Origin validation compared the request's Origin against the daemon's own direct connection: scheme from r.TLS, host from r.Host. Behind a TLS- terminating reverse proxy the browser sends Origin: https://host while the daemon sees a plain HTTP connection, so every dashboard mutation was refused as cross-origin. Taking X-Forwarded-Proto/-Host unconditionally would be worse than the bug: any client can set them, so an exposed daemon would let an attacker forge the scheme half of the CSRF check. They are honoured only when web.trust_forwarded_headers is explicitly enabled, which is off by default and documented as safe only behind a proxy that overwrites or strips untrusted forwarded headers. authorizeTaskMutation becomes a method because the decision now depends on server configuration; its three call sites move with it.
345 lines
14 KiB
Go
345 lines
14 KiB
Go
// Package webui is archied's dashboard: a control room over the daemon's
|
|
// task board, activity stream and configuration.
|
|
//
|
|
// The Go side owns serving and the JSON API only; the frontend lives in the
|
|
// repo-root ui/ package, which exposes the built assets as an embedded FS.
|
|
// Handlers are split by concern -- one file per API area -- so a section can
|
|
// be added without growing a single file without end.
|
|
package webui
|
|
|
|
import (
|
|
"context"
|
|
"encoding/json"
|
|
"io/fs"
|
|
"log/slog"
|
|
"net/http"
|
|
"sync"
|
|
"sync/atomic"
|
|
"time"
|
|
|
|
"github.com/samcharles93/archie-core/internal/channels"
|
|
"github.com/samcharles93/archie-core/internal/config"
|
|
"github.com/samcharles93/archie-core/internal/domain/workflow"
|
|
"github.com/samcharles93/archie-core/internal/events"
|
|
"github.com/samcharles93/archie-core/internal/gateway"
|
|
"github.com/samcharles93/archie-core/internal/logging"
|
|
"github.com/samcharles93/archie-core/internal/memory"
|
|
"github.com/samcharles93/archie-core/internal/store"
|
|
"github.com/samcharles93/archie-core/internal/webhookguard"
|
|
"github.com/samcharles93/archie-core/ui"
|
|
)
|
|
|
|
type Server struct {
|
|
Store store.TaskStore
|
|
Log *slog.Logger
|
|
|
|
// Cfg backs the setup checklist and configuration views. Optional: the
|
|
// dashboard degrades to task data alone when it is nil. Held through a
|
|
// Holder so a reload swaps the snapshot atomically across handler
|
|
// goroutines that would otherwise race on the raw shared pointer.
|
|
//
|
|
// In production this is the daemon's OWN Holder (main.go wires
|
|
// web.Cfg = d.Cfg), so Set here publishes to the running daemon. A
|
|
// config-mutating handler must never call Set directly: it goes
|
|
// through the same path as reload -- apply to a copy, validate the
|
|
// materialised config, persist, then Set.
|
|
Cfg *config.Holder
|
|
|
|
// Workflows is the executable registry snapshot supplied by composition.
|
|
Workflows []workflow.Definition
|
|
// WorkRequests admits dashboard requests through the same task-creation
|
|
// boundary used by chat; it never invokes a workflow runner directly.
|
|
WorkRequests gateway.TaskCreator
|
|
// LogFeed is the daemon diagnostic stream. It is separate from Events,
|
|
// which contains persisted task lifecycle activity only.
|
|
LogFeed *logging.Feed
|
|
// TaskLogs holds each task's persisted log output. Archiving a task
|
|
// removes its log files here too, via TaskLogs.Remove -- nil (task
|
|
// logging not configured) makes that a no-op.
|
|
TaskLogs *logging.TaskRegistry
|
|
|
|
// ConfigProvenance lists the files that produced the running config,
|
|
// in precedence order. Held through atomic.Pointer so a reload can
|
|
// swap it independently of Cfg. A reader may therefore observe the
|
|
// new config alongside the old provenance (or vice versa); this is
|
|
// display-only, and the skew is accepted rather than serialised.
|
|
ConfigProvenance atomic.Pointer[[]ConfigOrigin]
|
|
|
|
// LastReload reports the most recent config reload outcome, so the
|
|
// dashboard can tell the operator when the running config is stale
|
|
// (a reload failed validation and the daemon kept the previous
|
|
// config). Optional: when nil, the /api/config response omits the
|
|
// reload status.
|
|
LastReload func() config.ReloadStatus
|
|
|
|
// UpdateConfig applies a set of dotted-path config updates through
|
|
// the same validate-persist-publish path as reload (wired by the
|
|
// composition root). Optional: when nil, PATCH /api/config answers
|
|
// 503. The handler never touches the Cfg Holder directly -- see the
|
|
// Cfg field doc.
