Compare commits

...

11 Commits

Author SHA1 Message Date
Dai Ha b5bc5d4ab5 Merge PR #701: fleetd #669 Unit D — the resolver, a live spawned member outranks every tab map
CI / shell-tests (push) Failing after 8s
CI / contract (push) Successful in 49s
CI / build (push) Failing after 1m55s
2026-10-04 01:43:17 +02:00
Dai Ha d83972ede2 fleetd #669 Unit D correction: confirm live slot role before granting a spawned architect
CI / shell-tests (pull_request) Failing after 9s
CI / contract (pull_request) Successful in 50s
CI / build (pull_request) Failing after 1m44s
The new spawned-member roster step returned Principal.architect from the roster's
own MemberRole.ARCHITECT alone, never reconfirming against memberSlotRoles. A slot
revoked after the bind kept granting ARCHITECT to the already-bound session,
regressing fleetd #424's "config governs what a bound slot still grants" half. The
roster now only answers that the pane is a live spawned member; a confirmed live
slot role still decides whether that grants ARCHITECT, falling through to WORKER
otherwise — mirroring the existing architect-slot step a few lines below.

Also corrects LeadTabScanner.buildTabIndex's javadoc: a colliding lead/collaborator
key resolves to the lead because the lead entry is put last (overwriting), not
because it is put first.
2026-10-04 01:37:13 +02:00
Dai Ha 02c6909546 fleetd #669 Unit D: a live spawned member outranks every tab map
CI / shell-tests (pull_request) Failing after 10s
CI / contract (pull_request) Successful in 56s
CI / build (pull_request) Failing after 1m43s
CallerResolver.resolve() now checks the spawned-member roster first, ahead
of every tab map, so a live member's own role wins over a lead or
collaborator tab naming the same terminal (closes #661 at the resolver
level). Adds collaborator resolution (Role.COLLABORATOR) and a real
knownLeadOrCollaborator() classifier, wired into both FleetMcp.denyFor and
FleetApp.allow in place of the inert NO_KNOWN_LEAD_OR_COLLABORATOR stand-in.

LeadTabScanner is generalized to match lead and collaborator tabs in one
pass, keeping every existing liveness/caching/grace-scan property for both
kinds. FleetdAssembly wires the collaborator tab map and a roster-backed
spawned-member lookup; a collaborator-only fleet (no leaders configured)
falls back to a 10s scan interval, matching FleetConfig.Leader's own default.
2026-10-04 01:13:25 +02:00
Dai Ha b92a669ddc Merge PR #699: fleetd #669 Unit C — the COLLABORATOR role and its authorization row
CI / shell-tests (push) Failing after 11s
CI / contract (push) Successful in 52s
CI / build (push) Failing after 1m49s
2026-10-04 00:34:13 +02:00
Dai Ha bb29b001e4 fleetd #669 Unit C: add the COLLABORATOR role and its authorization row
CI / shell-tests (pull_request) Failing after 7s
CI / contract (pull_request) Successful in 53s
CI / build (pull_request) Failing after 1m48s
Adds Role.COLLABORATOR, Principal.collaborator(), and the Authz.permits grant:
READ/METRICS open, REPLY/ASK via ownership, SEND limited to a configured lead
or collaborator via a new classifier parameter, everything else denied.
isSpawnedMember() stays WORKER || ARCHITECT. fleet_whoami reports role and
collaborator name with no leader key.

Per a same-day ticket correction, the existing 3-argument Authz.permits is
kept (fail-closed default via the new NO_KNOWN_LEAD_OR_COLLABORATOR
classifier) rather than deleted, so the 47 pre-existing call sites in
AuthzTest/CallerResolverTest are untouched; both production gates
(FleetMcp#denyFor, FleetApp#allow via the new permitsFor seam) call the
4-argument form explicitly with the shared constant.

Mutation-verified: removing the classifier conjunct from the SEND arm kills
exactly 3 tests (AuthzTest, FleetMcpAuthzTest, FleetAppAuthTest), one per
gate, no cascade.

Full mvn clean install at this commit: 1974 tests, 0 failures (baseline at
f0ff252 was 1964; +10 are the new collaborator-matrix tests). Independently
counted from target/surefire-reports/*.xml after rm -rf: 173 report files,
aggregate tests=1974 failures=0 errors=0.
2026-10-04 00:22:27 +02:00
Dai Ha f0ff25221e Merge PR #697: fleetd #669 Unit B — recognise-only fleet.collaborators config block
CI / shell-tests (push) Failing after 14s
CI / contract (push) Successful in 55s
CI / build (push) Failing after 2m2s
2026-10-04 00:02:10 +02:00
Dai Ha 780cb342ad fleetd #669: pin the fleet-wide tabLabel-vs-collaborator-tab check
CI / shell-tests (pull_request) Failing after 8s
CI / contract (pull_request) Successful in 46s
CI / build (pull_request) Failing after 1m58s
Adds a test driving validateLeadTabPrefixes()'s fleet-wide
fleet.tabLabel branch directly against a collaborator tab, with no
profile override involved. Every existing collaborator test exercised
only the profile-override branch below it, leaving the fleet-wide
branch without a pinning test.
2026-10-03 23:55:47 +02:00
Dai Ha 5c563f02c8 fleetd #669: comments-only fixes — drop history/narration and premature behaviour claims
CI / shell-tests (pull_request) Failing after 6s
CI / contract (pull_request) Successful in 53s
CI / build (pull_request) Failing after 1m54s
The duplicate-collaborator test's javadoc narrated the before/after and a
timestamp; replaced with the present-tense behaviour it protects, per the
project's code-comment rule.

The Collaborator record javadoc and fleetd.example.yaml both claimed the
daemon "learns to address" a collaborator tab. Nothing does yet — that is
Units C-E. Dropped the claim from the record's contract and the example
now says the block is parsed and validated today, nothing more.

Also drops a confidence marker and a cross-class pointer from the
Collaborator record javadoc. No logic, assertion, or file outside this
PR's existing three changed.
2026-10-03 23:47:16 +02:00
Dai Ha ad593c9bb9 fleetd #669: add collaborators to the duplicate-slot-key guard
CI / shell-tests (pull_request) Failing after 8s
CI / contract (pull_request) Successful in 52s
CI / build (pull_request) Failing after 2m3s
FLEET_POOL_KEYS was missing "collaborators", so a duplicated
fleet.collaborators.<name> key took the skipValue branch in
rejectDuplicateSlotsInPools and silently collapsed last-wins, unlike
the other five pools. Measured: before this change, a test loading a
config with a duplicated collaborator name observed no exception;
after, it fails exactly like the existing architect-pool case.

Also updates the javadoc's "five pools" count to six.
2026-10-03 23:41:13 +02:00
Dai Ha 736fd9cf4b Merge PR #696: fleetd #692 — bound the unbounded userinfo mask in redact()
CI / shell-tests (push) Failing after 10s
CI / contract (push) Successful in 49s
CI / build (push) Failing after 1m46s
2026-10-03 23:33:37 +02:00
Dai Ha 05244a82b3 fleetd #669 Unit B: recognise-only fleet.collaborators config block
CI / shell-tests (pull_request) Failing after 8s
CI / contract (pull_request) Successful in 58s
CI / build (pull_request) Failing after 1m46s
Adds fleet.collaborators.<name>.tab (a Collaborator record with only a tab
field — no profile, no instances, no kind; never auto-launched) and widens
the two startup refusals that stop it being a privilege hole:

- validatePanePlacementAgainstLeadTabs() now fires on either a lead tab or
  a collaborator tab, and no longer early-returns when fleet.leaders is
  empty but fleet.collaborators is not.
- validateLeadTabPrefixes() now also refuses a tabLabel template able to
  render as a collaborator tab, two collaborators sharing one exact tab,
  and a collaborator tab equal to a lead tab (case-insensitively).

validateMembers() refuses a collaborator with no (or blank) tab, next to
the existing leader-must-name-a-tab check.

Does not touch Authz, CallerResolver, LeadTabScanner, MemberRole or
FleetMcp — those are later units per the ticket's plan.
2026-10-03 23:33:28 +02:00
17 changed files with 1287 additions and 122 deletions
+9
View File
@@ -669,6 +669,15 @@ fleet:
# kind: opencode
# model: openai/gpt-5.6-terra
# A collaborator tab, keyed by name (fleetd #669). This block is parsed and validated today;
# nothing yet recognises or addresses the tab it names. Recognise-only, like a profile-less
# `leaders:` entry above: there is no `profile:`, no `instances:` and no `kind:`. `tab:` is
# REQUIRED and is the only field identity depends on, matched case-insensitively — the same
# GET-THE-VALUE-RIGHT warning above the `leaders:` block applies here too.
# collaborators:
# reviewer-alex:
# tab: "collab: alex"
# architects:
# architect-1:
# profile: opus # a strong model, on the operator's subscription
@@ -43,6 +43,7 @@ import dev.ltms.fleet.msg.MessageService;
import dev.ltms.fleet.msg.Rendezvous;
import dev.ltms.fleet.msg.ReplyInbox;
import dev.ltms.fleet.msg.ReplyPushLoop;
import dev.ltms.fleet.peer.MemberRole;
import dev.ltms.fleet.peer.PeerLauncher;
import dev.ltms.fleet.placement.BackendOutagePolicy;
import dev.ltms.fleet.placement.BackendQuarantine;
@@ -50,6 +51,7 @@ import dev.ltms.fleet.power.CaffeinateSleepAssertionMechanism;
import dev.ltms.fleet.power.IdleSleepGuard;
import dev.ltms.fleet.rest.FleetApp;
import dev.ltms.fleet.session.GitWorktrees;
import dev.ltms.fleet.session.MemberSession;
import dev.ltms.fleet.session.SessionManager;
import dev.ltms.fleet.session.SessionReaper;
import io.javalin.Javalin;
@@ -250,24 +252,44 @@ final class FleetdAssembly {
log.info("leads: {} panes recognised {}", leadTerminals.size(), leadTerminals.values());
}
// CB-531/CB-579: discover leads by the tab labels the operator writes, one scanner per
// configured lead's own exact `tab:` label.
// configured lead's own exact `tab:` label. fleetd #669: the same scan also recognises a
// configured collaborator's tab, so one herdr pass answers both.
final Supplier<Map<String, String>> leads;
final Supplier<Map<String, String>> collaboratorTerminals;
var leaders = cfg.fleet().leaders();
if (!leaders.isEmpty()) {
var collaboratorsConfig = cfg.fleet().collaborators();
if (!leaders.isEmpty() || !collaboratorsConfig.isEmpty()) {
Map<String, String> tabToName = new LinkedHashMap<>();
leaders.forEach((name, leader) -> {
if (leader != null && leader.tab() != null && !leader.tab().isBlank()) {
tabToName.put(leader.tab(), name);
}
});
int scanIntervalSeconds = leaders.values().iterator().next().scanIntervalSeconds();
Map<String, String> collaboratorTabToName = new LinkedHashMap<>();
collaboratorsConfig.forEach((name, collaborator) -> {
if (collaborator != null && collaborator.tab() != null && !collaborator.tab().isBlank()) {
collaboratorTabToName.put(collaborator.tab(), name);
}
});
// A collaborator-only fleet configures no `leaders:` entry to read a scan interval from
// — FleetConfig.Collaborator carries no scanIntervalSeconds of its own. Falling back to
// FleetConfig.Leader's own compact-constructor default keeps a collaborator-only
// deployment on the same rescan cadence as the default lead cadence, instead of
// inventing a second number for the same kind of scan.
int scanIntervalSeconds = leaders.isEmpty()
? 10
: leaders.values().iterator().next().scanIntervalSeconds();
// This must use the lead daemon: scanning member tabs would demote the lead to a worker.
leads = new LeadTabScanner(herdr, tabToName, Set.of(),
LeadTabScanner scanner = new LeadTabScanner(herdr, tabToName, collaboratorTabToName, Set.of(),
TimeUnit.SECONDS.toNanos(scanIntervalSeconds), ports.nanoClock());
log.info("lead scan: tabs {} host a lead (rescan every {}s, shared fleet space)",
tabToName.keySet(), scanIntervalSeconds);
leads = scanner;
collaboratorTerminals = scanner::collaborators;
log.info("lead/collaborator scan: tabs {} host a lead, tabs {} host a collaborator "
+ "(rescan every {}s, shared fleet space)",
tabToName.keySet(), collaboratorTabToName.keySet(), scanIntervalSeconds);
} else {
leads = () -> leadTerminals;
collaboratorTerminals = Map::of;
}
leadsRef.set(leads);
@@ -452,6 +474,15 @@ final class FleetdAssembly {
ConnectionIdentity identity = new ConnectionIdentity(
new PaneLocator(herdr, memberHerdr), new LsofPeerPidLookup(), new LsofProcessCwdLookup());
// fleetd #669 Unit D: a live spawned member resolves as its own role, whatever a tab map
// says about the same terminal — read from the roster meant for a hot path (SessionManager
// javadoc), never rosterResolved(), since resolve() runs on every request.
