding only
PATH and the profile's env:, so a member looks like it gets a small explicit environment.
That map is an OVERLAY: WorkspaceControl.createTab/splitPane send only the keys it
contains, and herdr spawns the pane from its own login-shell environment, so every key we
never mention passes straight through — admin token included.
The fix puts a non-blank sentinel over the key in baseEnv, applied AFTER the profile's
env: so no profile, present or future, can restore the real token by naming it in config.
One place, every adapter, including peer kinds not yet written — deliberately not a
per-profile bridged.yaml entry, which is the silent-default shape this repo has shipped
nine times.
A non-blank sentinel rather than the empty string, on purpose: whether an empty overlay
value overrides an inherited variable or is skipped as blank cannot be settled from this
repo, because herdr's merge happens in an external process. baseEnv's own PATH seeding
(CB-511) already relies on a non-blank value replacing an inherited one, so this reuses
the shape that is demonstrated to work rather than the one that is merely plausible.
CB-302's repo-scoped GITEA_TOKEN grant is untouched — a worker can still open its own PR.
The subscription boundary was checked and is unaffected: the primary's pane carries no
ANTHROPIC_* at all, so nothing is inherited there.
Closes gitea #77. Live verification follows separately: the daemon must be redeployed
before this reaches any pane.
herdr spawns a pane from its own login-shell process env and layers our map on
top, so any key baseEnv never mentions passes straight through — including the
admin forge token. baseEnv now puts a non-blank sentinel for
GITEA_ACCESS_TOKEN, applied after the profile's own env: so no profile can
restore it. One place, every adapter, every profile including future ones.
CB-302's GITEA_TOKEN grant (applyGitToken) is untouched.
Operator's rule, 2026-08-15: only the leader and architects may use GITEA_ACCESS_TOKEN;
everyone else uses WORKER_GITEA_TOKEN.
opencode.json is TRACKED, so it ships in every worker worktree, and it mounted the gitea
MCP with {env:GITEA_ACCESS_TOKEN}. A live probe confirmed that variable actually resolves
inside a member: herdr spawns each pane from its own login-shell environment and layers
the launcher's map on top, so a member sees 108 variables rather than the small explicit
set baseEnv appears to build. That gave an opencode member admin forge TOOLS — enough to
merge its own PR, which both CLAUDE.md and the member contract forbid.
This is the narrow half of the fix: it removes the tooling. The admin token is still
present as a string in every member's environment, which is the real defect and is
tracked as CB-592 (gitea #77) — that fix belongs in the launcher, in one place, not
per-profile in bridged.yaml where a sixth profile would silently reopen it.
.mcp.json keeps GITEA_ACCESS_TOKEN and is correct to: it is skip-worktree, the primary's
own local copy, and the primary is the lead. That is the pattern this change follows —
the shared tracked file grants least privilege, and anything needing more overrides
locally.
The gateway (llm.ltms.dev) replaced Bifrost on 2026-08-15 and serves an Anthropic
surface and an OpenAI surface, so both member kinds can point at it. The plan is in
docs/CB-591-Gateway-Migration.md; gitea #76 tracks the work.
The opencode half needs no code: OpenCodeLauncher already pins an OpenAI-compatible
endpoint (CB-508), so baseUrl + tokenEnv + provider/model is a config change. That
matters more than it looks — every opencode member today is sol or terra, and both sit
on one OpenAI account via credentialId: openai-shared, so an exhaustion on either locks
out both. A gateway-backed opencode profile is free and off that credential, which
retires a single point of failure rather than only adding capacity.
Also extends the redeploy script's --check to AI_GATEWAY_TOKEN. A profile's tokenEnv is
resolved from the DAEMON's own environment by HerdrPeerLauncher.resolveEnv, so a token
added to secrets.sh after the daemon started is simply absent: the launcher injects an
empty token and the gateway answers 401, long after the restart and with nothing tying
the two together. That is the same trap as WORKER_GITEA_TOKEN, and it gets the same
login-shell check that never prints the value.
An async delegation ticket (bridge_send wait:false) resolves on MessageService.reply's
rendezvous fast path, which returns before onReplyQueued. So CB-307's push loop only ever
heard about the durable-inbox case, and the mode CLAUDE.md tells leads to prefer never
nudged anyone. Closes gitea #72.