|
|
UpdateConfig func(context.Context, map[string]any) error
|
|
|
|
// ConfigOverrides lists the dotted config keys currently overridden
|
|
// by the runtime overlay, so the dashboard can mark those rows and
|
|
// offer a reset. Optional: when nil, the /api/config response omits
|
|
// the overridden list.
|
|
ConfigOverrides func(context.Context) ([]string, error)
|
|
|
|
// ResetConfig deletes one runtime-overlay row and republishes file +
|
|
// remaining overlay. Optional: when nil, POST /api/config/reset
|
|
// answers 503. Like UpdateConfig, it never touches the Cfg Holder
|
|
// directly.
|
|
ResetConfig func(context.Context, string) error
|
|
|
|
// Channels reports actual adapter lifecycle, independently of configuration
|
|
// presence. Nil preserves the configuration-only fallback for tests and
|
|
// minimal embedding.
|
|
Channels *channels.Manager
|
|
|
|
// ReloadChannel invokes a channel-specific reload seam. Only adapters that
|
|
// explicitly support reload are wired here; nil means reload is unavailable.
|
|
ReloadChannel func(context.Context, string) error
|
|
|
|
// Memory backs the memory view. Optional: the section reports memory as
|
|
// unavailable rather than failing when it is nil.
|
|
Memory *memory.Manager
|
|
|
|
// Chat exposes the same gateway router used by the conversational
|
|
// channels. Optional: the dashboard simply omits chat when it is nil.
|
|
Chat *ChatService
|
|
|
|
// Token gates access. Empty means no check, which is how loopback binds
|
|
// stay frictionless -- see IsLoopback.
|
|
Token string
|
|
|
|
// Issues closes the forge issue behind a rejected task. Optional: the
|
|
// action still records the operator's decision without it, and says so
|
|
// in the log. Deliberately the narrowest interface that does the job
|
|
// rather than the whole forge client -- the dashboard has no business
|
|
// commenting, labelling or opening PRs.
|
|
Issues IssueCloser
|
|
|
|
// Events publishes operator actions so they reach the task timeline and
|
|
// the live activity stream. Optional: nil means the action is recorded
|
|
// in the store but invisible to anyone watching.
|
|
Events EventPublisher
|
|
|
|
// TaskStopper reaches work that is actually executing. A store row cannot
|
|
// stop a goroutine or container; the daemon supplies this narrow runtime
|
|
// seam so the Stop action cancels work before parking its task record.
|
|
TaskStopper TaskStopper
|
|
|
|
// Captures persists unbound webhook captures (docs/prds/event-capture-storage.md).
|
|
// Optional: nil makes POST /webhooks/capture/{source} answer 503 rather
|
|
// than the dashboard failing to start -- capture is a precondition for
|
|
// the no-code playbook epic, not a hard dependency of the dashboard.
|
|
Captures store.CaptureStore
|
|
// CaptureLimiter is the per-remote-address token bucket applied before a
|
|
// capture request's body is read (keyed by remote address, not the
|
|
// unregistered/attacker-chosen source path segment -- see
|
|
// handleCapture). Optional: nil disables rate limiting (every request is
|
|
// allowed), which composition never actually does in production -- it
|
|
// is nil only in tests that don't exercise this path.
|
|
CaptureLimiter *webhookguard.RateLimiter
|
|
// CaptureRetention and CaptureMaxEvents are passed straight through to
|
|
// InsertCapture's prune-on-write bounds. See config.CaptureConfig.
|
|
CaptureRetention time.Duration
|
|
CaptureMaxEvents int
|
|
// CaptureMaxBodyBytes rejects (413) a capture POST body larger than
|
|
// this via http.MaxBytesReader, before redaction or storage sees it.
|
|
CaptureMaxBodyBytes int64
|
|
|
|
// Mappings persists payload field mappings (docs/prds/payload-field-mapping.md).
|
|
// Optional: nil makes every /api/mappings route answer 503 rather than
|
|
// the dashboard failing to start.
|
|
Mappings store.MappingStore
|
|
|
|
// UpdateReportPath is where the update watchdog leaves the phase-2
|
|
// outcome of a dashboard-initiated install for this process to relay on
|
|
// its next boot -- the webui counterpart of
|
|
// channels/telegram.Gateway.UpdateReportPath. Empty (the default unless
|
|
// composition wires it) means dashboard-initiated updates get no
|
|
// phase-2 report (archie-core-nln7): the operator only sees the
|
|
// synchronous install result, never restart/health/version outcome.
