Skip to content

refactor(distributed): put the NATS-only paths behind interfaces - #12395

Merged
mudler merged 37 commits into
masterfrom
refactor/distributed-seams
Oct 2, 2026
Merged

mudler merged 37 commits into
masterfrom
refactor/distributed-seams

Conversation

@mudler-agent

Copy link
Copy Markdown
Collaborator

What this changes

Distributed mode talks to other processes through NATS in many places. This PR puts each of those places behind an interface that production code uses, with NATS and direct dialing as the only implementations. No second implementation is added here.

The goal is to make a second transport an addition instead of a rewrite. It is the first step of landing the broker-free work in #11812 next to NATS rather than in place of it, so a deployment can choose a broker or a PostgreSQL-only mode.

Area Interface Implementation here
Fan-out events messaging.Broadcaster NATS client
Competing consumers messaging.WorkQueue, messaging.WorkConsumer NATS queue groups
Control verbs, frontend side nodes.NodeCommandSender, nodes.FileStager NATS request/reply, file transfer
Control verbs, worker side controlServer (package worker, unexported) NATS subscriptions
MCP calls to agent workers mcp.AgentControl, agentRPCServer NATS request/reply
Dialing a worker nodes.BackendClientFactory (now takes the node id), nodes.ModelProber, nodes.WorkerNetDialerFor direct dial

Supporting changes:

  • Every subject goes through one set of rules (messaging.ValidateSubject): a closed set of roots, and * as the only wildcard. The NATS client and the in-memory test bus both enforce it.
  • nodes.ErrNoRoute names the "nothing is listening for that node" condition. The carrier error is mapped onto it in one file, so no scheduler code matches a NATS error any more. A timeout and a worker's own refusal are not ErrNoRoute.
  • The control request and reply structs moved from messaging to a new workerctl package. The move is mechanical and the JSON bytes are pinned by golden specs.
  • Every dial of a worker goes through the seam, including the reconciler probe, the file transfer client and the admin backend-logs proxy.
  • Code that nothing in production reached is deleted: the jobs.new consumer, NATSDispatcher.Dispatch, QueueSubscribeJSON, the frontend jobs.*.cancel subscription and the JobWorkerConcurrency config field, which no flag, option or decoder ever read. Publishing jobs.new is unchanged.
  • A short guide for contributors is in .agents/distributed-seams.md.

Behaviour

A NATS deployment should see no change, with the differences below.

  • A subject outside the served roots is refused with ErrUnservedSubject. The one operator-set subject, LOCALAI_AGENT_SUBJECT, is now checked at startup and the error names the setting.
  • MCP tool errors read remote MCP tool request failed: ... instead of NATS MCP tool request failed: ....
  • An empty list-dir answer from a worker is {} instead of {"files":null}. The frontend decodes both to an empty list.
  • The per-node file transfer clients each keep their own idle connections, so MaxIdleConns is now per node instead of global.
  • The backend-logs proxy keeps connections alive for 15s instead of 30s.
  • Malformed worker requests now log Ignoring malformed control request. Registration errors read serving <verb>: ....
  • The log line No NATS responders for node, marking unhealthy now reads No route to node, marking unhealthy. An alert that matches the old text needs the new one.
  • An explicitly empty LOCALAI_AGENT_QUEUE still gives a plain subscription, as before.

How to verify

go build ./core/... ./pkg/...
go vet ./core/... ./pkg/... ./tests/...
go test -count=1 ./core/services/... ./core/cli/... ./core/http/... ./core/application/... ./pkg/...
make lint

The NATS conformance suite and the distributed end-to-end specs start a NATS container, so they need Docker:

go test -count=1 ./core/services/messaging/...
go test -count=1 ./tests/e2e/distributed/

On the branch the unit run passes 90 packages and the distributed end-to-end suite passes 238 of 239 specs. The skipped spec needs a multi-node vLLM setup.

The wire formats that must not drift are pinned by specs: the worker reply bytes for every control verb, the queue group of each work kind, the in-flight limit each consumer asks for, and the node id that reaches every dialer.

Review notes

  • The branch has many small commits. Most are one step each: a mechanical rename, a seam, a deletion. The payload move to workerctl is one commit of about 40 files and is safe to skim.
  • Known follow-ups for the second implementation are listed in .agents/distributed-seams.md. The main one: the agent-run consumer publishes events through the process-wide event bridge, so the next step has to change that consumer.

🤖 Generated with Claude Code

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…arrier

Broadcaster is the fan-out half of MessagingClient. The NATS client and
the in-memory FakeBus now refuse a subject outside the served roots and any
wildcard other than a whole single token, and FakeBus shares MatchSubject
instead of its own copy. FakeBus Unsubscribe now removes its own
subscription instead of the first one with the same subject.