Function<String, MemberRole> spawnedMemberRole = terminal -> sessions.roster().stream()
.filter(s -> terminal.equals(s.terminalId()))
.map(MemberSession::role)
.findFirst()
.orElse(null);
// CB-501: one resolver behind both entry paths. Worker identity still comes from the
// connection and is never token-gated, so enabling token mode cannot lock the fleet out.
final CallerResolver callers;
@@ -461,11 +492,13 @@ final class FleetdAssembly {
throw new IllegalStateException("auth.mode=token but env var " + cfg.auth().tokenEnv()
+ " is unset or empty — export it before starting fleetd");
}
callers = CallerResolver.withLeadsAndMembers(identity, true, token, leads, members);
callers = CallerResolver.withLeadsAndMembers(identity, true, token, leads, members,
spawnedMemberRole, collaboratorTerminals);
log.info("auth: token mode (bearer required for non-worker callers, env {})",
cfg.auth().tokenEnv());
} else {
callers = CallerResolver.withLeadsAndMembers(identity, false, null, leads, members);
callers = CallerResolver.withLeadsAndMembers(identity, false, null, leads, members,
spawnedMemberRole, collaboratorTerminals);
log.info("auth: loopback-trust (any loopback non-worker caller is the primary)");
}
@@ -1,5 +1,7 @@
package dev.ltms.fleet.auth;
import java.util.function.Predicate;
/**
* The authorization table (CB-505), stated once and enforced on both entry paths.
*
@@ -60,58 +62,92 @@ public final class Authz {
}
/**
* Whether {@code caller} may perform {@code action} against {@code targetSession}.
*
* @param targetSession the session id in the request path; only consulted for the worker-scoped
* actions ({@code REPLY}, {@code ASK}), ignored otherwise, may be
* {@code null}
* The fail-closed classifier: answers no for every target, so a collaborator's {@code SEND}
* is refused unless a caller supplies a real one. {@code CallerResolver#knownLeadOrCollaborator()}
* is the real one, read from the same lead and collaborator maps {@code CallerResolver#resolve}
* consults, so a target that classifier calls known is one {@code resolve} would actually
* resolve as a lead or collaborator.
*/
public static final Predicate<String> NO_KNOWN_LEAD_OR_COLLABORATOR = target -> false;
/**
* Convenience form for a caller with no classifier to supply. Fails closed: a collaborator's
* {@code SEND} is refused, as if no terminal were a configured lead or collaborator — the
* same decision {@link #NO_KNOWN_LEAD_OR_COLLABORATOR} gives explicitly. Every other action's
* result is identical to the four-argument form's, since none of them consult the classifier.
*/
public static boolean permits(Principal caller, Action action, String targetSession) {
return permits(caller, action, targetSession, NO_KNOWN_LEAD_OR_COLLABORATOR);
}
/**
* Whether {@code caller} may perform {@code action} against {@code targetSession}.
*
* @param targetSession the session id in the request path; only consulted for the
* worker-scoped actions ({@code REPLY}, {@code ASK}) and for a
* collaborator's {@code SEND}, ignored otherwise, may be
* {@code null}
* @param knownLeadOrCollaborator whether a terminal is a configured lead or collaborator —
* consulted only for a collaborator's {@code SEND}, to confine
* it to another named peer and never a spawned member's
* terminal
*/
public static boolean permits(Principal caller, Action action, String targetSession,
Predicate<String> knownLeadOrCollaborator) {
if (caller == null || caller.isAnonymous()) {
return false; // authenticated as nothing ⇒ authorized for nothing
}
return switch (action) {
// Fleet lifecycle is the primary's alone — spawn, stop, drain. An architect
// deliberately does NOT get these (CB-548), so it cannot tear down or stand up workers
// even though it coordinates them; and a worker driving any of these would be a worker
// escalating into the orchestrator role.
// Fleet lifecycle is the primary's alone — spawn, stop, drain. An architect and a
// collaborator deliberately do NOT get these, so neither can tear down or stand up
// workers even though one of them coordinates them; and a worker driving any of these
// would be a worker escalating into the orchestrator role.
case SPAWN, STOP, DRAIN, HANDOVER -> caller.isPrimary();
// Delivering a turn to a local session is open to the primary and the architect: an
// architect delegates to workers (that is the role's point) but still has no lifecycle
// rights. A worker is excluded — sending would be it escalating.
case SEND -> caller.isPrimary() || caller.isArchitect();
// Delivering a turn to a local session is open to the primary, the architect, and a
// collaborator whose target is itself a configured lead or collaborator: the architect
// delegates to workers (that is the role's point); a collaborator may reach only
// another named peer, never a spawned member's terminal. A worker is excluded —
// sending would be it escalating.
case SEND -> caller.isPrimary() || caller.isArchitect()
|| (caller.isCollaborator() && knownLeadOrCollaborator.test(targetSession));
// Same grant as SEND. Resolving a worker's blocked question is part of delegating to
// it, not a separate capability.
// Resolving a worker's blocked question is part of delegating to it, open to the same
// two roles that may stand up that delegation in the first place. Not a collaborator:
// resuming another session's turn is lifecycle-adjacent, not peer messaging.
case ANSWER -> caller.isPrimary() || caller.isArchitect();
// Same grant as SEND. This leaves the daemon over the coordination broker rather than
// addressing a local session, but the caller who may do one may do the other.
// Leaves the daemon over the coordination broker rather than addressing a local
// session, open to the same two roles as ANSWER. Not a collaborator: it is a
// local-tab peer with no cross-host route.
case COORD_SEND -> caller.isPrimary() || caller.isArchitect();
// The load-bearing rule: a caller acts only as the pane it occupies. CB-532 widened who
// that can be — a lead answering another lead is replying for its OWN terminal, which
// this already permits — while the rule itself is unchanged, and is what stops anyone
// forging a reply for a rendezvous someone else is waiting on. An architect's own pane
// passes through the same check, so it can answer a funnel that delegated to it. An
// unnamed primary (token/loopback, no pane) owns nothing and is still excluded.
// forging a reply for a rendezvous someone else is waiting on. An architect's or a
// collaborator's own pane passes through the same check, so each can answer a funnel
// that delegated to it. An unnamed primary (token/loopback, no pane) owns nothing and
// is still excluded.
case REPLY, ASK -> caller.ownsSession(targetSession);
// READ is roster, profile, and identity observation — fleet_list, fleet_profiles, and
// fleet_whoami — and carries no secrets: no ticket reply, no pending question, and no
// other session's turn state. Those live under TASK_READ. METRICS is the separate
// Prometheus scrape. Both stay open to every authenticated role.
case READ, METRICS -> caller.isPrimary() || caller.isWorker() || caller.isArchitect();
// Prometheus scrape. Both are open to every authenticated role, including a
// collaborator.
case READ, METRICS -> caller.isPrimary() || caller.isWorker() || caller.isArchitect()
|| caller.isCollaborator();
// Ticket polling and session status, open to every authenticated role the same as READ.
// Unlike READ, a holder may poll a ticket it did not create, or read another session's
// pending question and the turnId that answers it.
// Ticket polling and session status, open to every role READ is open to except a
// collaborator: ticket ids are a sequential counter with no owner check, so a holder
// could walk every ticket and read another session's delegation reply.
case TASK_READ -> caller.isPrimary() || caller.isWorker() || caller.isArchitect();
// fleetd #421: reading held lead-to-lead mail is the primary's alone. An architect
// holds READ today (CB-548), so "not primary" must mean not-architect here too — this
// is coordination between leads, not observation of the roster.
// is coordination between leads, not observation of the roster. The same reasoning
// excludes a collaborator.
case COORD_READ -> caller.isPrimary();
};
}
@@ -7,6 +7,7 @@ import java.nio.charset.StandardCharsets;
import java.security.MessageDigest;
import java.util.Map;
import java.util.function.Function;
import java.util.function.Predicate;
import java.util.function.Supplier;
/**
@@ -19,6 +20,13 @@ import java.util.function.Supplier;
*
* <p><strong>Resolution order</strong> — connection identity first, token second, nothing third:
* <ol>
* <li>A loopback peer PID that maps to a pane this gateway itself spawned ⇒ that member's own
* role: {@link Role#WORKER} for a dev, hunter, or reviewer; {@link Role#ARCHITECT} for an
* architect, but only while the live slot role still confirms it (fleetd #424 — a slot
* revoked from config demotes an already-bound session on its very next request, so the
* roster's own role is never granted on its word alone). No tab map is consulted — a live
* spawned member's identity comes from the registry that spawned it, never from a label a
* pane could also carry.</li>
* <li>A loopback peer PID that maps to a pane named by {@code leaders:}, by the legacy
* {@code primary.terminal} pin, or by an operator-labelled lead tab (CB-307, CB-530, CB-531)
* ⇒ {@link Role#PRIMARY}, carrying that lead's
@@ -28,8 +36,10 @@ import java.util.function.Supplier;
* so two leads can work as peers rather than one being demoted.</li>
* <li>A loopback peer PID that maps to a pane bound to a CB-548 architect slot ⇒
* {@link Role#ARCHITECT}, carrying the slot name. Just unforgeable as a worker's, and
* resolved from the <em>live</em> terminal→slot binding (never a request argument), before
* the generic worker fallback.</li>
* resolved from the <em>live</em> terminal→slot binding (never a request argument). This is
* the case the previous step does not catch: a binding with no live spawned-member session.</li>
* <li>A loopback peer PID that maps to an operator-labelled collaborator tab ⇒
* {@link Role#COLLABORATOR}, carrying that collaborator's name.</li>
* <li>A loopback peer PID that maps to any other herdr pane ⇒ {@link Role#WORKER}. This is
* unforgeable (the OS reports the PID, herdr owns the PID→pane map) and is honoured
* regardless of auth mode, so enabling auth never breaks the fleet.</li>
@@ -64,6 +74,20 @@ public final class CallerResolver {
private final Supplier<Map<String, String>> architectTerminals;
private final Function<String, MemberRole> memberSlotRoles;
private final Function<String, String> memberSlotNames;
/**
* terminal_id → the role of the live spawned member occupying it, or {@code null} for a
* terminal no spawned member occupies. Consulted first, ahead of every tab map: a live
* spawned member's identity is its own, whatever a tab map says about the same terminal.
* A function rather than the roster itself, so a resolve on the hot path never scans a list —
* the lookup strategy is the caller's to choose.