|
|
UpdateReportPath string
|
|
|
|
// RunningVersions reports, per component ID (see releaseupdate.Report.Verify),
|
|
// the version that component reports for itself right now. Optional:
|
|
// nil leaves every claim in a relayed update report Unverified rather
|
|
// than Confirmed.
|
|
RunningVersions func() map[string]string
|
|
|
|
// TrustForwardedHeaders controls whether X-Forwarded-Proto and
|
|
// X-Forwarded-Host are trusted when validating Origin on mutating
|
|
// requests. When false, Origin scheme must match the direct connection (r.TLS).
|
|
// When Cfg is present, Cfg.Get().Web.TrustForwardedHeaders is also checked.
|
|
TrustForwardedHeaders bool
|
|
|
|
mu sync.Mutex
|
|
conns map[chan events.Event]struct{}
|
|
}
|
|
|
|
func (s *Server) trustForwardedHeaders() bool {
|
|
if s.TrustForwardedHeaders {
|
|
return true
|
|
}
|
|
if s.Cfg == nil {
|
|
return false
|
|
}
|
|
return s.Cfg.Get().Web.TrustForwardedHeaders
|
|
}
|
|
|
|
// ConfigOrigin explains one source file contributing to the effective
|
|
// configuration. Sources are ordered from lowest to highest precedence.
|
|
type ConfigOrigin struct {
|
|
Path string `json:"path"`
|
|
Role string `json:"role"`
|
|
Layer string `json:"layer"`
|
|
Feature string `json:"feature,omitempty"`
|
|
}
|
|
|
|
// SetProvenance publishes a fresh provenance list after a config reload.
|
|
func (s *Server) SetProvenance(origins []ConfigOrigin) {
|
|
s.ConfigProvenance.Store(&origins)
|
|
}
|
|
|
|
// IssueCloser closes the forge issue behind a task.
|
|
type IssueCloser interface {
|
|
CloseIssue(ctx context.Context, owner, repo string, number int, comment string) error
|
|
}
|
|
|
|
// EventPublisher accepts events for the store and the live stream. The bus
|
|
// in cmd/archied satisfies it.
|
|
type EventPublisher interface {
|
|
Publish(events.Event)
|
|
}
|
|
|
|
// TaskStopper interrupts one task currently executing in the daemon.
|
|
type TaskStopper interface {
|
|
CancelTask(taskID int64) bool
|
|
}
|
|
|
|
// Broadcast fans an (ID-stamped) event out to every connected SSE
|
|
// client; stalled clients drop events rather than blocking the caller.
|
|
func (s *Server) Broadcast(e events.Event) {
|
|
s.mu.Lock()
|
|
defer s.mu.Unlock()
|
|
for c := range s.conns {
|
|
select {
|
|
case c <- e:
|
|
default:
|
|
}
|
|
}
|
|
}
|
|
|
|
func (s *Server) Handler() http.Handler {
|
|
mux := http.NewServeMux()
|
|
mux.HandleFunc("GET /api/summary", s.handleSummary)
|
|
mux.HandleFunc("GET /api/setup", s.handleSetup)
|
|
mux.HandleFunc("GET /api/tasks", s.handleTasks)
|
|
mux.HandleFunc("POST /api/tasks/{id}/action", s.handleTaskAction)
|
|
mux.HandleFunc("GET /api/tasks/{id}", s.handleTask)
|
|
mux.HandleFunc("GET /api/workflows", s.handleWorkflows)
|
|
mux.HandleFunc("POST /api/work-requests", s.handleWorkRequest)
|
|
mux.HandleFunc("GET /api/skills", s.handleSkills)
|
|
mux.HandleFunc("GET /api/captures", s.handleCaptures)
|
|
mux.HandleFunc("GET /api/mappings", s.handleMappingsList)
|
|
mux.HandleFunc("POST /api/mappings", s.handleMappingCreate)
|
|
mux.HandleFunc("GET /api/mappings/{id}", s.handleMappingGet)
|
|
mux.HandleFunc("PATCH /api/mappings/{id}", s.handleMappingUpdate)
|
|
mux.HandleFunc("DELETE /api/mappings/{id}", s.handleMappingDelete)
|
|
mux.HandleFunc("POST /api/mappings/preview", s.handleMappingPreview)
|
|
mux.HandleFunc("GET /api/channels", s.handleChannels)
|
|
mux.HandleFunc("POST /api/channels/{id}/reload", s.handleChannelReload)
|
|
mux.HandleFunc("GET /api/config", s.handleConfig)
|
|
mux.HandleFunc("GET /api/version", s.handleVersion)
|
|
mux.HandleFunc("PATCH /api/config", s.handleConfigUpdate)
|
|