A shared conformance suite in messagingtest runs against both carriers.
The distributed e2e specs that used invented test.* subjects, and the one
that subscribed with a > filter, now use subjects from subjects.go.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…used

Narrowed to messaging.Broadcaster: nodes/staging_progress.go,
nodes/install_progress_publisher.go, galleryop/operation.go,
galleryop/service.go, agentpool/user_services.go, agentpool/agent_jobs.go,
openresponses/store.go, openresponses/sync.go, syncstate/syncstate.go,
finetune/service.go, quantization/service.go and
failover/distsync/distsync.go. SubscribeJSON now takes a Broadcaster
because it only calls Subscribe, which lets the narrowed consumers use it.

Stayed wide: worker/supervisor.go, because its client field also serves
the SubscribeReply handlers in worker/lifecycle.go. The request/reply,
queue and wiring files (nodes/unloader.go, nodes/file_stager_s3.go,
jobs/dispatcher.go, agents/dispatcher.go, agents/events.go,
worker/file_staging.go, cli/agent_worker.go, http/app.go) are unchanged by
design.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…error

Consumers matched nats.ErrNoResponders, which names an absence, to demote a node. They now match ErrNoRoute, the control path maps the carrier's failure onto it, and timeouts and worker refusals are pinned as not being no-route.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The messaging client now refuses a subject whose root no carrier serves.
An agent worker started with a custom LOCALAI_AGENT_SUBJECT such as
tenant-a.agent.execute used to start and then wait on a subject the
frontend never publishes to. After the subject rules landed it exited
at subscribe time with an error that did not name the setting.

Behaviour change: the worker now checks LOCALAI_AGENT_SUBJECT before it
registers or connects, and exits with an error that names the variable
and says to use a served subject under the agent root, for example
agent.execute. The served roots are not widened: a custom root was
never delivered by the frontend, and a wider set would reopen the
drift the subject rules exist to close. The flag help and the agent
worker docs state the constraint.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Three callers react to ErrNoRoute and had no spec: the reconciler's
upgrade drain falls back to the legacy forced install, the reconciler
marks the node unhealthy when a pending op has no route, and the
backend-op fan-out marks the node unhealthy. Each spec drives the real
caller with a scripted no-responders reply and reads the result from
the registry or the recorded requests.

A fourth spec pins the other side: a pending op that times out leaves
the node healthy and only counts the attempt, so mapping timeouts onto
ErrNoRoute would fail here.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
… in CI

Add specs that call Publish, Request, Subscribe, QueueSubscribe,
SubscribeReply and QueueSubscribeReply on a client with no connection.
Each call must return ErrUnservedSubject for bogus.thing and
ErrUnsupportedWildcard for jobs.>. This proves that the subject check
runs before the connection is used, and needs no server.

The NATS conformance suite is the only check that runs the subject rules
against a real carrier. Before this change it skipped without output
when Docker was missing. Now it fails when CI is set, so a Linux runner
without Docker cannot hide it. It still skips on local runs and on macOS
CI, which has no Docker.

Add SubjectNodeBackendInstallProgress to the list of constructors that
must build served subjects, and ask contributors to extend the list.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The seams note said MarkUnhealthy was the only state change allowed
on ErrNoRoute. A pending backend op still records the failed attempt,
counts toward the reconciler's retry limit and is dead-lettered after
the maximum attempts. The note now says that MarkUnhealthy is the only
change to the node's own state, and that the per-op accounting is not
a verdict about the node.

The note also documents that the NATS conformance run fails under CI
when Docker is missing. The distributed-seams row moves next to the
distributed-state row in the topics table. The liveness ping spec
header now says no route is a reason to skip the worker, not proof
that the worker is gone.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Mechanical: the method gains a nodeID parameter and the eight test fakes are updated. No behaviour change.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The node id is now in the main method, so the optional interface and its helper had no behaviour of their own.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The admin backend-logs proxy (list, lines and the WebSocket stream) now reaches a worker through the same per-node dialer as the HTTP file stager, so every frontend-to-worker dial goes through one seam. The shared direct dialer keeps alive for 15s where the proxy used 30s. Harmless for requests bounded at 15s.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…ction

The proxy request had no context before the dialer change and is bounded only by its 15s timeout. Keep it that way.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Mechanical move of the request and reply structs, the install progress event and the file payloads out of messaging. The verbs no longer belong to one carrier. No alias is left behind.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The worker registers one handler per verb and a NATS server maps each verb to its subject. Registration errors now name the verb. node.stop is served with SubscribeReply, which is identical on the wire because the handler never replies.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Install and upgrade now emit download progress through the sink the control server hands them. The debounce and the terminal flush stay in the handler path, built over that sink by the new nodes.NewDebouncedInstallProgressSink, which replaces NewDebouncedInstallProgressPublisher. The subject and payload on the wire are unchanged. The supervisor no longer holds the bus, and installFn and upgradeFn let specs drive both verbs without a gallery.

