Files
pipelines/internal/server/websocket.go
T
alan 6cb61578de Fan a chat session out to every client and share approvals exactly-once
A session was owned by one connection: the claim map in chat_claims.go let the
first client hold it, every other client saw the session Busy and read-only, and
a second client's prompt was rejected. Replace the claim with a per-session
"room" so the operator's phone, tablet and terminal all watch and write the same
conversation, and a pending approval is a shared, exactly-once object.

Server (internal/server, internal/protocol):
- New chat_room.go: one room per session holding the session's single broker
  stream, its one bounded prompt queue (8) and the fan-out to every attached
  client. A chat_subscribe frame attaches a connection to a session's shared
  stream and replays the still-unresolved approval requests the room holds.
- chat_stream.go: handleChatPrompt accepts a prompt from any client. It attaches
  the sender, either starts a turn or queues the prompt behind the in-flight turn
  and acknowledges it with chat_queued:true ("queued" is the chosen behaviour,
  stated on the wire and in docs/cli.md), and echoes the user turn to every other
  client attributed to the sender's label. The prompt handler no longer owns the
  stream; one turn runner per room drains the queue, so two turns never
  interleave. Per-connection write serialisation is unchanged (each chatSub has
  its own writer goroutine and bounded buffer).
- chat_permissions.go / chat_commands.go: an approval is reserved by the first
  answer and confirmed once delivered; the first answer wins, is recorded with
  the answering client's label, and the resolution is pushed to every watching
  client as chat_permission with resolved/decision/resolved_by so their cards
  clear. A later answer is refused with an error naming the client that answered,
  and never reaches the agent twice; a reservation whose reply failed is
  released.
- chat_hub.go / chat_sessions.go / chat.go: the hub now owns rooms and the
  user-turn id sequence; the resulting user turns are retained (role and client
  carried) so a reconnect re-syncs who said what. chat_sessions no longer marks a
  session Busy and its clientID parameter is gone; the protocol keeps the Busy
  field for wire compatibility only. websocket.go detaches a connection's chat
  subscriptions when the read loop ends.
- chat_claims.go is removed, not left vestigial: the only exclusive thing left is
  a turn, and the room's queue is that.

App (android):
- PipelinesRepository carries a stable per-device client label (Build.MODEL, with
  a fallback) on prompts and replies, and adds chatSubscribe; AppViewModel
  subscribes on opening a chat (best-effort, an older server's error frame does
  not stop history loading) and records the queued ack.
- ChatStore folds a user frame from another client into the transcript once,
  attributed by the sender's label, and never receives its own prompt back, so
  the sender's own message is never duplicated. ChatThread.readOnly is gone:
  canSend no longer depends on another client, the composer says "Queued - it
  will send when the current turn finishes." instead of disabling input, and the
  status label is Running/Idle only. Scroll/follow behaviour is untouched.
- Tests: ChatProtocolTest round-trips the new wire fields, ChatFormatTest asserts
  the stale Busy flag cannot produce a read-only label, MultiChatStateTest asserts
  a session another client watches stays writable.

Reflected where it belongs: no packaged template changes (no bundled starter
embeds the chat protocol). Checked the packaged authoring skill references: none
of them describes the chat wire beyond operations.md, and
internal/author/skill/references/operations.md is now stale - its "Brokered
OpenCode chat" section still says chat_sessions marks "busy" when another client
holds the session and that a second client's chat_prompt is rejected, and it does
not mention chat_subscribe, the per-session prompt queue/chat_queued, the
per-client label or the shared exactly-once approval (it also still describes the
per-subscription permission poller, which the room's single poller replaced). The
supported semantics are the ones in docs/cli.md, updated in this commit; the
operations.md section needs a rewrite to match. No README or other shared
document was edited here.

Checks run on the committed tree: gofmt -l internal cmd examples (nothing), go
build ./..., go vet ./internal/... ./cmd/..., go test ./internal/server
./internal/protocol ./internal/stage ./internal/pipeline.
2026-10-18 00:09:14 +01:00

