fix(coderd): invalidate chatd provider cache on AI provider changes (#26987)

chatd subscribes to `ChatConfigEventChannel` and invalidates its
provider cache on a `providers` event kind, but nothing ever published
that kind. AI provider CRUD only publishes on
`AIProvidersChangedChannel` (consumed by aibridged and aibridgeproxyd),
so chatd's provider cache only converged via its 10 second TTL.

Subscribe chatd to the same `AIProvidersChangedChannel` publish instead
of adding a second publish, per the review feedback on #26207: one
publish, multiple subscribers. The now-unused `ChatConfigEventProviders`
kind is removed so `ChatConfigEvent` stays scoped to model configs, user
prompts, and advisor config, and can't regrow a dead subscriber.

Follow-up to CRF-5 from the review of #25673. Supersedes #26207.

Closes CODAGT-499

coderd/ai_providers.go

@@ -492,8 +492,8 @@ func (api *API) aiProvidersDelete(rw http.ResponseWriter, r *http.Request) {
 }
 
 // publishAIProvidersChanged notifies subscribers (aibridged,
-// aibridgeproxyd) that the live provider set changed and they should
-// refetch from the database. Pubsub failures are logged but not
+// aibridgeproxyd, chatd) that the live provider set changed and they
+// should refetch from the database. Pubsub failures are logged but not
 // propagated: subscribers refresh authoritatively from the DB, so a
 // dropped notification only delays convergence.
 func (api *API) publishAIProvidersChanged(ctx context.Context) {

coderd/pubsub/aiproviderschangedevent.go

@@ -2,8 +2,8 @@ package pubsub
 
 // AIProvidersChangedChannel is the pubsub channel that carries AI
 // provider lifecycle events: provider create / update / soft-delete
-// and key insert / delete. Subscribers (aibridged, aibridgeproxyd)
-// reload their in-memory provider snapshot on receipt.
+// and key insert / delete. Subscribers (aibridged, aibridgeproxyd,
+// chatd) reload their in-memory provider snapshot on receipt.
 //
 // The payload is an empty invalidation hint; subscribers refetch the
 // authoritative state from the database, so dropped messages only

coderd/pubsub/chatconfigevent.go

@@ -9,7 +9,7 @@ import (
 )
 
 // ChatConfigEventChannel is the pubsub channel for chat config
-// changes (providers, model configs, user prompts, advisor config).
+// changes (model configs, user prompts, advisor config).
 // All replicas subscribe to this channel to invalidate their local
 // caches.
 const ChatConfigEventChannel = "chat:config_change"
@@ -33,13 +33,11 @@ func HandleChatConfigEvent(cb func(ctx context.Context, payload ChatConfigEvent,
 }
 
 // ChatConfigEvent is published when chat configuration changes
-// (provider CRUD, model config CRUD, user prompt updates, or advisor
-// config updates). Subscribers use this to invalidate their local
-// caches.
+// (model config CRUD, user prompt updates, or advisor config
+// updates). Subscribers use this to invalidate their local caches.
 type ChatConfigEvent struct {
 	Kind ChatConfigEventKind `json:"kind"`
 	// EntityID carries context for the invalidation:
-	//   - For providers: uuid.Nil (all providers are invalidated).
 	//   - For model configs: the specific config ID.
 	//   - For user prompts: the user ID.
 	//   - For advisor config: uuid.Nil (singleton site-config row).
@@ -49,7 +47,6 @@ type ChatConfigEvent struct {
 type ChatConfigEventKind string
 
 const (
-	ChatConfigEventProviders     ChatConfigEventKind = "providers"
 	ChatConfigEventModelConfig   ChatConfigEventKind = "model_config"
 	ChatConfigEventUserPrompt    ChatConfigEventKind = "user_prompt"
 	ChatConfigEventAdvisorConfig ChatConfigEventKind = "advisor_config"

coderd/x/chatd/chatd.go

@@ -179,6 +179,7 @@ type Server struct {
 	debugSvcInit                   sync.Once
 	configCache                    *chatConfigCache
 	configCacheUnsubscribe         func()
+	providerCacheUnsubscribe       func()
 
 	usageTracker      *workspacestats.UsageTracker
 	clock             quartz.Clock
@@ -3300,8 +3301,6 @@ func New(ps pubsub.Pubsub, cfg Config) *Server {
 				return
 			}
 			switch ev.Kind {
-			case coderdpubsub.ChatConfigEventProviders:
-				p.configCache.InvalidateProviders()
 			case coderdpubsub.ChatConfigEventModelConfig:
 				p.configCache.InvalidateModelConfig(ev.EntityID)
 			case coderdpubsub.ChatConfigEventUserPrompt:
@@ -3317,6 +3316,22 @@ func New(ps pubsub.Pubsub, cfg Config) *Server {
 		p.configCacheUnsubscribe = cancelConfigSub
 	}
 
+	cancelProviderSub, err := p.pubsub.SubscribeWithErr(
+		coderdpubsub.AIProvidersChangedChannel,
+		func(cbCtx context.Context, _ []byte, err error) {
+			if err != nil {
+				p.logger.Warn(cbCtx, "ai providers changed event error", slog.Error(err))
+				return
+			}
+			p.configCache.InvalidateProviders()
+		},
+	)
+	if err != nil {
+		p.logger.Error(ctx, "subscribe to ai providers changed events", slog.Error(err))
+	} else {
+		p.providerCacheUnsubscribe = cancelProviderSub
+	}
+
 	p.ctx = ctx
 
 	// Spawn background goroutines that all servers need.
@@ -4903,6 +4918,10 @@ func (p *Server) Close() error {
 		p.configCacheUnsubscribe = nil
 		unsub()
 	}
+	if unsub := p.providerCacheUnsubscribe; unsub != nil {
+		p.providerCacheUnsubscribe = nil
+		unsub()
+	}
 	if p.chatWorker != nil {
 		if err := p.chatWorker.Close(); err != nil {
 			p.logger.Warn(context.Background(), "failed to close chat worker", slog.Error(err))

coderd/x/chatd/configcache_internal_test.go

@@ -14,6 +14,9 @@ import (
 	"golang.org/x/xerrors"
 
 	"github.com/coder/coder/v2/coderd/database"
+	"github.com/coder/coder/v2/coderd/database/dbtestutil"
+	coderdpubsub "github.com/coder/coder/v2/coderd/pubsub"
+	"github.com/coder/coder/v2/coderd/x/chatd/chatprovider"
 	"github.com/coder/coder/v2/codersdk"
 	"github.com/coder/coder/v2/testutil"
 	"github.com/coder/quartz"
@@ -1198,3 +1201,19 @@ func TestConfigCache_InvalidateAdvisorConfig_BlocksStaleInFlight(t *testing.T) {
 	require.EqualValues(t, 2, third.MaxUsesPerRun)
 	require.Equal(t, int32(2), store.advisorConfigCalls.Load())
 }
+
+func TestConfigCache_InvalidatesProvidersOnAIProvidersChangedEvent(t *testing.T) {
+	t.Parallel()
+
+	db, ps := dbtestutil.NewDB(t)
+	server := newInternalTestServer(t, db, ps, chatprovider.ProviderAPIKeys{})
+
+	// The generation counter only advances through InvalidateProviders,
+	// so this cannot false-pass via TTL expiry.
+	gen := server.configCache.providersGeneration()
+	require.NoError(t, ps.Publish(coderdpubsub.AIProvidersChangedChannel, nil))
+
+	require.Eventually(t, func() bool {
+		return server.configCache.providersGeneration() > gen
+	}, testutil.WaitShort, testutil.IntervalFast)
+}