*/
private final Function<String, MemberRole> spawnedMemberRole;
/**
* terminal_id → collaborator name; empty when none are configured. A supplier for the same
* reason as {@link #leadTerminals}: a collaborator tab recognised after construction (the tab
* scan discovering a newly-labelled tab) takes effect without a restart.
*/
private final Supplier<Map<String, String>> collaboratorTerminals;
/** Loopback-trust resolver: no token required, historical behaviour. Test-only. */
CallerResolver(ConnectionIdentity identity) {
@@ -124,17 +148,42 @@ public final class CallerResolver {
/**
* Live registry form that can confirm a bound slot is an architect slot.
*
* <p>This is the only public construction path. It keeps terminal bindings and slot roles in
* the same {@link MemberRegistry}, so a configured architect can resolve as an architect.
* <p>It keeps terminal bindings and slot roles in the same {@link MemberRegistry}, so a
* configured architect can resolve as an architect. No spawned-member roster or collaborator
* registry is consulted — equivalent to {@link #withLeadsAndMembers(ConnectionIdentity,
* boolean, String, Supplier, MemberRegistry, Function, Supplier)} with both absent. Kept for
* every caller that has neither to offer, so adding them did not churn every construction site.
*/
public static CallerResolver withLeadsAndMembers(ConnectionIdentity identity,
boolean tokenMode, String token,
Supplier<Map<String, String>> leadTerminals,
MemberRegistry members) {
return withLeadsAndMembers(identity, tokenMode, token, leadTerminals, members, null, null);
}
/**
* Live registry form that also resolves a live spawned member to its own role, and a
* configured collaborator tab to {@link Role#COLLABORATOR}.
*
* <p>This is the only public construction path that exercises the full resolution order.
*
* @param spawnedMemberRole terminal_id → the role of the live spawned member occupying
* it, or {@code null} for a terminal no spawned member occupies.
* {@code null} here means no roster is consulted at all (every
* terminal falls through to the tab maps), not that none matches.
* @param collaboratorTerminals terminal_id → collaborator name, live like {@code leadTerminals}
*/
public static CallerResolver withLeadsAndMembers(ConnectionIdentity identity,
boolean tokenMode, String token,
Supplier<Map<String, String>> leadTerminals,
MemberRegistry members,
Function<String, MemberRole> spawnedMemberRole,
Supplier<Map<String, String>> collaboratorTerminals) {
return new CallerResolver(identity, tokenMode, token, leadTerminals,
members == null ? null : members::snapshot,
members == null ? null : members::roleForSlot,
members == null ? null : members::nameForSlot);
members == null ? null : members::nameForSlot,
spawnedMemberRole, collaboratorTerminals);
}
private static Supplier<Map<String, String>> fixed(Map<String, String> leadTerminals) {
@@ -160,6 +209,17 @@ public final class CallerResolver {
Supplier<Map<String, String>> architectTerminals,
Function<String, MemberRole> memberSlotRoles,
Function<String, String> memberSlotNames) {
this(identity, tokenMode, token, leadTerminals, architectTerminals, memberSlotRoles,
memberSlotNames, null, null);
}
private CallerResolver(ConnectionIdentity identity, boolean tokenMode, String token,
Supplier<Map<String, String>> leadTerminals,
Supplier<Map<String, String>> architectTerminals,
Function<String, MemberRole> memberSlotRoles,
Function<String, String> memberSlotNames,
Function<String, MemberRole> spawnedMemberRole,
Supplier<Map<String, String>> collaboratorTerminals) {
if (tokenMode && (token == null || token.isBlank())) {
throw new IllegalArgumentException(
"auth.mode=token requires a non-empty token; check that the env var named by "
@@ -172,6 +232,8 @@ public final class CallerResolver {
this.architectTerminals = architectTerminals == null ? Map::of : architectTerminals;
this.memberSlotRoles = memberSlotRoles == null ? _ -> null : memberSlotRoles;
this.memberSlotNames = memberSlotNames == null ? Function.identity() : memberSlotNames;
this.spawnedMemberRole = spawnedMemberRole == null ? _ -> null : spawnedMemberRole;
this.collaboratorTerminals = collaboratorTerminals == null ? Map::of : collaboratorTerminals;
}
/**
@@ -198,6 +260,27 @@ public final class CallerResolver {
return architectTerminals.get();
}
/**
* The currently-recognised collaborator tabs, {@code terminal_id → name}.
*
* <p>Read from the same supplier {@link #resolve} consults, for the reason given in
* {@link #leads()}. Live for the same reason as {@link #leads()}.
*/
public Map<String, String> collaborators() {
return collaboratorTerminals.get();
}
/**
* Whether {@code target} names a terminal this resolver would resolve as a lead or a
* collaborator — the classifier a collaborator's {@code SEND} is checked against, read from the
* exact maps {@link #resolve} consults so a target that would resolve as a lead or collaborator
* is never the one a collaborator is refused to reach, or the reverse.
*/
public Predicate<String> knownLeadOrCollaborator() {
return target -> leadTerminals.get().containsKey(target)
|| collaboratorTerminals.get().containsKey(target);
}
/**
* Resolve the caller of a request.
*
@@ -208,6 +291,25 @@ public final class CallerResolver {
public Principal resolve(String remoteAddr, int remotePort, String authorizationHeader) {
ConnectionIdentity.Caller c = identity.resolve(remoteAddr, remotePort);
if (c.terminal() != null) {
MemberRole spawnedRole = spawnedMemberRole.apply(c.terminal());
if (spawnedRole != null) {
// A live spawned member occupies this pane. Its identity is its own, whatever a tab
// map says about the same terminal — checked before every tab map, consulting none
// of them, so a tab label can never override a roster entry for the same terminal.
if (spawnedRole == MemberRole.ARCHITECT) {
// The roster only answers THAT this pane is a live spawned member; config still
// decides WHAT that member's slot grants (fleetd #424). A slot revoked after the
// bind must still demote this session on its very next request, so the roster's
// own ARCHITECT role is confirmed against the live slot role, exactly as the
// architect-slot step below confirms a binding with no live member session.
String slot = architectTerminals.get().get(c.terminal());
if (slot != null && memberSlotRoles.apply(slot) == MemberRole.ARCHITECT) {
return Principal.architect(memberSlotNames.apply(slot), c.terminal(), c.pid());
}
return Principal.worker(c.terminal(), c.pid());
}
return Principal.worker(c.terminal(), c.pid());
}
String lead = leadTerminals.get().get(c.terminal());
if (lead != null) {
// The config names this pane as a lead's own. The pane mapping is exactly as
@@ -221,10 +323,17 @@ public final class CallerResolver {
// The config/live binding names this pane as an architect slot's own. Same
// unforgeable pane mapping; the live binding, never a request argument, decides.
// Check the slot role too: this defence in depth prevents a bad lifecycle bind from
// escalating a dev, hunter or reviewer into an architect. Checked before
// the worker fallback.
// escalating a dev, hunter or reviewer into an architect. This is the case the
// spawned-member step above does not catch: a binding with no live member session.
return Principal.architect(memberSlotNames.apply(slot), c.terminal(), c.pid());
}
String collaborator = collaboratorTerminals.get().get(c.terminal());
if (collaborator != null) {
// An operator-labelled collaborator tab, confirmed live by the same scan that
// confirms a lead tab. Checked last among the tab maps so a pane also matching one
// of the above keeps that stronger role.
return Principal.collaborator(collaborator, c.terminal(), c.pid());
}
return Principal.worker(c.terminal(), c.pid()); // unforgeable; never token-gated
}
@@ -11,7 +11,9 @@ package dev.ltms.fleet.auth;
* @param pid the connecting process id, or {@code -1} when not resolvable (audit context)
* @param name for a lead resolved from the CB-530 {@code leaders:} registry, which lead it is;
* for an architect resolved from the CB-548 {@code architects:} registry, which
* slot it occupies; {@code null} for every other caller, including an unnamed primary
* slot it occupies; for a collaborator resolved from the {@code collaborators:}
* registry, which collaborator it is; {@code null} for every other caller,
* including an unnamed primary
*/
public record Principal(Role role, String terminal, long pid, String name) {
@@ -73,6 +75,19 @@ public record Principal(Role role, String terminal, long pid, String name) {
return new Principal(Role.ARCHITECT, terminal, pid, slotName);
}
/**
* A collaborator: a human-opened tab recognised by its exact label in the
* {@code collaborators:} registry.
*
* <p>Carries {@link Role#COLLABORATOR}. {@code name} is reporting only — it lets
* {@code fleet_whoami} say which collaborator is asking. Identity is the {@code terminal}:
* like a worker's it comes from the connection, so {@code ownsSession} works exactly as it
* does for a worker — a collaborator acts as its own pane and no other.
*/
public static Principal collaborator(String name, String terminal, long pid) {
return new Principal(Role.COLLABORATOR, terminal, pid, name);
}
public boolean isPrimary() {
return role == Role.PRIMARY;
}
@@ -81,6 +96,10 @@ public record Principal(Role role, String terminal, long pid, String name) {
return role == Role.ARCHITECT;
}
public boolean isCollaborator() {
return role == Role.COLLABORATOR;
}
public boolean isWorker() {
return role == Role.WORKER;
}
@@ -120,6 +139,7 @@ public record Principal(Role role, String terminal, long pid, String name) {
return switch (role) {
case WORKER -> "worker:" + terminal;
case ARCHITECT -> "architect:" + name;
case COLLABORATOR -> "collaborator:" + name;
case PRIMARY -> name == null ? "primary" : "leader:" + name;
case ANONYMOUS -> "anonymous";
};
@@ -34,6 +34,17 @@ public enum Role {
*/
ARCHITECT,
/**
* A config-declared, human-opened tab recognised by its exact label (the {@code
* fleet.collaborators.<name>.tab} registry). Never spawned — identity comes from the
* connection, never a request argument, exactly like {@link #WORKER} and {@link #ARCHITECT}.
* May {@code SEND} only to a configured lead or collaborator, {@code REPLY}/{@code ASK} only
* as its own pane, and {@code READ}/{@code METRICS}; may not {@code SPAWN}/{@code STOP}/
* {@code DRAIN}/{@code HANDOVER}, poll a ticket ({@code TASK_READ}), or reach the
* coordination broker ({@code COORD_SEND}/{@code COORD_READ}).
*/
COLLABORATOR,
/** Authenticated as nothing. Authorized for nothing but {@code /healthz}. */
ANONYMOUS
}
@@ -1176,6 +1176,25 @@ public record FleetConfig(
}
}
/**
* A tab fleetd recognises as a collaborator, keyed by name (fleetd #669).
*
* <p>Recognise-only: there is no {@code profile}, no {@code instances} and no {@code kind}.
* Nothing here ever launches a pane.
*
* <p>{@code tabPrefix} is absent. Identity is matched on the exact {@code tab} alone.
*
* @param tab the exact tab label hosting this collaborator, matched case-insensitively; the
* only field identity depends on. Required — an entry with no {@code tab} can
* never be discovered.
*/
@JsonIgnoreProperties(ignoreUnknown = true)
public record Collaborator(String tab) {
public Collaborator {
tab = (tab == null || tab.isBlank()) ? null : tab.strip();
}
}
/**
* One entry of a {@code fleet:} role pool — a role paired with the backend it runs on.
*
@@ -1213,15 +1232,18 @@ public record FleetConfig(
* is exactly compatible with that. The pool is also what replaced {@code defaultProfile:} — an
* unqualified spawn names a role, and the role's pool supplies the candidates.
*
* @param leaders panes that orchestrate rather than are orchestrated, keyed by lead name
* @param architects profiles the {@code architect} role may run on
* @param developers profiles the {@code dev} role may run on
* @param hunters profiles the {@code hunter} role may run on
* @param reviewers profiles the {@code reviewer} role may run on
* @param charters optional launch-charter text keyed by singular role wire name
* @param tabLabel template for a member tab's label; {@code {role}}, {@code {profile}},
* {@code {model}} and {@code {n}} (a per role+profile counter) are
* substituted. Default {@link #DEFAULT_TAB_LABEL}
* @param leaders panes that orchestrate rather than are orchestrated, keyed by lead name
* @param architects profiles the {@code architect} role may run on
* @param developers profiles the {@code dev} role may run on
* @param hunters profiles the {@code hunter} role may run on
* @param reviewers profiles the {@code reviewer} role may run on
* @param charters optional launch-charter text keyed by singular role wire name
* @param tabLabel template for a member tab's label; {@code {role}}, {@code {profile}},
* {@code {model}} and {@code {n}} (a per role+profile counter) are
* substituted. Default {@link #DEFAULT_TAB_LABEL}
* @param collaborators tabs fleetd recognises as collaborators (fleetd #669), keyed by name.
* Recognise-only, exactly like a {@code profile}-less {@link Leader}:
* nothing here is ever auto-launched.
*/
@JsonIgnoreProperties(ignoreUnknown = true)
public record Fleet(Map<String, Leader> leaders,
@@ -1230,7 +1252,8 @@ public record FleetConfig(
Map<String, Slot> hunters,
Map<String, Slot> reviewers,
Map<String, String> charters,
String tabLabel) {
String tabLabel,
Map<String, Collaborator> collaborators) {
/**
* Role first, so the tab bar identifies the member's fleet role.
@@ -1245,26 +1268,30 @@ public record FleetConfig(
reviewers = unmodifiableOrEmpty(reviewers);
charters = unmodifiableOrEmpty(charters);
tabLabel = (tabLabel == null || tabLabel.isBlank()) ? DEFAULT_TAB_LABEL : tabLabel;
collaborators = unmodifiableOrEmpty(collaborators);
}
/**
* A fleet with no configured launch charters — the shape every deployment had before
* CB-566, and what most tests want.
* A fleet with no configured launch charters and no collaborators — the shape every
* deployment had before CB-566, and what most tests want.
*
* <p>Kept deliberately, even though an overload that drops a new field is normally the
* shape to avoid. It is safe here because nothing <em>reads</em> a charter through a
* constructor: the launcher reads {@code fleet.charters()} from the live config. Jackson
* binds the canonical constructor, so this one cannot swallow an operator's YAML.
* {@code collaborators} is dropped the same way and for the same reason: no caller of
* this overload has ever needed to set it, so it defaults to empty here exactly as the
* canonical constructor would default an absent YAML key.
*/
public Fleet(Map<String, Leader> leaders, Map<String, Slot> architects,
Map<String, Slot> developers, Map<String, Slot> reviewers,
Map<String, String> charters, String tabLabel) {
this(leaders, architects, developers, null, reviewers, charters, tabLabel);
this(leaders, architects, developers, null, reviewers, charters, tabLabel, null);
}
public Fleet(Map<String, Leader> leaders, Map<String, Slot> architects,
Map<String, Slot> developers, Map<String, Slot> reviewers, String tabLabel) {
this(leaders, architects, developers, null, reviewers, null, tabLabel);
this(leaders, architects, developers, null, reviewers, null, tabLabel, null);
}
/**
@@ -1910,7 +1937,7 @@ public record FleetConfig(
/** The {@code fleet:} child blocks whose direct children are slot names. */
private static final Set<String> FLEET_POOL_KEYS =
Set.of("leaders", "architects", "developers", "hunters", "reviewers");
Set.of("leaders", "architects", "developers", "hunters", "reviewers", "collaborators");
/**
* Reject a {@code fleet:} role pool whose slot names repeat (CB-548, re-homed by CB-557).
@@ -1920,7 +1947,7 @@ public record FleetConfig(
* daemon would never know. Jackson's YAML parser does not fail on duplicate mapping keys by
* default, so duplicates are caught here, at parse time, before the map is built.
*
* <p>Only the five pools <em>directly under the top-level {@code fleet:}</em> are considered,
* <p>Only the six pools <em>directly under the top-level {@code fleet:}</em> are considered,
* and only their direct child keys (the slot names). A nested field elsewhere, even one also
* named {@code developers:}, is ignored, so parsing of the rest of the config is unaffected.
*
@@ -2679,17 +2706,21 @@ public record FleetConfig(
}
/**
* Reject a member tab-label template that could render as a configured lead tab or match a
* lead-tab naming convention, and reject two {@code fleet.leaders} entries that share one exact
* tab.
* Reject a member tab-label template that could render as a configured lead or collaborator
* tab or match a lead-tab naming convention, and reject two {@code fleet.leaders} or
* {@code fleet.collaborators} entries — across either registry — that share one exact tab.
*
* <p>{@code fleet.collaborators} has no {@code tabPrefix}: identity is matched on the exact
* {@code tab} alone, so only the exact-render check applies there, not the prefix check.
*
* @throws IllegalStateException when the fleet template or a profile {@code tabLabel} override
* can render as a configured lead tab or match a lead-tab prefix,
* or when two {@code fleet.leaders} entries carry the same exact
* {@code tab} (case-insensitively)
* can render as a configured lead or collaborator tab or match a
* lead-tab prefix, or when two entries — of either registry, or
* one of each — carry the same exact {@code tab}
* (case-insensitively)
*/
public void validateLeadTabPrefixes() {
if (fleet == null || fleet.leaders().isEmpty()) {
if (fleet == null) {
return;
}
List<String> bad = new ArrayList<>();
@@ -2722,11 +2753,32 @@ public record FleetConfig(
}
});
});
fleet.collaborators().forEach((collabName, collaborator) -> {
if (collaborator == null) {
return;
}
String tab = collaborator.tab();
if (templateCanRenderAs(fleet.tabLabel(), tab)) {
bad.add("fleet.tabLabel=\"" + fleet.tabLabel() + "\" can render as the tab of "
+ "collaborator '" + collabName + "' (\"" + tab + "\")");
}
profiles().entrySet().stream()
.map(Map.Entry::getKey)
.sorted()
.forEach(p -> {
String label = profiles().get(p).tabLabel();
if (templateCanRenderAs(label, tab)) {
bad.add("profile '" + p + "' overrides tabLabel with \"" + label
+ "\", which can render as the tab of collaborator '"
+ collabName + "' (\"" + tab + "\")");
}
});
});
if (!bad.isEmpty()) {
throw new IllegalStateException("refusing to start: " + String.join("; ", bad)
+ ". Every member labelled that way would be read back as a lead and granted "
+ "spawn/stop/send on the whole fleet. Change one of the two so member tabs "
+ "and lead tabs cannot be confused.");
+ ". Every member labelled that way would be read back as a lead or "
+ "collaborator and granted that identity's authority. Change one of the two "
+ "so member tabs cannot be confused with a lead's or collaborator's tab.");
}
List<String> collisions = new ArrayList<>();
@@ -2749,13 +2801,49 @@ public record FleetConfig(
}
}
}
List<String> collabNames = fleet.collaborators().keySet().stream().sorted().toList();
for (int i = 0; i < collabNames.size(); i++) {
String nameA = collabNames.get(i);
Collaborator a = fleet.collaborators().get(nameA);
if (a == null || a.tab() == null || a.tab().isBlank()) {
continue;
}
for (int j = i + 1; j < collabNames.size(); j++) {
String nameB = collabNames.get(j);
Collaborator b = fleet.collaborators().get(nameB);
if (b == null || b.tab() == null || b.tab().isBlank()) {
continue;
}
if (a.tab().equalsIgnoreCase(b.tab())) {
collisions.add("collaborator '" + nameA + "' and collaborator '" + nameB
+ "' both use tab \"" + a.tab() + "\"");
}
}
}
for (String leadName : leadNames) {
Leader lead = fleet.leaders().get(leadName);
if (lead == null || lead.tab() == null || lead.tab().isBlank()) {
continue;
}
for (String collabName : collabNames) {
Collaborator collaborator = fleet.collaborators().get(collabName);
if (collaborator == null || collaborator.tab() == null
|| collaborator.tab().isBlank()) {
continue;
}
if (lead.tab().equalsIgnoreCase(collaborator.tab())) {
collisions.add("lead '" + leadName + "' and collaborator '" + collabName
+ "' both use tab \"" + lead.tab() + "\"");
}
}
}
if (collisions.isEmpty()) {
return;
}
throw new IllegalStateException("refusing to start: " + String.join("; ", collisions)
+ ". Tab identity is matched exactly, so only one of two leads sharing a tab can "
+ "ever be found — the other is silently unreachable. Give each lead its own "
+ "exact tab.");
+ ". Tab identity is matched exactly, so only one of two entries sharing a tab can "
+ "ever be found — the other is silently unreachable. Give each lead and "
+ "collaborator its own exact tab.");
}
private static boolean templateCanRenderAs(String template, String tab) {
@@ -2785,25 +2873,28 @@ public record FleetConfig(
/**
* Reject a profile that places its members by {@code "pane"} while any {@code fleet.leaders}
* entry names a {@code tab}. A pane-placed member lands inside the focused tab rather than its
* own, so it can land inside a lead's own labelled tab. {@link
* dev.ltms.fleet.herdr.LeadTabScanner} identifies a lead purely by that tab's label — it does
* not exclude the member space — so a member that ends up there would be read back as the lead
* and granted spawn/stop/send on the whole fleet.
* or {@code fleet.collaborators} entry names a {@code tab}. A pane-placed member lands inside
* the focused tab rather than its own, so it can land inside a lead's or collaborator's own
* labelled tab. {@link dev.ltms.fleet.herdr.LeadTabScanner} identifies a lead or collaborator
* purely by that tab's label — it does not exclude the member space — so a member that ends up
* there would be read back as that lead or collaborator and granted that identity's authority.
*
* <p>Only a leader with a non-blank {@code tab} is in scope: one with no {@code tab} feeds
* <p>Only an entry with a non-blank {@code tab} is in scope: one with no {@code tab} feeds
* nothing into {@link dev.ltms.fleet.herdr.LeadTabScanner}, so it creates no hazard here.
*
* @throws IllegalStateException when any {@code profiles:} entry is pane-placed while any
* {@code fleet.leaders} entry names a non-blank {@code tab}
* {@code fleet.leaders} or {@code fleet.collaborators} entry
* names a non-blank {@code tab}
*/
public void validatePanePlacementAgainstLeadTabs() {
if (fleet == null || fleet.leaders().isEmpty()) {
if (fleet == null) {
return;
}
boolean anyLeaderHasTab = fleet.leaders().values().stream()
.anyMatch(leader -> leader != null && leader.tab() != null && !leader.tab().isBlank());
if (!anyLeaderHasTab) {
boolean anyCollaboratorHasTab = fleet.collaborators().values().stream()
.anyMatch(c -> c != null && c.tab() != null && !c.tab().isBlank());
if (!anyLeaderHasTab && !anyCollaboratorHasTab) {
return;
}
List<String> bad = new ArrayList<>();
@@ -2816,10 +2907,11 @@ public record FleetConfig(
return;
}
throw new IllegalStateException("refusing to start: profile(s) " + bad
+ " use placement: pane while fleet.leaders names a tab. A pane-placed member can "
+ "land inside a lead's labelled tab and be read back as the lead, granted "
+ "spawn/stop/send on the whole fleet. Set placement: tab for each named profile, "
+ "or remove the tab from every fleet.leaders entry.");
+ " use placement: pane while fleet.leaders or fleet.collaborators names a tab. A "
+ "pane-placed member can land inside that labelled tab and be read back as the "
+ "lead or collaborator, granted that identity's authority. Set placement: tab for "
+ "each named profile, or remove the tab from every fleet.leaders and "
+ "fleet.collaborators entry.");
}
/**
@@ -2926,8 +3018,14 @@ public record FleetConfig(
* so duplicates are unrepresentable by construction once loaded — and {@link #load(Path)}
* already rejects a duplicated slot name at parse time, before the map collapses.