Adds a second, independent reminder schedule keyed by the lead terminal, so several
tickets finishing together coalesce into one nudge. The CB-307 path is untouched.
Three defects were found in review and fixed before merge:
* a pendingTickets entry outlived the ticket it named. poll() returns null once
pruneTerminalTickets drops a ticket, so ticketCollected was never reached and the
entry leaked for the daemon's life, riding along on every later nudge and sending
the lead after a ticket bridge_poll can no longer find.
* a lost nudge: a ticket landing between decideTickets returning STOP and
activeLeads.remove coalesced onto a schedule that was about to die. That is the
exact failure this ticket exists to remove, reintroduced in a narrow window.
* the success direction was unpinned in tests, and the comment listing the paths that
complete the future was short by several.
The obvious fix for the second one was wrong: restarting on any pending ticket defeats
the reminder cap, because a never-collected ticket at cap is expected to still be there.
The fix diffs against a snapshot taken before the decision, so only a ticket that truly
arrived during the window restarts the schedule.
Verified on my own unpiped build: 802 tests, 0 failures, BUILD SUCCESS.
Two reviewers on the diff; the loop-gating finding they raised is split out as CB-590.
- ReplyPushLoop.stopOrRestartTicketLoop: after releasing a lead's active-schedule
slot on STOP, restart only if a ticket landed that the pre-decision snapshot
did not already account for. A naive "restart on any pending ticket" version
was tried first and reverted: it defeated the reminder cap by restarting
forever on a stale, never-collected ticket (broke
successfulTicketNudgeIncrementsDelivered and ticketNudgesSendUpToCapThenStop).
Diffing against a pendingBefore snapshot distinguishes a genuine race arrival
from stale cap-exhausted backlog.
- Two new ReplyPushLoopTest cases exercise stopOrRestartTicketLoop directly
(now package-private) rather than forcing the underlying thread race:
aTicketStillPendingWhenTheLoopStopsIsNotStranded (the race must restart) and
aStaleUncollectedTicketAtCapDoesNotRestartTheLoop (the cap must still hold).
Both were verified to fail against deliberately-reverted versions of the fix
before being restored to green.
- MessageServiceTest: pin the success-path nudge test's negative direction too
(must not contain "FAILED"), not just the failure-path test.
- MessageService: complete the whenComplete comment's list of completion paths
(TIMED_OUT/BUSY/BACKEND_EXHAUSTED via finishAsyncTask, completeExceptionally
on throw, and answer() -> finishAsyncTask(turnId, result)).
tasks is the sole authority on whether a ticket exists, but
pruneTerminalTickets dropped entries from it without telling
ReplyPushLoop. ticketCollected(ticket) was only ever called from
MessageService.poll's terminal branch, which pruneTerminalTickets
short-circuits past once a ticket is gone (poll returns null at the
top). A ticket the lead never polled — or one the reminder cap already
gave up on — was pruned from tasks but never collected in
ReplyPushLoop.pendingTickets, so it rode along on every later nudge to
the same lead forever, naming a ticket bridge_poll could no longer
find, and the map itself never shrank.
pruneTerminalTickets now calls pushLoop.ticketCollected for every
ticket it actually removes (guarded on pushLoop != null), so a pending
nudge entry lives exactly as long as its ticket is pollable. Reused
tasks as the only removal trigger rather than adding a second live
query back into MessageService — no new source of truth.
Added an injectable clock (LongSupplier nowNanos, defaulting to
System::nanoTime) to MessageService, mirroring the SessionManager/
SessionReaper nowNanos seam, so a test can cross the 10-minute
TICKET_TTL_NANOS deterministically instead of sleeping for real.
Confirmed the new regression test fails against the prior
pruneTerminalTickets (a stale ticket rides along on a later coalesced
nudge) before restoring the fix.
An async bridge_send(wait:false) registers a rendezvous waiter, so its
reply always takes MessageService.reply's fast path and returns before
ReplyPushLoop.onReplyQueued is ever called — the exact mode the charter
tells leads to prefer never nudged.