mux.HandleFunc("POST /api/config/reset", s.handleConfigReset)
|
|
mux.HandleFunc("GET /api/logs", s.handleLogs)
|
|
mux.HandleFunc("GET /api/tasks/{id}/logs", s.handleTaskLogs)
|
|
mux.HandleFunc("GET /api/logs/stream", s.handleLogStream)
|
|
mux.HandleFunc("GET /api/memory", s.handleMemory)
|
|
mux.HandleFunc("GET /api/chat/sessions", s.handleChatSessions)
|
|
mux.HandleFunc("GET /api/chat/sessions/{id}/messages", s.handleChatMessages)
|
|
mux.HandleFunc("POST /api/chat/message", s.handleChatMessage)
|
|
mux.HandleFunc("POST /api/chat/stream", s.handleChatStream)
|
|
mux.HandleFunc("POST /api/chat/cancel", s.handleChatCancel)
|
|
mux.HandleFunc("POST /api/chat/persona", s.handleChatPersona)
|
|
mux.HandleFunc("GET /api/chat/update", s.handleChatUpdate)
|
|
mux.HandleFunc("POST /api/chat/update/defer", s.handleChatUpdateDefer)
|
|
mux.HandleFunc("POST /api/chat/update/install", s.handleChatUpdateInstall)
|
|
mux.HandleFunc("GET /api/chat/dangerous", s.handleChatDangerousState)
|
|
mux.HandleFunc("POST /api/chat/dangerous/{kind}", s.handleChatDangerousRequest)
|
|
mux.HandleFunc("POST /api/chat/dangerous/{id}/decision", s.handleChatDangerousDecision)
|
|
mux.HandleFunc("GET /events", s.handleSSE)
|
|
mux.Handle("GET /", s.assets())
|
|
|
|
top := http.NewServeMux()
|
|
top.HandleFunc("GET /healthz", s.handleHealthz)
|
|
top.HandleFunc("POST /webhooks/capture/{source}", s.handleCapture)
|
|
top.Handle("/", s.requireToken(mux))
|
|
return top
|
|
}
|
|
|
|
// handleHealthz is a liveness probe for local, unauthenticated callers --
|
|
// most notably the update watchdog script, which polls it after restarting
|
|
// archied to decide whether the new version came up or the update needs to
|
|
// be rolled back (see scripts/archie-update-watchdog). It deliberately
|
|
// bypasses requireToken: the token protects the dashboard from remote
|
|
// access, not this process's own local restart tooling.
|
|
func (s *Server) handleHealthz(w http.ResponseWriter, _ *http.Request) {
|
|
w.WriteHeader(http.StatusOK)
|
|
_, _ = w.Write([]byte("ok"))
|
|
}
|
|
|
|
// assets serves the embedded dashboard, falling back to index.html so client
|
|
// routes deep-link correctly. Under the no_ui build tag the assets are absent
|
|
// and a plain explanation is served instead of a confusing 404.
|
|
func (s *Server) assets() http.Handler {
|
|
if ui.DistDirFS == nil {
|
|
return http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
|
|
http.Error(w, "dashboard not built into this binary (no_ui)", http.StatusNotFound)
|
|
})
|
|
}
|
|
files := http.FileServerFS(ui.DistDirFS)
|
|
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
|
if _, err := fs.Stat(ui.DistDirFS, trimLeadingSlash(r.URL.Path)); err != nil {
|
|
r = r.Clone(r.Context())
|
|
r.URL.Path = "/"
|
|
}
|
|
files.ServeHTTP(w, r)
|
|
})
|
|
}
|
|
|
|
func trimLeadingSlash(p string) string {
|
|
if len(p) > 0 && p[0] == '/' {
|
|
return p[1:]
|
|
}
|
|
if p == "" {
|
|
return "."
|
|
}
|
|
return p
|
|
}
|
|
|
|
// Run serves until ctx ends.
|
|
func (s *Server) Run(ctx context.Context, listen string) error {
|
|
s.reportPendingUpdate(ctx)
|
|
srv := &http.Server{Addr: listen, Handler: s.Handler(), ReadHeaderTimeout: 5 * time.Second}
|
|
go func() {
|
|
<-ctx.Done()
|
|
shutdownCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 2*time.Second)
|
|
defer cancel()
|
|
_ = srv.Shutdown(shutdownCtx)
|
|
}()
|
|
s.Log.Info("web ui listening", "addr", "http://"+listen)
|
|
if err := srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func writeJSON(w http.ResponseWriter, v any) {
|
|
w.Header().Set("Content-Type", "application/json")
|
|
_ = json.NewEncoder(w).Encode(v)
|
|
}
|