323 lines
12 KiB
Go

package server
import (
"context"
"log"
"net/http"
"strings"
"sync"
"time"
"github.com/coder/websocket"
"github.com/coder/websocket/wsjson"
"git.a-wan.net/alan/pipelines/examples"
"git.a-wan.net/alan/pipelines/internal/buildinfo"
"git.a-wan.net/alan/pipelines/internal/pipeline"
"git.a-wan.net/alan/pipelines/internal/protocol"
"git.a-wan.net/alan/pipelines/internal/summarise"
)
// Options configures the HTTP handler.
type Options struct {
// ArtifactsRoot overrides where stage artifacts live. Empty uses the store's
// directory (the database's parent), matching the data-directory default.
ArtifactsRoot string
// Chat brokers OpenCode chat traffic over /connect. Nil answers every
// chat_* request with a protocol error; a configured server passes an
// *opencode.Chat bound to the owned endpoint.
Chat chatBroker
// OperatorRegistry is the operator-session registry chat_sessions lists and
// chat_create/chat_directories read. When nil, HandlerWithOptions derives it
// from the first store; when neither is available chat_sessions returns an
// empty list (never the endpoint's whole session list) and chat_create is
// refused. Purely pipeline stage ownership is unaffected.
OperatorRegistry operatorRegistry
// RunTitles generates the operator-facing title stored with each new run.
// When nil, run creation stores the prompt-derived fallback without a model
// call, which keeps tests and unconfigured servers deterministic.
RunTitles func(context.Context, string) summarise.Title
// StepVerdicts generates the bounded one-line reason stored with each
// completed stage visit. When nil, the store derives the verdict's reason at
// read time with no model call, which keeps tests and unconfigured servers
// deterministic.
StepVerdicts func(context.Context, map[string]any) string
}
// Handler serves the persistent connection with default options. Canceling ctx
// stops active clients.
func Handler(ctx context.Context, stores ...*pipeline.Store) http.Handler {
return HandlerWithOptions(ctx, Options{}, stores...)
}
// HandlerWithOptions serves the persistent connection and the stage-artifact
// endpoint. Canceling ctx stops active clients.
func HandlerWithOptions(ctx context.Context, options Options, stores ...*pipeline.Store) http.Handler {
build := buildinfo.Current()
started := time.Now()
hub := newChatHub()
events := newRunEventHub()
mux := http.NewServeMux()
var runs *runner
var runErr error
if len(stores) > 0 {
runs, runErr = newRunner(ctx, stores[0])
if runErr == nil {
runs.events = events
}
if runErr == nil && options.ArtifactsRoot != "" {
runs.artifactsRoot = options.ArtifactsRoot
}
if runErr == nil && options.RunTitles != nil {
runs.title = options.RunTitles
}
if runErr == nil && options.StepVerdicts != nil {
runs.verdict = options.StepVerdicts
}
if runErr == nil {
runs.schedulerWG.Add(1)
go runs.scheduleLoop()
}
}
// chat_sessions lists the operator's own recorded sessions. Prefer an
// injected registry; otherwise derive it from the store. With neither, the
// handler fails closed to an empty list.
registry := options.OperatorRegistry
if registry == nil && len(stores) > 0 {
registry = stores[0]
}
hub.baseCtx = ctx
hub.registry = registry
mux.HandleFunc("GET /artifacts/{run}/{step}/{name...}", func(w http.ResponseWriter, r *http.Request) {
if runs == nil {
http.NotFound(w, r)
return
}
runs.serveArtifact(w, r)
})
mux.HandleFunc("GET /healthz", func(w http.ResponseWriter, r *http.Request) {
w.Header().Set("Content-Type", "application/json")
if runErr != nil {
w.WriteHeader(http.StatusServiceUnavailable)
_, _ = w.Write([]byte(`{"healthy":false}`))
return
}
_, _ = w.Write([]byte(`{"healthy":true}`))
})
mux.HandleFunc("GET /connect", func(w http.ResponseWriter, r *http.Request) {
conn, err := websocket.Accept(w, r, nil)
if err != nil {
return
}
defer conn.CloseNow()
conn.SetReadLimit(2 << 20)
// connCtx is canceled when this connection's read loop returns, so a
// chat subscription started for the client stops on disconnect. A single
// write mutex serializes the read loop's responses and the chat stream
// goroutine, which must never interleave partial websocket writes.
connCtx, connCancel := context.WithCancel(ctx)
defer connCancel()
var writeMu sync.Mutex
write := func(message protocol.Message) error {
// Every server frame (responses, the welcome and chat events) carries
// the build identity and protocol so an older client fails fast or
// warns once instead of silently misreading a field. build is
// computed once per handler.
if message.Build == nil {
message.Build = &build
}
if message.Protocol == 0 {
message.Protocol = protocol.ProtocolVersion
}
writeMu.Lock()
defer writeMu.Unlock()
writeCtx, cancel := context.WithTimeout(connCtx, 5*time.Second)
defer cancel()
return wsjson.Write(writeCtx, conn, message)
}
// clientID is this connection's identity for chat session claims: a
// second client is refused a session another connection holds.
clientID := chatClientToken()
// Drop every run subscription this connection holds when the read loop
// returns, so no subscriber, goroutine or timer outlives the connection.
defer events.unsubscribeClient(clientID)
// Drop every chat subscription too, so a chat room held only by this
// connection releases its broker stream and turn runner.
defer hub.detachClient(clientID)
if len(stores) > 0 {
items, err := stores[0].List(ctx)
if err != nil {
log.Printf("list pipelines: %v", err)
return
}