Add ReplyPushLoop.onTicketTerminal(ticket, target, failed), a second
entry point reusing the loop's status gating, bounded/backoff reminders
and metrics, keyed by the nudge-receiving lead so several tickets
finishing together coalesce into one nudge naming the count. MessageService
wires it via task.future.whenComplete in sendAsync (covers reply,
completion fallback, wedge, and abandon() alike) and calls the new
ticketCollected(ticket) from poll() once a terminal view is handed back,
so an already-collected ticket is never nudged again. The CB-307 inbox
path (onReplyQueued/decide/NUDGE_FORMAT) is untouched.
The comment said non-positive means unlimited at load. That stopped being
true when CB-585 made an explicit maxLoad: 0 survive as a real cap of zero
and made a negative value refuse config load. The code below it was already
right — only the comment described the old normalisation. Flagged by the
CB-585 worker, which correctly stayed out of a file not on its list.
Wires up the two links that made session resume unreachable: MemberSession
now records agentSessionId from PeerHandle at acquire (both the plain and
worktree paths), and it survives onto the roster (rosterView, so both
bridge_list and GET /members show it). bridge_spawn accepts sessionName and
resumeSessionId, threading them into the SpawnRequest fields that already
existed but were never reachable from the MCP surface.
A resumeSessionId now requires an explicit profile (a resumed conversation
is tied to the specific backend that started it, so an unqualified spawn
routed by placement has no safe candidate to check) and is refused, naming
Capability.SESSION_RESUME, when that profile's adapter does not declare it
— PeerLauncher gains capabilitiesFor(profileName) so a mixed fleet is
checked per-adapter rather than against the fleet-wide capability union.
PeerHandle.agentSessionId() loses its default, the same fix CB-571 (c00a86b)
made for charterReceipt() one method above it — both existing adapters
already overrode it, so this only closes the landmine for a future one.
Out of scope, left for follow-up: carrying agentSessionId on a failed
ticket's ReleaseDetail (issue #65 criterion 5) — CB-578 stage C just
landed in that file and this ticket deliberately stayed out of it.
--skip-worktree is an index flag. GitWorktrees.snapshot staged into a fresh
empty temp index, which carried none of the real index's skip-worktree bits,
so add -A staged local on-disk content for files git status correctly hides
(e.g. .mcp.json). Copy the real index (resolved via git rev-parse --git-path
index, correct for linked worktrees) into the temp index before staging, so
add -A skips exactly what git status skips.
The doc claimed an in-flight failure re-surfaces the messages on a later
drain, because the ack is local. That was true when InMemoryReplyInbox was
the only inbox, and became false without anyone noticing when the AMQP
adapter landed: there the ack is a broker-side basicAck, so a crash while
writing the response loses the reply outright. Re-polling cannot recover it,
since the broker has already forgotten it.
Behaviour is unchanged and the window stays accepted — acking on the next
poll instead would double-deliver on every normal drain. The point is that
the comment sat exactly where the next implementer would read it and said
the opposite of what happens.
Closes#12.
Found by running it. The script launched the daemon with a relative jar path
(cwd is bridged/) but detected it with an absolute one, so pgrep never matched.
The daemon restarted correctly and booted clean, and the script still failed
with 'no process appeared' — the worst shape of bug for a deploy tool, because
it invites a second restart on a daemon that is already healthy.
Detection now matches both path forms, and the launch uses the absolute path so
ps names which checkout is running.
The lead already owned the redeploy, but the command classifier refuses a bare
kill on the daemon, so in practice every deploy still needed the operator to
approve a stop and a start by hand. A single script is the seam that fixes
that: the operator allow-lists one auditable command instead of two ad-hoc
ones.
It also stops the procedure from living only in a checklist people read after
things go wrong. It builds before it stops anything, so a failed build never
leaves the fleet down; waits for the old process to exit instead of assuming;
polls /healthz; and anchors its log checks to a line marker taken before the
restart, so old errors cannot be misread as new ones.
The check with no log line anywhere in the daemon is the reason --check exists:
bridged inherits WORKER_GITEA_TOKEN from the shell that starts it, and starting
from a non-login shell empties it. The daemon boots fine, healthz is green, and
the failure only appears later as workers that cannot open a PR. --check tests
whether the name resolves and never prints the value.