*
* @throws IllegalStateException when a slot names no profile or an unknown one, or when a lead
* can be neither found nor created, naming the offending entry
* <p>Also rejects a {@code fleet.collaborators} entry with no (or a blank) {@code tab}. A
* {@code profile}-less lead is still useful recognise-only — {@code tab} is the only field
* that matters to it either way. A collaborator carries no other field at all, so a blank
* {@code tab} leaves nothing for the entry to mean.
*
* @throws IllegalStateException when a slot names no profile or an unknown one, when a lead
* can be neither found nor created, or when a collaborator names
* no tab, naming the offending entry
*/
public void validateMembers() {
if (fleet == null) {
@@ -2965,6 +3063,16 @@ public record FleetConfig(
+ "auto-launched, labelled) purely by its tab, so every entry must name one.");
}
});
fleet.collaborators().forEach((name, collaborator) -> {
if (collaborator == null) {
return;
}
if (collaborator.tab() == null || collaborator.tab().isBlank()) {
bad.add("fleet.collaborators." + name + " has no tab: — a collaborator is "
+ "recognised purely by its tab, and carries no other field, so every "
+ "entry must name one.");
}
});
if (!bad.isEmpty()) {
throw new IllegalStateException("refusing to start: " + String.join(" ", bad));
}
@@ -96,13 +96,19 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
private static final Logger log = LoggerFactory.getLogger(LeadTabScanner.class);
/** What a matched tab names: a lead or a collaborator. */
private enum Kind { LEAD, COLLABORATOR }
/** One matched tab's name and what it names. */
private record Entry(String name, Kind kind) {}
private final HerdrClient herdr;
private final Map<String, String> tabToName;
private final Map<String, Entry> tabToEntry;
private final Set<String> excludedWorkspaceLabels;
private final long ttlNanos;
private final LongSupplier clock;
private Map<String, String> cached = Map.of();
private Map<String, Entry> cached = Map.of();
private long scannedAtNanos;
private boolean everScanned;
@@ -126,26 +132,53 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
*/
public LeadTabScanner(HerdrClient herdr, Map<String, String> tabToName,
Set<String> excludedWorkspaceLabels, long ttlNanos, LongSupplier clock) {
this(herdr, tabToName, Map.of(), excludedWorkspaceLabels, ttlNanos, clock);
}
/**
* As {@link #LeadTabScanner(HerdrClient, Map, Set, long, LongSupplier)}, additionally scanning
* for configured collaborator tabs in the same pass.
*
* @param collaboratorTabToName every configured collaborator's exact tab label → its name
* ({@code fleet.collaborators.<name>.tab}), matched the same way as
* {@code tabToName}
*/
public LeadTabScanner(HerdrClient herdr, Map<String, String> tabToName,
Map<String, String> collaboratorTabToName,
Set<String> excludedWorkspaceLabels, long ttlNanos, LongSupplier clock) {
this.herdr = herdr;
this.tabToName = normalize(tabToName);
this.tabToEntry = buildTabIndex(tabToName, collaboratorTabToName);
this.excludedWorkspaceLabels = excludedWorkspaceLabels == null
? Set.of() : Set.copyOf(excludedWorkspaceLabels);
this.ttlNanos = ttlNanos;
this.clock = clock;
}
/** Keys stripped and lower-cased once, so every lookup is a plain map hit. */
private static Map<String, String> normalize(Map<String, String> tabToName) {
if (tabToName == null || tabToName.isEmpty()) {
return Map.of();
/**
* Keys stripped and lower-cased once, so every lookup is a plain map hit. Leads and
* collaborators merge into a single index, so {@link #scan()} matches both kinds in one pass
* over the tab list; a label naming both a lead and a collaborator takes the lead entry —
* leads are put last, so a colliding key's lead entry is the one that overwrites — since a lead
* can already do everything a collaborator can. Config validation already refuses a lead and a
* collaborator sharing one exact tab, so this ordering is defence in depth, not the control.
*/
private static Map<String, Entry> buildTabIndex(Map<String, String> tabToName,
Map<String, String> collaboratorTabToName) {
Map<String, Entry> out = new LinkedHashMap<>();
putNormalized(out, collaboratorTabToName, Kind.COLLABORATOR);
putNormalized(out, tabToName, Kind.LEAD);
return Collections.unmodifiableMap(out);
}
private static void putNormalized(Map<String, Entry> out, Map<String, String> tabToName, Kind kind) {
if (tabToName == null) {
return;
}
Map<String, String> out = new LinkedHashMap<>();
tabToName.forEach((tab, name) -> {
if (tab != null && !tab.isBlank() && name != null && !name.isBlank()) {
out.put(tab.strip().toLowerCase(Locale.ROOT), name);
out.put(tab.strip().toLowerCase(Locale.ROOT), new Entry(name, kind));
}
});
return Collections.unmodifiableMap(out);
}
/**
@@ -156,6 +189,29 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
*/
@Override
public synchronized Map<String, String> get() {
return byKind(refresh(), Kind.LEAD);
}
/**
* The current {@code terminal_id → collaborator name} map, sharing the same scan and cache as
* {@link #get()} — both kinds are matched in one pass, so this never costs a second herdr call.
*/
public synchronized Map<String, String> collaborators() {
return byKind(refresh(), Kind.COLLABORATOR);
}
private static Map<String, String> byKind(Map<String, Entry> entries, Kind kind) {
Map<String, String> out = new LinkedHashMap<>();
entries.forEach((terminal, entry) -> {
if (entry.kind() == kind) {
out.put(terminal, entry.name());
}
});
return Collections.unmodifiableMap(out);
}
/** Rescans if the cache has expired, otherwise returns the cached answer. */
private Map<String, Entry> refresh() {
long now = clock.getAsLong();
if (everScanned && now - scannedAtNanos < ttlNanos) {
return cached;
@@ -165,21 +221,21 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
scannedAtNanos = now;
everScanned = true;
try {
Map<String, String> fresh = scan();
Map<String, Entry> fresh = scan();
if (!fresh.equals(cached)) {
log.info("lead panes: {}", fresh);
log.info("lead/collaborator panes: {}", fresh);
}
cached = fresh;
} catch (HerdrException e) {
log.warn("lead-tab scan failed, keeping the {} lead(s) already known: {}",
log.warn("lead-tab scan failed, keeping the {} entr(y/ies) already known: {}",
cached.size(), e.getMessage());
}
return cached;
}
/** One full pass: labelled tabs → live agents in them → those panes' terminals. */
private Map<String, String> scan() {
Map<String, String> nameByTab = new LinkedHashMap<>();
private Map<String, Entry> scan() {
Map<String, Entry> entryByTab = new LinkedHashMap<>();
for (JsonNode w : herdr.call("workspace.list").path("workspaces")) {
Workspace ws = Workspace.from(w);
if (ws.workspaceId() == null || excludedWorkspaceLabels.contains(ws.label())) {
@@ -187,21 +243,22 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
}
for (JsonNode t : herdr.call("tab.list", Map.of("workspace_id", ws.workspaceId())).path("tabs")) {
Tab tab = Tab.from(t);
String name = leadNameOf(tab.label());
if (name != null && tab.tabId() != null) {
nameByTab.put(tab.tabId(), name);
Entry entry = entryOf(tab.label());
if (entry != null && tab.tabId() != null) {
entryByTab.put(tab.tabId(), entry);
}
}
}
if (nameByTab.isEmpty()) {
if (entryByTab.isEmpty()) {
gracedTerminals = Set.of();
return Map.of();
}
// fleetd #359: a labelled tab is only a lead when herdr also reports a running agent in
// it — the same liveness signal LeadLauncher.countLeads trusts for the identical purpose.
// Without this, a tab left behind by a session that has since died reads as live forever.
// fleetd #359: a labelled tab is only a lead (or collaborator) when herdr also reports a
// running agent in it — the same liveness signal LeadLauncher.countLeads trusts for the
// identical purpose. Without this, a tab left behind by a session that has since died reads
// as live forever.
Set<String> tabsWithAgent = new HashSet<>();
for (JsonNode a : herdr.call("agent.list").path("agents")) {
String tabId = a.path("tab_id").asText(null);
@@ -210,18 +267,18 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
}
}
Map<String, String> byTerminal = new LinkedHashMap<>();
Map<String, Entry> byTerminal = new LinkedHashMap<>();
Set<String> stillGraced = new HashSet<>();
// One pane.list for every tab: panes carry tab_id, so the join is local.
for (JsonNode p : herdr.call("pane.list", Map.of()).path("panes")) {
String tabId = p.path("tab_id").asText(null);
String name = nameByTab.get(tabId);
Entry entry = entryByTab.get(tabId);
String terminal = p.path("terminal_id").asText(null);
if (name == null || terminal == null || terminal.isBlank()) {
if (entry == null || terminal == null || terminal.isBlank()) {
continue;
}
if (tabsWithAgent.contains(tabId)) {
byTerminal.put(terminal, name);
byTerminal.put(terminal, entry);
continue;
}
// No agent reported for this tab, but its tab/pane are still here — this is the
@@ -230,7 +287,7 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
// reported as live; a terminal we never reported live gets none, so the original #359
// fix (a genuinely dead tab is never reported) is unaffected for the common case.
if (cached.containsKey(terminal) && !gracedTerminals.contains(terminal)) {
byTerminal.put(terminal, name);
byTerminal.put(terminal, entry);
stillGraced.add(terminal);
}
}
@@ -239,18 +296,19 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
}
/**
* The lead name a tab label declares, or {@code null} if it names none of the configured leads.
* The entry a tab label declares, or {@code null} if it names neither a configured lead nor a
* configured collaborator.
*
* <p>Exact match (case-insensitive, ends stripped) against {@link #tabToName} — no prefix
* <p>Exact match (case-insensitive, ends stripped) against {@link #tabToEntry} — no prefix
* stripping, so an operator's {@code "lead: something-else"} tab is never mistaken for a
* configured lead just because it shares a prefix. The match strips a trailing
* {@link PendingCloseMarker} first, so a tab {@code LeadLauncher} has flagged as maybe-dead but
* not yet closed keeps resolving normally while that reconcile is pending.
*/
private String leadNameOf(String label) {
private Entry entryOf(String label) {
if (label == null) {
return null;
}
return tabToName.get(PendingCloseMarker.strip(label).toLowerCase(Locale.ROOT));
return tabToEntry.get(PendingCloseMarker.strip(label).toLowerCase(Locale.ROOT));
}
}
@@ -114,6 +114,13 @@ public final class FleetMcp {
* dev.ltms.fleet.herdr.PaneLocator} — that a real assembly wired up. See {@link #identity()}.
*/
private final ConnectionIdentity identity;
/**
* Kept as a field (rather than only captured by the {@code contextExtractor} closure) so
* {@link #denyFor} can read {@link CallerResolver#knownLeadOrCollaborator()} — the classifier a
* collaborator's {@code SEND} is checked against, built from the same lead and collaborator
* maps {@link #identity}-based resolution reads.
*/
private final CallerResolver callers;
private final Metrics metrics; // CB-502: null → auth failures not counted
private final CapacitySource capacity;
private final HealthCoverageSource healthCoverage;
@@ -409,6 +416,7 @@ public final class FleetMcp {
this.authorizationEnforced = Objects.requireNonNull(authorizationMode, "authorizationMode")
== AuthorizationMode.ENFORCED;
this.identity = identity;
this.callers = callers;
this.leadChannel = leadChannel;
this.peers = peers == null ? List.of() : List.copyOf(peers);
this.capacity = capacity;
@@ -689,7 +697,7 @@ public final class FleetMcp {
if (!authorizationEnforced) {
return null; // AuthorizationMode.UNENFORCED: authorization not enforced (fleetd #518)
}
if (Authz.permits(caller, action, target)) {
if (Authz.permits(caller, action, target, callers.knownLeadOrCollaborator())) {
if (action != Authz.Action.READ && action != Authz.Action.TASK_READ) {
AuditLog.allowed(caller, action, target); // reads would drown the trail
}
@@ -1309,6 +1317,18 @@ public final class FleetMcp {
}
return text(json(m));
}
if (caller.isCollaborator()) {
// A collaborator's name is its slot in the collaborators: registry; sessionId is its
// pane so a peer knows where to reach it. No leader key: a collaborator is not a
// primary for authorization, unlike a lead.
if (caller.name() != null) {
m.put("collaborator", caller.name());
}
if (caller.terminal() != null) {
m.put("sessionId", caller.terminal());
}
return text(json(m));
}
if (!caller.isWorker()) {
// CB-530: which lead, once more than one pane is configured as one. `role` deliberately
// still reads "primary" — the fallback ladder in CLAUDE.md keys on it, and a lead IS a
@@ -266,6 +266,21 @@ public final class FleetApp {
return app;
}
/**
* The authorization decision behind {@link #allow}, taking the caller directly rather than
* pulling it from a servlet {@link Context} — unit-testable without fabricating a live
* request, the same reason {@code FleetMcp#denyFor} is split from {@code FleetMcp#deny}.
*
* @param knownLeadOrCollaborator the classifier a collaborator's {@code SEND} is checked
* against; pass {@link #auth}'s own {@code
* knownLeadOrCollaborator()} to exercise the real production
* gate, as {@link #allow} does
*/
static boolean permitsFor(Principal caller, Authz.Action action, String target,
Predicate<String> knownLeadOrCollaborator) {
return Authz.permits(caller, action, target, knownLeadOrCollaborator);
}
/**
* Gate a handler on the CB-505 authorization table. Returns {@code true} when the request may
* proceed; otherwise writes the error response and returns {@code false}.
@@ -279,7 +294,7 @@ public final class FleetApp {
return true; // legacy: authorization not enforced
}
Principal caller = ctx.attribute(CALLER);
if (Authz.permits(caller, action, target)) {
if (permitsFor(caller, action, target, auth.knownLeadOrCollaborator())) {
if (action != Authz.Action.READ && action != Authz.Action.METRICS
&& action != Authz.Action.TASK_READ) {
AuditLog.allowed(caller, action, target); // reads would drown the trail
@@ -14,6 +14,7 @@ class AuthzTest {
private static final Principal ANON = Principal.anonymous();
private static final Principal ARCH_DESIGN = Principal.architect("lead-designer", "term_design", 400);
private static final Principal ARCH_OTHER = Principal.architect("reviewer", "term_review", 500);
private static final Principal COLLABORATOR = Principal.collaborator("ops", "term_collab", 600);
@Test
void anonymousIsAuthorizedForNothing() {
@@ -166,4 +167,87 @@ class AuthzTest {
assertFalse(Authz.isUnauthenticated(WORKER_A));
assertFalse(Authz.isUnauthenticated(PRIMARY));
}
// ── the collaborator matrix ─────────────────────────────────────────────────────────────────
/**
* {@code SEND} for a collaborator is the one grant that is conditional rather than fixed:
* flipping only the classifier's answer for the target flips only this outcome.
*/
@Test
void aCollaboratorMaySendOnlyWhenTheClassifierAcceptsTheTarget() {
assertTrue(Authz.permits(COLLABORATOR, SEND, "term_lead", target -> true),
"the classifier accepting the target must grant SEND");
assertFalse(Authz.permits(COLLABORATOR, SEND, "term_lead", target -> false),
"the classifier refusing the target must deny SEND");
assertFalse(Authz.permits(COLLABORATOR, SEND, "term_lead"),
"the real production classifier recognises no terminal yet, so SEND is refused today");
}
/**
* Control for the test above: every other action's result for a collaborator does not move
* when the classifier does. Only {@code SEND} is wired to it.
*/
@Test
void theClassifierMovesOnlySendForACollaborator() {
for (Authz.Action a : Authz.Action.values()) {
if (a == SEND) {
continue;
}
assertEquals(
Authz.permits(COLLABORATOR, a, "term_lead"),
Authz.permits(COLLABORATOR, a, "term_lead", target -> true),
a + " must not depend on the classifier at all");
}
}
@Test
void aCollaboratorMayReadAndScrapeMetrics() {
assertTrue(Authz.permits(COLLABORATOR, READ, null));
assertTrue(Authz.permits(COLLABORATOR, METRICS, null));
}
@Test
void aCollaboratorMayReplyAndAskOnlyAsItsOwnPane() {
assertTrue(Authz.permits(COLLABORATOR, REPLY, "term_collab"),
"its own pane is its own");
assertTrue(Authz.permits(COLLABORATOR, ASK, "term_collab"));
assertFalse(Authz.permits(COLLABORATOR, REPLY, "term_design"),
"a collaborator must not reply on another pane");
assertFalse(Authz.permits(COLLABORATOR, REPLY, null),
"an absent target must not pass the own-session rule");
}
/**
* Every action denied to a collaborator, asserted denied even when the classifier would
* accept any target — proving none of these is actually gated on the classifier at all.
*/
@Test
void aCollaboratorIsDeniedLifecycleCoordinationAndTicketPolling() {
for (Authz.Action a : new Authz.Action[]{SPAWN, STOP, DRAIN, HANDOVER, ANSWER, COORD_SEND,
COORD_READ, TASK_READ}) {
assertFalse(Authz.permits(COLLABORATOR, a, "term_lead", target -> true),
"a collaborator must not " + a + " even when the classifier accepts every target");
}
}
@Test
void aCollaboratorIsNotCountedAsPrimaryWorkerOrArchitect() {
assertFalse(COLLABORATOR.isPrimary());
assertFalse(COLLABORATOR.isWorker());
assertFalse(COLLABORATOR.isArchitect());
assertTrue(COLLABORATOR.isCollaborator());
}
/**
* A collaborator is never spawned, so it must not be enrolled in the presence map as an
* available member. Control: both a worker and an architect — which ARE spawned — still are.
*/
@Test
void isSpawnedMemberIsFalseForACollaboratorButTrueForAWorkerAndAnArchitect() {
assertFalse(COLLABORATOR.isSpawnedMember());
assertTrue(WORKER_A.isSpawnedMember());
assertTrue(ARCH_DESIGN.isSpawnedMember());
}
}
@@ -523,4 +523,129 @@ class CallerResolverTest {
}
}
// ── fleetd #669 Unit D: a live spawned member outranks every tab map ────────────────────────
/**
* Criterion 1: a terminal present in BOTH the spawned-member roster AND the lead tab map
* resolves as its member role, not as a lead — the roster is checked first, consulting no tab
* map at all when it matches.
*/
@Test
void aSpawnedMemberWinsOverALeadTabForTheSamePane() {
Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
() -> Map.of("term_a", "opus-5.0"), new MemberRegistry(null),
t -> "term_a".equals(t) ? MemberRole.DEV : null, Map::of)
.resolve("127.0.0.1", 42, null);
assertEquals(Role.WORKER, p.role(),
"a live spawned member's own identity must win over a tab map naming the same pane a lead");
assertEquals("term_a", p.terminal());
}
/** A spawned architect in the roster resolves ARCHITECT, carrying its bound slot's name. */
@Test
void aSpawnedArchitectInTheRosterResolvesArchitectWithItsSlotName() {
Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
boundMembers("architect:lead-designer", MemberRole.ARCHITECT),
t -> "term_a".equals(t) ? MemberRole.ARCHITECT : null, Map::of)
.resolve("127.0.0.1", 42, null);
assertEquals(Role.ARCHITECT, p.role());
assertEquals("lead-designer", p.name());
assertEquals("term_a", p.terminal());
}
/**
* fleetd #424 regression: the roster only answers THAT a pane is a live spawned member; config
* still decides WHAT that member's slot grants. A slot revoked after the bind must still demote
* the session on its very next request, exactly as it would for a pane with no roster entry at
* all — the roster's own ARCHITECT role must never be granted on its word alone.
*
* <p>{@code bind} refuses an unconfigured slot, so the revoked state can only be reached by
* binding while the slot is configured and then swapping the config out from under it, the way
* a live reload does.
*/
@Test
void aRevokedArchitectSlotDemotesALiveSpawnedArchitectToWorker() {
FleetConfig.Fleet configured = new FleetConfig.Fleet(Map.of(),
Map.of("lead-designer", new FleetConfig.Slot("sonnet")), Map.of(), Map.of(), null);
java.util.concurrent.atomic.AtomicReference<FleetConfig.Fleet> live =
new java.util.concurrent.atomic.AtomicReference<>(configured);
MemberRegistry members = MemberRegistry.live(live::get);
assertTrue(members.bind("architect:lead-designer", "term_a"));
live.set(new FleetConfig.Fleet(Map.of(), Map.of(), Map.of(), Map.of(), null)); // slot revoked
// Setup controls: the slot is really gone from config, but the occupancy is still there —
// otherwise this test would pass for the wrong reason.
assertNull(members.roleForSlot("architect:lead-designer"), "setup control: the slot must be gone from config");
assertEquals("architect:lead-designer", members.snapshot().get("term_a"),
"setup control: the binding itself must still be there");
Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
members, t -> "term_a".equals(t) ? MemberRole.ARCHITECT : null, Map::of)
.resolve("127.0.0.1", 42, null);
assertEquals(Role.WORKER, p.role(),
"a revoked slot must demote a live spawned architect on its very next request");
}
/** Criterion 3: a configured collaborator tab that is not a spawned member resolves COLLABORATOR. */
@Test
void aConfiguredCollaboratorTabResolvesToCollaboratorCarryingItsName() {
Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
new MemberRegistry(null), t -> null, () -> Map.of("term_a", "ops"))
.resolve("127.0.0.1", 42, null);
assertEquals(Role.COLLABORATOR, p.role());
assertEquals("ops", p.name());
assertEquals("term_a", p.terminal());
}
/** Regression: an empty collaborator registry leaves every pane exactly as before. */
@Test
void anEmptyCollaboratorRegistryLeavesEveryPaneAsBefore() {
Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
new MemberRegistry(null), t -> null, Map::of)
.resolve("127.0.0.1", 42, null);
assertEquals(Role.WORKER, p.role());
assertNull(p.name());
}
@Test
void describeNamesTheCollaborator() {
assertEquals("collaborator:ops", Principal.collaborator("ops", "term_a", 1).describe());
}
// ── fleetd #669 Unit D: knownLeadOrCollaborator() reads the same maps resolve() does ───────────
@Test
void knownLeadOrCollaboratorIsTrueForALeadTerminal() {
CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
() -> Map.of("term_lead", "opus-5.0"), new MemberRegistry(null), t -> null, Map::of);
assertTrue(r.knownLeadOrCollaborator().test("term_lead"));
assertFalse(r.knownLeadOrCollaborator().test("term_other"));
}
@Test
void knownLeadOrCollaboratorIsTrueForACollaboratorTerminal() {
CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
new MemberRegistry(null), t -> null, () -> Map.of("term_collab", "ops"));
assertTrue(r.knownLeadOrCollaborator().test("term_collab"));
assertFalse(r.knownLeadOrCollaborator().test("term_other"));
}
@Test
void knownLeadOrCollaboratorIsFalseForASpawnedMembersTerminal() {
// The exact scenario a collaborator's SEND must never reach: a live spawned member's own
// terminal, which is neither a configured lead nor a configured collaborator.
CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
new MemberRegistry(null), t -> "term_a".equals(t) ? MemberRole.DEV : null, Map::of);
assertFalse(r.knownLeadOrCollaborator().test("term_a"));
}
}
@@ -1000,6 +1000,273 @@ class FleetConfigTest {
"no primary.terminal pin ⇒ nothing registered, even with fleet.leaders configured");
}
// ── fleetd #669: the collaborators registry ────────────────────────────────────────────────
/**
* {@code Fleet} is {@code @JsonIgnoreProperties(ignoreUnknown = true)}, so a config naming
* {@code fleet.collaborators.<name>.tab} loads with no exception whether or not the key is
* ever read into the object model. Asserting only "no exception" would pass both before and
* after the real fix, so this asserts the parsed value is actually reachable from the loaded
* {@code FleetConfig} — the one thing a vacuous "no exception" test cannot tell apart.
*/
@Test
void collaboratorsBlockIsActuallyParsedNotSilentlyDropped(@TempDir Path dir) throws Exception {
Path f = dir.resolve("collaborators.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
reviewer-alex:
tab: "collab: alex"
""");
FleetConfig cfg = FleetConfig.load(f);
assertEquals("collab: alex", cfg.fleet().collaborators().get("reviewer-alex").tab());
}
/** A collaborator carries no field other than {@code tab}, so a blank one is meaningless. */
@Test
void aCollaboratorWithNoTabRefusesToStart(@TempDir Path dir) throws Exception {
Path f = dir.resolve("useless-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
ghost: {}
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e = assertThrows(IllegalStateException.class, cfg::validateMembers);
assertTrue(e.getMessage().contains("ghost"), "the message must name the useless entry");
assertTrue(e.getMessage().contains("tab:"), "the message must say what is missing");
}
/** Control for {@link #aCollaboratorWithNoTabRefusesToStart}: a named tab loads cleanly. */
@Test
void aCollaboratorWithATabIsAllowed(@TempDir Path dir) throws Exception {
Path f = dir.resolve("named-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
reviewer-alex:
tab: "collab: alex"
""");
FleetConfig cfg = FleetConfig.load(f);
assertDoesNotThrow(cfg::validateMembers);
}
/**
* fleetd #669: the fleet-wide {@code tabLabel} template can render as a collaborator tab, the
* same hazard {@link #aFleetTabLabelTemplateThatCanRenderAsALeadTabRefusesToStart} covers on
* the lead side. Drives the fleet-wide branch directly, with no profile override involved.
*/
@Test
void aFleetTabLabelTemplateThatCanRenderAsACollaboratorTabRefusesToStart(@TempDir Path dir)
throws Exception {
Path f = dir.resolve("collide-template-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
tabLabel: "al{profile}"
collaborators:
alex:
tab: "alpha"
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e =
assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
assertTrue(e.getMessage().contains("fleet.tabLabel"));
assertTrue(e.getMessage().contains("alex"), "the message must name the offending collaborator");
}
/**
* fleetd #669: a member tabLabel that can render as a configured collaborator tab is the same
* hazard as the lead case above — a member labelled that way is read back as the collaborator.
*/
@Test
void aProfileTabLabelOverrideMatchingACollaboratorTabRefusesToStart(@TempDir Path dir)
throws Exception {
Path f = dir.resolve("collide-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
profiles:
gx10:
tabLabel: "collab-tab"
fleet:
collaborators:
alex:
tab: "collab-tab"
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e =
assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
assertTrue(e.getMessage().contains("collab-tab"), "the message must name the offending label");
}
/** Control: a profile tabLabel that cannot render as the collaborator tab is allowed. */
@Test
void aProfileTabLabelThatCannotRenderAsACollaboratorTabIsAllowed(@TempDir Path dir)
throws Exception {
Path f = dir.resolve("ok-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
profiles:
gx10:
tabLabel: "worker-{profile}"
fleet:
collaborators:
alex:
tab: "collab-tab"
""");
FleetConfig cfg = FleetConfig.load(f);
assertDoesNotThrow(cfg::validateLeadTabPrefixes);
}
/** fleetd #669: identity is matched on a collaborator's exact tab, so two sharing one are unreachable. */
@Test
void twoCollaboratorsSharingTheSameExactTabRefusesToStart(@TempDir Path dir) throws Exception {
Path f = dir.resolve("shared-collaborator-tab.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
alex:
tab: "shared tab"
sam:
tab: "Shared Tab"
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e =
assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
assertTrue(e.getMessage().contains("alex"), "the message must name one offending collaborator");
assertTrue(e.getMessage().contains("sam"), "the message must name the other offending collaborator");
}
/** Control for {@link #twoCollaboratorsSharingTheSameExactTabRefusesToStart}: distinct tabs load cleanly. */
@Test
void twoCollaboratorsWithDistinctExactTabsAreAllowed(@TempDir Path dir) throws Exception {
Path f = dir.resolve("distinct-collaborator-tabs.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
alex:
tab: "alex tab"
sam:
tab: "sam tab"
""");
FleetConfig cfg = FleetConfig.load(f);
assertDoesNotThrow(cfg::validateLeadTabPrefixes);
}
/**
* fleetd #669: a collaborator tab equal to a lead tab crosses a privilege boundary — the worst
* of the three new collisions, since only one of the two identities is ever found.
*/
@Test
void aCollaboratorTabEqualToALeadTabRefusesToStart(@TempDir Path dir) throws Exception {
Path f = dir.resolve("lead-collaborator-collision.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
leaders:
opus:
tab: "shared tab"
collaborators:
alex:
tab: "Shared Tab"
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e =
assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
assertTrue(e.getMessage().contains("opus"), "the message must name the offending lead");
assertTrue(e.getMessage().contains("alex"), "the message must name the offending collaborator");
}
/** Control for {@link #aCollaboratorTabEqualToALeadTabRefusesToStart}: distinct tabs load cleanly. */
@Test
void aLeadAndACollaboratorWithDistinctTabsAreAllowed(@TempDir Path dir) throws Exception {
Path f = dir.resolve("lead-collaborator-ok.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
leaders:
opus:
tab: "lead tab"
collaborators:
alex:
tab: "collab tab"
""");
FleetConfig cfg = FleetConfig.load(f);
assertDoesNotThrow(cfg::validateLeadTabPrefixes);
}
/**
* fleetd #669: a pane-placed member can land in a collaborator's labelled tab exactly as it
* can land in a lead's — {@code validatePanePlacementAgainstLeadTabs()} must fire even when
* {@code fleet.leaders} is empty, which is the early-return the brief flagged as the bug.
*/
@Test
void aPanePlacedProfileWithACollaboratorTabRefusesToStartEvenWithNoLeaders(@TempDir Path dir)
throws Exception {
Path f = dir.resolve("pane-hazard-collaborator.yaml");
Files.writeString(f, """
bind:
port: 8080
profiles:
gx10:
placement: pane
fleet:
collaborators:
alex:
tab: "collab: alex"
""");
FleetConfig cfg = FleetConfig.load(f);
IllegalStateException e = assertThrows(IllegalStateException.class,
cfg::validatePanePlacementAgainstLeadTabs);
assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
}
/** Control: a pane-placed profile with no lead or collaborator tab configured is allowed. */
@Test
void aPanePlacedProfileWithNoLeaderOrCollaboratorTabIsAllowed(@TempDir Path dir) throws Exception {
Path f = dir.resolve("pane-no-tab-at-all.yaml");
Files.writeString(f, """
bind:
port: 8080
profiles:
gx10:
placement: pane
fleet:
collaborators:
alex: {}
""");
assertDoesNotThrow(() -> FleetConfig.load(f).validatePanePlacementAgainstLeadTabs(),
"a collaborator with no tab feeds nothing into the scanner, so pane placement is safe");
}
// ── CB-548: the architects registry ────────────────────────────────────────────────────────
@Test
@@ -1272,6 +1539,60 @@ class FleetConfigTest {
assertEquals(Set.of("sonnet"), cfg.fleet().pool(MemberRole.REVIEWER).keySet());
}
/**
* A duplicated name in {@code fleet.collaborators} is refused at parse time, like any other
* {@code fleet:} pool. See {@link #duplicateSlotNamesInOnePoolAreRejectedAtParseTime}.
*/
@Test
void duplicateCollaboratorNamesAreRejectedAtParseTime(@TempDir Path dir) throws Exception {
Path f = dir.resolve("collaborator-dup.yaml");
Files.writeString(f, """
bind:
port: 8080
fleet:
collaborators:
alex:
tab: "collab: alex"
alex:
tab: "collab: alex, second"
""");
IllegalStateException e =
assertThrows(IllegalStateException.class, () -> FleetConfig.load(f));
assertTrue(e.getMessage().contains("alex"),
"the refusal names the duplicated entry, was: " + e.getMessage());
assertTrue(e.getMessage().contains("fleet.collaborators"),
"the refusal names the pool the duplicate is in, was: " + e.getMessage());
}
/**
* Control for {@link #duplicateCollaboratorNamesAreRejectedAtParseTime}: the same name reused
* across the collaborators registry and a member role pool is the role × profile matrix doing
* its job in the other pool, not a mistake — only a repeat within one pool loses an entry.
*/
@Test
void theSameNameInCollaboratorsAndAnotherPoolIsNotADuplicate(@TempDir Path dir) throws Exception {
Path f = dir.resolve("collaborator-cross-pool.yaml");
Files.writeString(f, """
bind:
port: 8080
profiles:
sonnet:
baseUrl: http://gx10.gw:8000
fleet:
architects:
alex:
profile: sonnet
collaborators:
alex:
tab: "collab: alex"
""");
FleetConfig cfg = assertDoesNotThrow(() -> FleetConfig.load(f));
assertEquals(Set.of("alex"), cfg.fleet().pool(MemberRole.ARCHITECT).keySet());
assertEquals(Set.of("alex"), cfg.fleet().collaborators().keySet());
}
@Test
void duplicateKeysOutsideTheFleetPoolsAreUnaffected(@TempDir Path dir) throws Exception {
// The duplicate check is scoped to the fleet pools — a duplicate elsewhere is not this
@@ -163,6 +163,13 @@ class LeadTabScannerTest {
return new LeadTabScanner(herdr, tabToName, Set.of("fleetd-workers"), TTL, clock::get);
}
private LeadTabScanner scannerWithCollaborators(TopologyHerdr herdr, Map<String, String> tabToName,
Map<String, String> collaboratorTabToName,
AtomicLong clock) {
return new LeadTabScanner(herdr, tabToName, collaboratorTabToName,
Set.of("fleetd-workers"), TTL, clock::get);
}
@Test
void everyConfiguredTabBecomesALeadNamedByItsEntry() {
Map<String, String> leads = scanner(twoLeads(), twoLeadsConfigured(), new AtomicLong()).get();
@@ -464,4 +471,75 @@ class LeadTabScannerTest {
assertEquals(afterFirst, herdr.calls, "the failure path must be rate-limited too");
}
// ── fleetd #669 Unit D: a collaborator tab is matched the same way as a lead tab, one pass ─────
@Test
void aConfiguredCollaboratorTabIsReportedByCollaboratorsNotByGet() {
TopologyHerdr herdr = new TopologyHerdr()
.workspace("w1", "main")
.tab("w1:t1", "w1", "collab: ops")
.pane("w1:p1", "w1:t1", "term_ops");
LeadTabScanner s = scannerWithCollaborators(herdr, Map.of(), Map.of("collab: ops", "ops"),
new AtomicLong());
assertEquals(Map.of("term_ops", "ops"), s.collaborators(),
"a collaborator tab is matched exactly like a lead tab");
assertEquals(Map.of(), s.get(), "a collaborator tab must never also appear as a lead");
}
/**
* Criterion 4, scanner level: a collaborator tab that is labelled but runs no agent is not
* reported — the same #359 liveness cross-check a lead tab gets.
*/
@Test
void aDeadCollaboratorTabIsNotReported() {
TopologyHerdr herdr = new TopologyHerdr()
.workspace("w1", "main")
.tab("w1:t1", "w1", "collab: ops")
.pane("w1:p1", "w1:t1", "term_ops")
.deadAgent("w1:t1");
LeadTabScanner s = scannerWithCollaborators(herdr, Map.of(), Map.of("collab: ops", "ops"),
new AtomicLong());
assertFalse(s.collaborators().containsKey("term_ops"),
"a dead collaborator tab must never resolve as a live collaborator");
}
/**
* Both kinds are matched in a single pass over the same tab list — not a second scanner, not a
* second scan. Proven by herdr call count: scanning one lead tab and one collaborator tab in the
* same instance costs exactly as many calls as scanning two lead tabs in {@link #twoLeads()}.
*/
@Test
void leadsAndCollaboratorsAreMatchedInOnePassOverTheSameScan() {
TopologyHerdr oneOfEach = new TopologyHerdr()
.workspace("w1", "main")
.tab("w1:t1", "w1", "lead: opus-5.0")
.tab("w1:t2", "w1", "collab: ops")
.pane("w1:p1", "w1:t1", "term_opus")
.pane("w1:p2", "w1:t2", "term_ops");
LeadTabScanner s = scannerWithCollaborators(oneOfEach, Map.of("lead: opus-5.0", "opus-5.0"),
Map.of("collab: ops", "ops"), new AtomicLong());
assertEquals(Map.of("term_opus", "opus-5.0"), s.get());
assertEquals(Map.of("term_ops", "ops"), s.collaborators());
TopologyHerdr twoLeadsBaseline = twoLeads();
scanner(twoLeadsBaseline, twoLeadsConfigured(), new AtomicLong()).get();
assertEquals(twoLeadsBaseline.calls, oneOfEach.calls,
"one lead tab + one collaborator tab must cost exactly as many herdr calls as two "
+ "lead tabs — proof this is one pass, not a second scan");
}
/** Regression: with no collaborators configured, every existing lead-only behaviour is unchanged. */
@Test
void anEmptyCollaboratorMapLeavesCollaboratorsEmptyAndGetUnaffected() {
LeadTabScanner s = scannerWithCollaborators(twoLeads(), twoLeadsConfigured(), Map.of(),
new AtomicLong());
assertEquals(Map.of(), s.collaborators());
assertEquals(Map.of("term_opus", "opus-5.0", "term_gpt", "gpt-sol-5.6"), s.get());
}
}
@@ -63,6 +63,17 @@ class FleetMcpAuthzTest {
/** A fully wired FleetMcp on fakes — constructing it is itself part of what is under test. */
private FleetMcp mcp(boolean enforce) {
ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
return mcp(enforce, CallerResolver.withLeadsAndMembers(identity, false, null,
Map::of, new MemberRegistry(null)));
}
/**
* As {@link #mcp(boolean)}, with an explicit {@link CallerResolver} — so a test can wire known
* leads/collaborators and drive {@code denyFor}'s real {@code knownLeadOrCollaborator()}
* classifier instead of the default empty one.
*/
private FleetMcp mcp(boolean enforce, CallerResolver callers) {
FleetConfig.Profile cfg = new FleetConfig.Profile(
"ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
"tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
@@ -83,9 +94,7 @@ class FleetMcpAuthzTest {
// to be omitted to reach "legacy" is now always real, and AuthorizationMode is the
// separate, explicit choice that governs enforcement.
mcp = new FleetMcp(messages, workers, sessions, identity, sessions.asPresence(),
new PrimaryRegistry(null),
CallerResolver.withLeadsAndMembers(identity, false, null,
Map::of, new MemberRegistry(null)),
new PrimaryRegistry(null), callers,
enforce ? FleetMcp.AuthorizationMode.ENFORCED : FleetMcp.AuthorizationMode.UNENFORCED,
metrics, FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
FleetMcp.QuarantineSource.none(), null, FleetMcp.OutageSource.none(),
@@ -97,6 +106,7 @@ class FleetMcpAuthzTest {
private static final Principal WORKER_A = Principal.worker("term_a", 200);
private static final Principal ANON = Principal.anonymous();
private static final Principal ARCH_DESIGN = Principal.architect("lead-designer", "term_design", 400);
private static final Principal COLLABORATOR = Principal.collaborator("ops", "term_collab", 600);
// --- the table, enforced on THIS path too ---------------------------------------------------
@@ -226,6 +236,39 @@ class FleetMcpAuthzTest {
"the caller IS authenticated — it is just not the right role");
}
/**
* {@code denyFor} passes the real production classifier, not a test-supplied one — no
* terminal is recognised as a configured lead or collaborator, so a collaborator's SEND is
* refused over MCP.
*/
@Test
void aCollaboratorMayNotSendOverMcpWithTheRealProductionClassifier() {
FleetMcp m = mcp(true);
McpSchema.CallToolResult denied = m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_lead");
assertNotNull(denied, "no terminal is recognised as a lead or collaborator yet");
assertTrue(denied.isError());
}
/**
* fleetd #669 Unit D: wires a real {@link CallerResolver} with a known lead and a known
* collaborator tab, and leaves a spawned member's own terminal recognised by neither map — so a
* collaborator's SEND reaches both named peers and is refused for the spawned member's terminal,
* over MCP's {@code denyFor}.
*/
@Test
void aCollaboratorMaySendToAKnownLeadOrCollaboratorButNotToASpawnedMembersTerminal() {
ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
() -> Map.of("term_lead_known", "lead-x"), new MemberRegistry(null),
t -> null, () -> Map.of("term_collab_known", "ops2"));
FleetMcp m = mcp(true, callers);
assertNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_lead_known"));
assertNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_collab_known"));
assertNotNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_a"),
"a spawned member's own terminal must stay unreachable, even once the classifier is real");
}
@Test
void theLegacyConstructorLeavesTheGateOpen() {
// The 22 pre-existing FleetMcpTest cases rely on no authorization being enforced.
@@ -1837,6 +1837,32 @@ class FleetMcpTest {
assertTrue(out.contains("\"sessionId\":\"term_design\""), out);
}
/**
* A collaborator reports its own role and name, never the {@code leader} key a lead gets —
* {@code role} already reads {@code "collaborator"}, so a {@code leader} key alongside it
* would be self-contradicting. Control: the same call shape fed a named lead must still carry
* {@code leader}, so this is not passing because the key stopped being emitted for everyone.
*/
@Test
void whoamiReportsACollaboratorWithNoLeaderKeyButALeadStillGetsOne() {
FakeHerdr h = new FakeHerdr();
SessionManager sessions = sessionManager(h, "http://gx00.gw:8000", Set.of("gx00.gw"));
McpSchema.CallToolResult collabRes = FleetMcp.whoami(
Principal.collaborator("ops", "term_collab", 700), sessions);
assertNotEquals(Boolean.TRUE, collabRes.isError());
String collabOut = textOf(collabRes);
assertTrue(collabOut.contains("\"role\":\"collaborator\""), collabOut);
assertTrue(collabOut.contains("\"collaborator\":\"ops\""), collabOut);
assertTrue(collabOut.contains("\"sessionId\":\"term_collab\""), collabOut);
assertFalse(collabOut.contains("leader"), collabOut);
McpSchema.CallToolResult leadRes = FleetMcp.whoami(
Principal.leader("opus", "term_lead", 100), sessions);
String leadOut = textOf(leadRes);
assertTrue(leadOut.contains("\"leader\":\"opus\""), leadOut);
}
/**
* CB-548: an architect SEND delegates as its own pane (recording the per-target delegation) but
* must NEVER become the legacy singleton "primary" fallback — the per-target map does not cure
@@ -6,6 +6,7 @@ import com.fasterxml.jackson.databind.ObjectMapper;
import dev.ltms.fleet.auth.CallerResolver;
import dev.ltms.fleet.auth.Authz;
import dev.ltms.fleet.auth.MemberRegistry;
import dev.ltms.fleet.auth.Principal;
import dev.ltms.fleet.config.FleetConfig;
import dev.ltms.fleet.guard.SubscriptionGuard;
import dev.ltms.fleet.herdr.AgentControl;
@@ -202,6 +203,74 @@ class FleetAppAuthTest {
}
}
/**
* {@code permitsFor} is the exact decision {@link FleetApp#allow} makes, passing the real
* production classifier rather than a test-supplied one — built from an empty {@link
* CallerResolver}, so no terminal is recognised as a configured lead or collaborator and a
* collaborator's SEND is refused through the REST gate.
*/
@Test
void aCollaboratorMayNotSendOverRestWithTheRealProductionClassifier() {
ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(new FakeHerdr()), _ -> 700L);
CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
Map::of, new MemberRegistry(null));
Principal collaborator = Principal.collaborator("ops", "term_collab", 700);
assertFalse(FleetApp.permitsFor(collaborator, Authz.Action.SEND, "term_lead",
callers.knownLeadOrCollaborator()),
"no terminal is recognised as a lead or collaborator yet");
}
/**
* fleetd #669 Unit D: wires a real {@link CallerResolver} with a known lead and a known
* collaborator tab, and a spawned member's own terminal recognised by neither map. A
* collaborator's SEND reaches the known lead and the known collaborator, and is refused for the
* spawned member's terminal — over the REST route, not just the unit-level classifier, so a
* test covering only MCP cannot leave this route open.
*/
@Test
void aCollaboratorMaySendToAKnownLeadOrCollaboratorButNotToASpawnedMembersTerminalOverRest() throws Exception {
int port = startWithRealClassifier(FakeHerdr.WORKER_PID,
Map.of("term_lead_known", "lead-x"), Map.of("term_a", "ops2"));
HttpResponse<String> toLead = send(port, "POST", "/sessions/term_lead_known/message",
"{\"content\":\"hi\",\"wait\":false}", null);
assertEquals(202, toLead.statusCode(), toLead.body());
HttpResponse<String> toSpawnedMembersTerminal = send(port, "POST", "/sessions/term_worker/message",
"{\"content\":\"hi\",\"wait\":false}", null);
assertEquals(403, toSpawnedMembersTerminal.statusCode(), toSpawnedMembersTerminal.body());
}
/**
* As {@link #start}, but with explicit lead/collaborator maps and no spawned-member roster, so
* a test can wire the real {@link CallerResolver#knownLeadOrCollaborator()} classifier instead
* of the default empty one.
*/
private int startWithRealClassifier(long pid, Map<String, String> leadTerminals,
Map<String, String> collaboratorTerminals) {
FakeHerdr herdr = new FakeHerdr();
FleetConfig.Profile wcfg = new FleetConfig.Profile(
"ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
"tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
AgentControl agents = new AgentControl(herdr);
ClaudeCodeLauncher workers = new ClaudeCodeLauncher(
agents, new WorkspaceControl(herdr), new SubscriptionGuard(Set.of("gx00.gw")),
Map.of(wcfg.profile(), wcfg), wcfg.profile(),
k -> "FLEETD_WORKER_TOKEN".equals(k) ? "tok-abc" : null);
SessionManager sessions = new SessionManager(workers, new FakeWorktrees());
Injector injector = new Injector(agents);
MessageService messages = new MessageService(agents, injector, new Rendezvous());
ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> pid);
CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
() -> leadTerminals, new MemberRegistry(null), t -> null, () -> collaboratorTerminals);
metrics = FleetMetrics.create(sessions, new dev.ltms.fleet.msg.InMemoryReplyInbox());
app = new FleetApp(herdr, workers, sessions, messages, sessions.asPresence(), null,
callers, metrics).build().start("127.0.0.1", 0);
return app.port();
}
// --- loopback-trust: the caller is the primary -------------------------------------------
@Test