if err := write(protocol.Message{Type: protocol.PipelineList, Pipelines: items, NotificationEndpoints: notificationEndpoints()}); err != nil {
return
}
}
for {
var request protocol.Message
if err := wsjson.Read(ctx, conn, &request); err != nil {
return
}
log.Printf("command received: client=%s type=%q id=%q pipeline=%d version=%d", r.RemoteAddr, request.Type, request.ID, request.PipelineID, request.Version)
if handleChat(connCtx, request, options.Chat, hub, registry, clientID, write) {
continue
}
if handleRunEvents(request, events, runs, clientID, write) {
continue
}
response := protocol.Message{Type: protocol.Pong, ID: request.ID}
if request.Type == protocol.List || request.Type == protocol.Submit {
response.NotificationEndpoints = notificationEndpoints()
}
if request.Type == protocol.GetStatus {
response.Type = protocol.Status
response.Status = systemStatus(r, started, runErr)
} else if request.Type == protocol.Ping {
response.Time = time.Now().UTC().Format(time.RFC3339)
} else if request.Type == protocol.ListTemplates {
response.Type = protocol.Templates
response.Templates = examples.Catalog()
model, backend := starterDefaults()
for i := range response.Templates {
response.Templates[i].Source = pipeline.WithDefaults(response.Templates[i].Source, model, backend)
}
} else if request.Type == protocol.Validate {
graph, err := pipeline.Parse(request.Source)
response.Type = protocol.Validated
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
} else {
response.Graph = &graph
}
} else if request.Type == protocol.Submit && len(stores) > 0 {
response.Type = protocol.Submitted
response.Name = strings.TrimSpace(request.Name)
version, err := stores[0].SubmitChecked(ctx, request.Name, request.Source, request.CreateOnly, request.ExpectedVersion)
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
} else {
response.Version = version
response.Pipelines, err = stores[0].List(ctx)
if err != nil {
log.Printf("list pipelines after submission: %v", err)
}
for _, item := range response.Pipelines {
if item.Name == response.Name {
response.Detail, err = stores[0].Get(ctx, item.ID, version)
if err != nil {
log.Printf("load submitted pipeline: %v", err)
}
break
}
}
}
} else if request.Type == protocol.GetPipeline && len(stores) > 0 {
response.Type = protocol.Pipeline
detail, err := stores[0].Get(ctx, request.PipelineID, request.Version)
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
} else {
response.Detail = detail
}
} else if request.Type == protocol.List && len(stores) > 0 {
response.Type = protocol.PipelineList
response.DefaultModel, response.DefaultHarness = starterDefaults()
items, err := stores[0].List(ctx)
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
} else {
response.Pipelines = items
}
} else if request.Type == protocol.Search {
if len(stores) == 0 {
response.Type = protocol.Error
response.Error = "search unavailable: no store"
} else if request.Search == nil {
response.Type = protocol.Error
response.Error = "search requires a query"
} else {
response.Type = protocol.SearchResults
var err error
response.SearchResults, err = stores[0].Search(ctx, *request.Search, time.Now())
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
}
}
} else if runs != nil && (request.Type == protocol.RunPipeline || request.Type == protocol.RetryRun || request.Type == protocol.ListRuns || request.Type == protocol.GetRun || request.Type == protocol.CancelRun || request.Type == protocol.DecideRun || request.Type == protocol.SetRunTitle || request.Type == protocol.GetRunEvidence) {
var err error
switch request.Type {
case protocol.GetRunEvidence:
response.Type = protocol.RunEvidence
response.Evidence, err = runs.evidence(ctx, request.RunID)
case protocol.RunPipeline:
response.Type = protocol.RunStarted
response.RunDetail, err = runs.start(request)
case protocol.RetryRun:
response.Type = protocol.RunStarted
response.RunDetail, err = runs.retry(request)
case protocol.DecideRun:
response.Type = protocol.Run
response.RunDetail, err = runs.decide(request)
case protocol.SetRunTitle:
response.Type = protocol.Run
response.RunDetail, err = runs.setTitle(request.RunID, request.Title)
case protocol.ListRuns:
response.Type = protocol.Runs
response.Runs, err = runs.listRuns()
if err == nil {
response.Usage, err = stores[0].UsageStats(ctx)
}
case protocol.GetRun:
response.Type = protocol.Run
response.RunDetail, err = runs.getRun(request.RunID)
case protocol.CancelRun:
response.Type = protocol.Run
err = runs.cancel(request.RunID)
if err == nil {
response.RunDetail, err = runs.getRun(request.RunID)
}
}
if err != nil {
response.Type = protocol.Error
response.Error = err.Error()
}
} else {
response.Type = protocol.Error
response.Error = "unknown command: " + request.Type
if runErr != nil {
response.Error = "execution unavailable: " + runErr.Error()
}
}
if err := write(response); err != nil {
log.Printf("write response: %v", err)
return
}
selectedVersion := response.Version
if response.Detail != nil {
selectedVersion = response.Detail.Version
}
log.Printf("response sent: client=%s type=%q id=%q time=%q error=%q version=%d", r.RemoteAddr, response.Type, response.ID, response.Time, response.Error, selectedVersion)
}
})
return &runHandler{Handler: mux, runner: runs}
}
type runHandler struct {
http.Handler
runner *runner
}
// WaitRuns drains active execution before the database is closed on shutdown.
func WaitRuns(handler http.Handler) {
if h, ok := handler.(*runHandler); ok && h.runner != nil {
h.runner.schedulerWG.Wait()
h.runner.wg.Wait()
}
}