A CLAUDE.md rule grants intent, not tool permission. Tried the redeploy right
after writing the section and the classifier refused `kill`, so the section
would have been misleading as written. States the honest split until a Bash
permission rule exists: the lead builds and verifies, the operator runs the
stop and start with `!`.
A merge is not a deployment: the running bridged holds the jar it started
with, so merged code does nothing until the daemon is rebuilt and restarted.
Calling that work shipped is a false report. This makes the redeploy the
lead's job rather than something handed back to the operator, and records the
five things that have gone wrong doing it here — chiefly that starting the
daemon from a non-login shell empties WORKER_GITEA_TOKEN, which nothing logs
and which only surfaces later as workers that cannot open a PR.
Goes in the project addendum, not the canonical block; the block is unchanged
and still byte-identical with the wiki template.
CB-578 stage C commits a preserved dirty worktree to refs/wip/<branch> with
git add -A. The parity overlay copies the primary's environment files into
every worktree, so a non-gitignored overlay path now reaches a durable git
object instead of only sitting on disk. add -A respects .gitignore, which is
what stops it — so the rule documented for CB-581 is now what keeps a secret
out of a commit, not just out of a directory.
Adds Worktrees.snapshot(worktreePath, branch, message): stages into a
temporary GIT_INDEX_FILE (never the worker's real index/HEAD), writes
the tree, commit-trees it onto the worktree's current HEAD, and points
refs/wip/<branch> at the result. add -A (never -f) respects .gitignore.
SessionManager.release() now snapshots any dirty worktree before the
preserve-or-remove decision, regardless of release cause (COMPLETED or
SHUTDOWN) — preserving on disk alone is one `worktree remove --force`
away from gone. A failing snapshot never escalates: the worktree is
still preserved, the pane still stops, and the release listener is
still notified.
onRelease's listener now receives a ReleaseDetail (terminal, worktree
path, branch, snapshot ref) instead of a bare terminal id, so Bridged's
WORKER_FAILED wiring can put the same three facts into a failed
ticket's detail — a lead can re-dispatch onto the same tree instead of
starting from the base commit.
A quarantined profile's capacity row now forces free:0 and names the
quarantine (credentialId, quarantinedForSeconds), reusing the same
QuarantineSource bridge_profiles already reads instead of a second
lookup. An ordinary fleet's capacity rows are unchanged (no new keys).
Closes issue #50. Verified by the lead: own unpiped build of the branch merged
onto main — 740 tests, BUILD SUCCESS, exit 0.
Stage A only classified a usage-limit refusal. Nothing acted on it, so the
fleet would spawn another member onto the same exhausted account and fail the
same way. Now a BACKEND_EXHAUSTED classification quarantines the CREDENTIAL
for a cooldown: an explicit spawn onto it is refused with the profile, the
credential and the seconds remaining, placement skips it under every policy,
and it lifts itself on an injected clock.
Keyed by credential, not profile name, because sol and terra are two models on
one OpenAI account. Quarantining only the profile that reported the refusal
would leave its sibling live, and the next spawn walks onto the same dead
account. A profile that sets no credentialId quarantines alone under its own
name, so an existing config behaves exactly as before.
The implementer also found a real bug outside the brief: ConfigRef's
sameLaunchSettings never compared exhaustedPattern, so a reload changing only
that key reported 'config reloaded' while the value is actually deferred —
the precise failure that file's own doc calls the worst outcome a reload can
produce. Fixed with a regression test.
Two things I changed at review. The example.yaml conflict with e09cac6 was
resolved toward the branch, which was a strict superset. And
BackendQuarantine.none() held a clock frozen at 0, so a quarantine() call on
it would have locked a credential out for the life of the daemon — two
production CompositePeerLauncher constructors default to none(). It is now a
real no-op with a test.
Left open deliberately: the implementer's bridge_ask about the credentialId
name timed out after 55s with no answer, because I was polling minutes apart.
It proceeded on its own judgment and chose well. The 55s ask window against an
async lead is a bridge problem, not a worker problem.
Added at merge review. none() held a clock frozen at 0 with a 1ns cooldown, so
a quarantine() call on it recorded a deadline that could never pass — the
credential would be locked out for the life of the daemon. Two production
CompositePeerLauncher constructors default to none(), so that failure would
have been silent and permanent.
The implementer documented the limitation honestly rather than hiding it, but
a stand-in named none() should not need the caveat. quarantine() is now a
no-op on that instance, with a test asserting it. An inert value must omit the
fact, never invent one.
A BACKEND_EXHAUSTED classification (stage A) now puts that profile's
credential into a BackendQuarantine for a configurable cooldown. A spawn
onto a quarantined profile is refused naming the credential and roughly
when it lifts; weighted/round-robin/fixed placement skip a quarantined
candidate; the quarantine lifts itself on the injected clock; and it is
visible on bridge_profiles.
Keyed by credential, not by profile name, via the new Profile.credentialId
(profiles sharing one credential quarantine together — e.g. two models on
one account) and effectiveCredentialId() (unset ⇒ quarantines alone,
today's behaviour unchanged). Fixed a related gap along the way: a reload
changing exhaustedPattern was silently reported "applied" even though it's
deferred — sameLaunchSettings() now catches it too.
Issue #57 criterion 5 asked for an explicit decision on this rather than
silence. It is closer to live than the issue assumed.
The DEFAULT parityOverlay is List.of(".env", ".envrc"), so it applies to every
profile, and neither path was gitignored. Since CB-576 a release preserves any
worktree that git status --porcelain calls dirty, and that deliberately counts
untracked files — the work lost in CB-576 was a file nobody had added. So one
.env at the repo root would make every COMPLETED release preserve its
worktree, and worktrees would accumulate with no error to notice.
Inert today only because neither file exists here. Both are now gitignored,
which they deserve on their own as environment files. The javadoc carries the
rule for the next overlay path: it must be gitignored, or tracked and
skip-worktree'd.
Verified by the lead: own build of the branch merged onto main — 717 tests,
BUILD SUCCESS, exit 0.
release() ran an unprotected sequence. hasUncommitted shells out to git status
and throws on a non-zero exit, which skipped both notifyReleased and
launcher.stop. The session was already out of the registry, so nothing retried
it: a live pane kept burning a fleet slot while absent from the roster, and any
send blocked on it was never resolved. reapIdle called release bare inside a
loop, so one such session aborted the whole pass and skipped every session
after it.
Now the dirty-check block is guarded and fails toward preserving the worktree —
'we could not tell' must not be treated as 'it is clean', because deleting on a
guess destroys work with no other copy. notifyReleased runs in a finally and
launcher.stop runs unconditionally, so the pane always stops. reapIdle catches
per session, matching the shape drainAll already used.
Checked while reviewing: notifyReleased cannot throw out of the finally — it
already guards each listener and only logs. The pane stop is genuinely
unconditional.
hasUncommitted shells out to git and can throw; release() now catches that
inside a try/finally so notifyReleased and launcher.stop always run, and
defaults to preserving the worktree on a throw (can't tell dirty vs clean,
so don't risk deleting unrecoverable work). reapIdle wraps each per-session
release in try/catch, matching drainAll, so one bad session no longer
skips the rest of the reaping pass.
Unit 2 was written as one block, but CB-571/576/580 and unit 2a have since
built parts of it. A symbol survey of main source plus the merge history puts
it at 1 of 21 criteria done, with three partials.
Criterion 15 says a normal COMPLETED release removes the worktree. Since
CB-576 that is false on purpose — a dirty worktree is preserved, because
deleting it destroys work that cannot be recovered. The criterion is wrong,
not the code.
The patterns are compiled once in Bridged.main from the startup config
snapshot, so adding one to a profile does nothing until a restart. Neither
the per-key docs nor the reload-class table said so.
Verified by the lead: own build of this branch merged onto main — 710 tests,
BUILD SUCCESS, exit 0.
Supersedes PR #54. A reviewer found that OpenCodeLauncher.SessionAwareHandle
wrapped the base WorkerHandle but never overrode charterReceipt(), so it
inherited the interface default of null while the real receipt sat on its
delegate — sol and terra would have shown no charterSource/charterSha256 on
the roster while Claude Code members showed both.
The fix is the root one, not the one-line override: PeerHandle.charterReceipt()
is no longer a default, so the compiler forces every implementation to answer.
This repo had shipped that same class of defect — a defaulted dependency that
compiles, passes tests, and quietly turns a feature off — eight times before
this one.
ltms
Dai Ha