The malformed-request log lines are restored for install, upgrade, backend.delete, model.unload, model.stop and model.delete, with the reply bytes unchanged. The signal adapter is renamed noReply, which also lets worker.go import os/signal without an alias again.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
An empty list-dir answer is now {} rather than {"files":null}, because the typed reply omits an empty Files slice. The frontend decodes both to a nil slice in nodes/file_stager_s3.go.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This is the producer side of the competing-consumer seam. The work kinds map one to one to today's subjects and queue groups: task to jobs.new and mcp-ci to jobs.mcp-ci.new (both in group workers), agent-run to agent.execute (group agent-workers). Enqueue publishes the payload as Publish does today, with one JSON marshal. FakeBus now records queue groups and keeps reply handlers so later specs can pin and drive them.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
An in-flight limit of one runs the handler inline on the delivery goroutine, as the MCP CI consumer does today. Any other limit spawns per delivery, as the agent consumer does. Queue groups are unchanged.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The job dispatcher, the agent pool and the agent scheduler enqueue through messaging.WorkQueue; the NATS implementation publishes to the same subjects as before. DistributedServices builds the queue next to the NATS client and hands it to the dispatcher and the agent pool, whose distributed mode switch now reads a non-nil WorkQueue. The unused AgentPoolService.SetNATSClient is removed.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The agent dispatcher and the MCP CI consumer register through messaging.WorkConsumer. The NATS implementation keeps the inline one-at-a-time model for MCP CI and the per-delivery model for agent runs. handleMCPCIJob reports on the events publisher the carrier hands it instead of a captured client.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
jobs.new has a producer and no production consumer, and the agent dispatcher's Dispatch was only called from tests. Publishing jobs.new is unchanged.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Timeouts still honour only the deadline, not cancellation, exactly as today. The NATS no-responders error maps to ErrNoRoute and a timeout does not.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…gentRPCServer

The agent worker's MCP tool and discovery reply subscriptions and its backend stop listener move behind an unexported agentRPCServer interface, served on NATS by nodes.NATSAgentRPCServer. The handlers become typed mcp.ToolHandler and mcp.DiscoveryHandler values that answer every failure with a reply carrying Error.

Queue group (agent-workers), inline execution on the delivery goroutine, the background handler context, the unmarshal error reply texts and the reply-less backend stop subscription are unchanged. The backend stop handler takes the decoded backend name, so it can still close that backend's MCP sessions.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
mudler added 7 commits October 1, 2026 00:17
BroadcastRoots, ControlRoots and SubjectRoot had no production caller. The roots spec now asserts every served root through ValidateSubject instead. MatchSubject moves back into the test support package, the only place that used it, with its table. NATSAgentRPCServer drops the subscription list it stored and never read, and NewNATSAgentRPCServer gets a doc comment.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
One spec sends a tool request and a discovery request through NATSAgentControl to NATSAgentRPCServer and checks that the handlers see the decoded requests and the replies come back. It also puts an undecodable body on the tool subject and checks the server answers with an unmarshal error instead of leaving the requester to time out.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The developer note now lists the final seams: fan-out, queues, both halves of the control verbs and of agent RPC, and the dial. It records the open items a second carrier has to handle.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
The work queue specs pin what Consume does for a given limit, but nothing
pinned which limit each production consumer passes. Changing the agent
worker's MCP CI limit from 1 to 0 would have let MCP CI jobs run
concurrently on each worker with every test green.

Move the MCP CI Consume call into startMCPCIConsumer with the same wiring
and pin that it asks for (WorkMCPCI, 1). Pin that NATSDispatcher.Start asks
for (WorkAgentRun, maxConcurrent) for several limits.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
SubjectJobCancelWildcard lost its last subscriber when the frontend
stopped listening on jobs.*.cancel; the NATS permissions and conformance
suite spell the subject out, so nothing reads the constant.

decodeBackendStopRequest returned a stopAll flag that production dropped
and only a test read. decodeBackendStop is now the single decoder with the
same semantics: an empty body is stop-all, an empty Backend is stop-all,
malformed JSON is an error. stopBackends still derives stop-all from
Backend, so no reply changes.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
…tion

Before the work queue seam the agent worker passed LOCALAI_AGENT_QUEUE
straight to QueueSubscribe, so an explicitly empty value made a plain
subscription and every agent worker ran every agent run.
WithAgentRunRoute replaced an empty queue with agent-workers, which
silently changed that.

Keep the queue as given once the option is applied. An empty subject still
falls back to agent.execute, since it never had a meaning of its own. The
flag default stays agent-workers, so only an explicitly empty value
reaches this.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Fix the recordingFactory comment (it also records the parallel flag),
document that a negative maxInFlight is unbounded and that Unsubscribe from
a handler deadlocks, and say a permanently undecodable payload returns nil.

Record controlHandler's undecodable return as a kept exception, and add the
second carrier notes to the developer note: the reconciler has no
ClientFactory option, the logs proxy honours HTTP_PROXY, verbs one carrier
serves need an opt-out, terminal replies come from the result event, and
agent runs publish through the NATS-bound EventBridge, which is not an
additive change.

Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
@mudler-agent mudler-agent added enhancement New feature or request area/golang labels Oct 1, 2026
@mudler
mudler merged commit e4fa051 into master Oct 2, 2026
72 checks passed
@mudler
mudler deleted the refactor/distributed-seams branch October 2, 2026 21:58
@mudler-agent mudler-agent removed the enhancement New feature or request label Oct 22, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants