refactor: make last_error nullable instead of NOT NULL DEFAULT ''

Adds a last_error column to the chats table so that when a chat
transitions to error status, the error reason is stored in the
database and returned in the Chat API payload. Previously the error
was only published as an ephemeral SSE stream event and lost if the
client wasn't connected.

The UpdateChatStatus query now accepts a last_error parameter. All
non-error status transitions pass an empty string, which clears any
previous error. The two error paths (panic recovery and runChat
failure) capture the error message into lastError before persisting.

The SDK Chat struct gains a LastError field (omitted from JSON when
empty) and convertChat maps it from the database row.

coderd/chatd/chatd.go

@@ -478,6 +478,7 @@ func (p *Server) EditMessage(
 			WorkerID:    uuid.NullUUID{},
 			StartedAt:   sql.NullTime{},
 			HeartbeatAt: sql.NullTime{},
+			LastError:   sql.NullString{},
 		})
 		if err != nil {
 			return xerrors.Errorf("set chat pending: %w", err)
@@ -739,6 +740,7 @@ func setChatPendingWithStore(
 		WorkerID:    uuid.NullUUID{},
 		StartedAt:   sql.NullTime{},
 		HeartbeatAt: sql.NullTime{},
+		LastError:   sql.NullString{},
 	})
 	if err != nil {
 		return database.Chat{}, xerrors.Errorf("set chat pending: %w", err)
@@ -753,6 +755,7 @@ func (p *Server) setChatWaiting(ctx context.Context, chatID uuid.UUID) (database
 		WorkerID:    uuid.NullUUID{},
 		StartedAt:   sql.NullTime{},
 		HeartbeatAt: sql.NullTime{},
+		LastError:   sql.NullString{},
 	})
 	if err != nil {
 		return database.Chat{}, err
@@ -810,6 +813,7 @@ func insertUserMessageAndSetPending(
 		WorkerID:    uuid.NullUUID{},
 		StartedAt:   sql.NullTime{},
 		HeartbeatAt: sql.NullTime{},
+		LastError:   sql.NullString{},
 	})
 	if err != nil {
 		return database.ChatMessage{}, database.Chat{}, xerrors.Errorf("set chat pending: %w", err)
@@ -1622,8 +1626,9 @@ func (p *Server) processChat(ctx context.Context, chat database.Chat) {
 		Valid: true,
 	})
 
-	// Determine the final status to set when we're done.
+	// Determine the final status and last error to set when we're done.
 	status := database.ChatStatusWaiting
+	lastError := ""
 	remainingQueuedMessages := []database.ChatQueuedMessage{}
 	shouldPublishQueueUpdate := false
 
@@ -1636,7 +1641,8 @@ func (p *Server) processChat(ctx context.Context, chat database.Chat) {
 		// Handle panics gracefully.
 		if r := recover(); r != nil {
 			logger.Error(cleanupCtx, "panic during chat processing", slog.F("panic", r))
-			p.publishError(chat.ID, panicFailureReason(r))
+			lastError = panicFailureReason(r)
+			p.publishError(chat.ID, lastError)
 			status = database.ChatStatusError
 		}
 
@@ -1707,6 +1713,7 @@ func (p *Server) processChat(ctx context.Context, chat database.Chat) {
 				WorkerID:    uuid.NullUUID{},
 				StartedAt:   sql.NullTime{},
 				HeartbeatAt: sql.NullTime{},
+				LastError:   sql.NullString{String: lastError, Valid: lastError != ""},
 			})
 			return updateErr
 		}, nil)
@@ -1746,7 +1753,8 @@ func (p *Server) processChat(ctx context.Context, chat database.Chat) {
 		}
 		logger.Error(ctx, "failed to process chat", slog.Error(err))
 		if reason, ok := processingFailureReason(err); ok {
-			p.publishError(chat.ID, reason)
+			lastError = reason
+			p.publishError(chat.ID, lastError)
 		}
 		status = database.ChatStatusError
 		return
@@ -2458,6 +2466,7 @@ func (p *Server) recoverStaleChats(ctx context.Context) {
 			WorkerID:    uuid.NullUUID{},
 			StartedAt:   sql.NullTime{},
 			HeartbeatAt: sql.NullTime{},
+			LastError:   sql.NullString{},
 		})
 		if err != nil {
 			p.logger.Error(ctx, "failed to recover stale chat",

coderd/chatd/chatd_test.go

@@ -577,6 +577,61 @@ func TestWaitingChatsAreNotRecoveredAsStale(t *testing.T) {
 		"waiting chat should not be modified by stale recovery")
 }
 
+func TestUpdateChatStatusPersistsLastError(t *testing.T) {
+	t.Parallel()
+
+	db, ps := dbtestutil.NewDB(t)
+	_ = newTestServer(t, db, ps, uuid.New())
+
+	ctx := testutil.Context(t, testutil.WaitLong)
+	user, model := seedChatDependencies(ctx, t, db)
+
+	chat, err := db.InsertChat(ctx, database.InsertChatParams{
+		OwnerID:           user.ID,
+		Title:             "error-persisted",
+		LastModelConfigID: model.ID,
+	})
+	require.NoError(t, err)
+
+	// Simulate a chat that failed with an error.
+	errorMessage := "stream response: status 500: internal server error"
+	chat, err = db.UpdateChatStatus(ctx, database.UpdateChatStatusParams{
+		ID:          chat.ID,
+		Status:      database.ChatStatusError,
+		WorkerID:    uuid.NullUUID{},
+		StartedAt:   sql.NullTime{},
+		HeartbeatAt: sql.NullTime{},
+		LastError:   sql.NullString{String: errorMessage, Valid: true},
+	})
+	require.NoError(t, err)
+	require.Equal(t, database.ChatStatusError, chat.Status)
+	require.Equal(t, sql.NullString{String: errorMessage, Valid: true}, chat.LastError)
+
+	// Verify the error is persisted when re-read from the database.
+	fromDB, err := db.GetChatByID(ctx, chat.ID)
+	require.NoError(t, err)
+	require.Equal(t, database.ChatStatusError, fromDB.Status)
+	require.Equal(t, sql.NullString{String: errorMessage, Valid: true}, fromDB.LastError)
+
+	// Verify the error is cleared when the chat transitions to a
+	// non-error status (e.g. pending after a retry).
+	chat, err = db.UpdateChatStatus(ctx, database.UpdateChatStatusParams{
+		ID:          chat.ID,
+		Status:      database.ChatStatusPending,
+		WorkerID:    uuid.NullUUID{},
+		StartedAt:   sql.NullTime{},
+		HeartbeatAt: sql.NullTime{},
+		LastError:   sql.NullString{},
+	})
+	require.NoError(t, err)
+	require.Equal(t, database.ChatStatusPending, chat.Status)
+	require.False(t, chat.LastError.Valid)
+
+	fromDB, err = db.GetChatByID(ctx, chat.ID)
+	require.NoError(t, err)
+	require.False(t, fromDB.LastError.Valid)
+}
+
 func newTestServer(
 	t *testing.T,
 	db database.Store,

coderd/chats.go

@@ -795,6 +795,7 @@ func (api *API) interruptChat(rw http.ResponseWriter, r *http.Request) {
 			WorkerID:    uuid.NullUUID{},
 			StartedAt:   sql.NullTime{},
 			HeartbeatAt: sql.NullTime{},
+			LastError:   sql.NullString{},
 		})
 		if updateErr != nil {
 			api.Logger.Error(ctx, "failed to mark chat as waiting",
@@ -2059,6 +2060,9 @@ func convertChat(c database.Chat, diffStatus *database.ChatDiffStatus) codersdk.
 		CreatedAt:         c.CreatedAt,
 		UpdatedAt:         c.UpdatedAt,
 	}
+	if c.LastError.Valid {
+		chat.LastError = &c.LastError.String
+	}
 	if c.ParentChatID.Valid {
 		parentChatID := c.ParentChatID.UUID
 		chat.ParentChatID = &parentChatID

coderd/database/dump.sql

@@ -1274,7 +1274,8 @@ CREATE TABLE chats (
     parent_chat_id uuid,
     root_chat_id uuid,
     last_model_config_id uuid NOT NULL,
-    archived boolean DEFAULT false NOT NULL
+    archived boolean DEFAULT false NOT NULL,
+    last_error text
 );
 
 CREATE TABLE connection_logs (

coderd/database/migrations/000424_chat_last_error.down.sql

@@ -0,0 +1 @@
+ALTER TABLE chats DROP COLUMN last_error;

coderd/database/migrations/000424_chat_last_error.up.sql

@@ -0,0 +1 @@
+ALTER TABLE chats ADD COLUMN last_error TEXT;

coderd/database/models.go

@@ -3886,21 +3886,22 @@ type BoundaryUsageStat struct {
 }
 
 type Chat struct {
-	ID                uuid.UUID     `db:"id" json:"id"`
-	OwnerID           uuid.UUID     `db:"owner_id" json:"owner_id"`
-	WorkspaceID       uuid.NullUUID `db:"workspace_id" json:"workspace_id"`
-	WorkspaceAgentID  uuid.NullUUID `db:"workspace_agent_id" json:"workspace_agent_id"`
-	Title             string        `db:"title" json:"title"`
-	Status            ChatStatus    `db:"status" json:"status"`
-	WorkerID          uuid.NullUUID `db:"worker_id" json:"worker_id"`
-	StartedAt         sql.NullTime  `db:"started_at" json:"started_at"`
-	HeartbeatAt       sql.NullTime  `db:"heartbeat_at" json:"heartbeat_at"`
-	CreatedAt         time.Time     `db:"created_at" json:"created_at"`
-	UpdatedAt         time.Time     `db:"updated_at" json:"updated_at"`
-	ParentChatID      uuid.NullUUID `db:"parent_chat_id" json:"parent_chat_id"`
-	RootChatID        uuid.NullUUID `db:"root_chat_id" json:"root_chat_id"`
-	LastModelConfigID uuid.UUID     `db:"last_model_config_id" json:"last_model_config_id"`
-	Archived          bool          `db:"archived" json:"archived"`
+	ID                uuid.UUID      `db:"id" json:"id"`
+	OwnerID           uuid.UUID      `db:"owner_id" json:"owner_id"`
+	WorkspaceID       uuid.NullUUID  `db:"workspace_id" json:"workspace_id"`
+	WorkspaceAgentID  uuid.NullUUID  `db:"workspace_agent_id" json:"workspace_agent_id"`
+	Title             string         `db:"title" json:"title"`
+	Status            ChatStatus     `db:"status" json:"status"`
+	WorkerID          uuid.NullUUID  `db:"worker_id" json:"worker_id"`
+	StartedAt         sql.NullTime   `db:"started_at" json:"started_at"`
+	HeartbeatAt       sql.NullTime   `db:"heartbeat_at" json:"heartbeat_at"`
+	CreatedAt         time.Time      `db:"created_at" json:"created_at"`
+	UpdatedAt         time.Time      `db:"updated_at" json:"updated_at"`
+	ParentChatID      uuid.NullUUID  `db:"parent_chat_id" json:"parent_chat_id"`
+	RootChatID        uuid.NullUUID  `db:"root_chat_id" json:"root_chat_id"`
+	LastModelConfigID uuid.UUID      `db:"last_model_config_id" json:"last_model_config_id"`
+	Archived          bool           `db:"archived" json:"archived"`
+	LastError         sql.NullString `db:"last_error" json:"last_error"`
 }
 
 type ChatDiffStatus struct {

coderd/database/queries.sql.go

@@ -2842,7 +2842,7 @@ WHERE
             1
     )
 RETURNING
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 `
 
 type AcquireChatParams struct {
@@ -2871,6 +2871,7 @@ func (q *sqlQuerier) AcquireChat(ctx context.Context, arg AcquireChatParams) (Ch
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
@@ -2939,7 +2940,7 @@ func (q *sqlQuerier) DeleteChatQueuedMessage(ctx context.Context, arg DeleteChat
 
 const getChatByID = `-- name: GetChatByID :one
 SELECT
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 FROM
     chats
 WHERE
@@ -2965,12 +2966,13 @@ func (q *sqlQuerier) GetChatByID(ctx context.Context, id uuid.UUID) (Chat, error
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
 
 const getChatByIDForUpdate = `-- name: GetChatByIDForUpdate :one
-SELECT id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived FROM chats WHERE id = $1::uuid FOR UPDATE
+SELECT id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error FROM chats WHERE id = $1::uuid FOR UPDATE
 `
 
 func (q *sqlQuerier) GetChatByIDForUpdate(ctx context.Context, id uuid.UUID) (Chat, error) {
@@ -2992,6 +2994,7 @@ func (q *sqlQuerier) GetChatByIDForUpdate(ctx context.Context, id uuid.UUID) (Ch
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
@@ -3288,7 +3291,7 @@ func (q *sqlQuerier) GetChatQueuedMessages(ctx context.Context, chatID uuid.UUID
 
 const getChatsByOwnerID = `-- name: GetChatsByOwnerID :many
 SELECT
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 FROM
     chats
 WHERE
@@ -3323,6 +3326,7 @@ func (q *sqlQuerier) GetChatsByOwnerID(ctx context.Context, ownerID uuid.UUID) (
 			&i.RootChatID,
 			&i.LastModelConfigID,
 			&i.Archived,
+			&i.LastError,
 		); err != nil {
 			return nil, err
 		}
@@ -3339,7 +3343,7 @@ func (q *sqlQuerier) GetChatsByOwnerID(ctx context.Context, ownerID uuid.UUID) (
 
 const getStaleChats = `-- name: GetStaleChats :many
 SELECT
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 FROM
     chats
 WHERE
@@ -3374,6 +3378,7 @@ func (q *sqlQuerier) GetStaleChats(ctx context.Context, staleThreshold time.Time
 			&i.RootChatID,
 			&i.LastModelConfigID,
 			&i.Archived,
+			&i.LastError,
 		); err != nil {
 			return nil, err
 		}
@@ -3407,7 +3412,7 @@ INSERT INTO chats (
     $7::text
 )
 RETURNING
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 `
 
 type InsertChatParams struct {
@@ -3447,6 +3452,7 @@ func (q *sqlQuerier) InsertChat(ctx context.Context, arg InsertChatParams) (Chat
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
@@ -3572,7 +3578,7 @@ func (q *sqlQuerier) InsertChatQueuedMessage(ctx context.Context, arg InsertChat
 
 const listChatsByRootID = `-- name: ListChatsByRootID :many
 SELECT
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 FROM
     chats
 WHERE
@@ -3606,6 +3612,7 @@ func (q *sqlQuerier) ListChatsByRootID(ctx context.Context, rootChatID uuid.UUID
 			&i.RootChatID,
 			&i.LastModelConfigID,
 			&i.Archived,
+			&i.LastError,
 		); err != nil {
 			return nil, err
 		}
@@ -3622,7 +3629,7 @@ func (q *sqlQuerier) ListChatsByRootID(ctx context.Context, rootChatID uuid.UUID
 
 const listChildChatsByParentID = `-- name: ListChildChatsByParentID :many
 SELECT
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 FROM
     chats
 WHERE
@@ -3656,6 +3663,7 @@ func (q *sqlQuerier) ListChildChatsByParentID(ctx context.Context, parentChatID
 			&i.RootChatID,
 			&i.LastModelConfigID,
 			&i.Archived,
+			&i.LastError,
 		); err != nil {
 			return nil, err
 		}
@@ -3711,7 +3719,7 @@ SET
 WHERE
     id = $2::uuid
 RETURNING
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 `
 
 type UpdateChatByIDParams struct {
@@ -3738,6 +3746,7 @@ func (q *sqlQuerier) UpdateChatByID(ctx context.Context, arg UpdateChatByIDParam
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
@@ -3817,19 +3826,21 @@ SET
     worker_id = $2::uuid,
     started_at = $3::timestamptz,
     heartbeat_at = $4::timestamptz,
+    last_error = $5::text,
     updated_at = NOW()
 WHERE
-    id = $5::uuid
+    id = $6::uuid
 RETURNING
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 `
 
 type UpdateChatStatusParams struct {
-	Status      ChatStatus    `db:"status" json:"status"`
-	WorkerID    uuid.NullUUID `db:"worker_id" json:"worker_id"`
-	StartedAt   sql.NullTime  `db:"started_at" json:"started_at"`
-	HeartbeatAt sql.NullTime  `db:"heartbeat_at" json:"heartbeat_at"`
-	ID          uuid.UUID     `db:"id" json:"id"`
+	Status      ChatStatus     `db:"status" json:"status"`
+	WorkerID    uuid.NullUUID  `db:"worker_id" json:"worker_id"`
+	StartedAt   sql.NullTime   `db:"started_at" json:"started_at"`
+	HeartbeatAt sql.NullTime   `db:"heartbeat_at" json:"heartbeat_at"`
+	LastError   sql.NullString `db:"last_error" json:"last_error"`
+	ID          uuid.UUID      `db:"id" json:"id"`
 }
 
 func (q *sqlQuerier) UpdateChatStatus(ctx context.Context, arg UpdateChatStatusParams) (Chat, error) {
@@ -3838,6 +3849,7 @@ func (q *sqlQuerier) UpdateChatStatus(ctx context.Context, arg UpdateChatStatusP
 		arg.WorkerID,
 		arg.StartedAt,
 		arg.HeartbeatAt,
+		arg.LastError,
 		arg.ID,
 	)
 	var i Chat
@@ -3857,6 +3869,7 @@ func (q *sqlQuerier) UpdateChatStatus(ctx context.Context, arg UpdateChatStatusP
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }
@@ -3871,7 +3884,7 @@ SET
 WHERE
     id = $3::uuid
 RETURNING
-    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived
+    id, owner_id, workspace_id, workspace_agent_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
 `
 
 type UpdateChatWorkspaceParams struct {
@@ -3899,6 +3912,7 @@ func (q *sqlQuerier) UpdateChatWorkspace(ctx context.Context, arg UpdateChatWork
 		&i.RootChatID,
 		&i.LastModelConfigID,
 		&i.Archived,
+		&i.LastError,
 	)
 	return i, err
 }

coderd/database/queries/chats.sql

@@ -266,6 +266,7 @@ SET
     worker_id = sqlc.narg('worker_id')::uuid,
     started_at = sqlc.narg('started_at')::timestamptz,
     heartbeat_at = sqlc.narg('heartbeat_at')::timestamptz,
+    last_error = sqlc.narg('last_error')::text,
     updated_at = NOW()
 WHERE
     id = @id::uuid

codersdk/chats.go

@@ -38,6 +38,7 @@ type Chat struct {
 	LastModelConfigID uuid.UUID       `json:"last_model_config_id" format:"uuid"`
 	Title             string          `json:"title"`
 	Status            ChatStatus      `json:"status"`
+	LastError         *string         `json:"last_error"`
 	DiffStatus        *ChatDiffStatus `json:"diff_status,omitempty"`
 	CreatedAt         time.Time       `json:"created_at" format:"date-time"`
 	UpdatedAt         time.Time       `json:"updated_at" format:"date-time"`

site/src/api/typesGenerated.ts

@@ -1060,6 +1060,7 @@ export interface Chat {
 	readonly last_model_config_id: string;
 	readonly title: string;
 	readonly status: ChatStatus;
+	readonly last_error: string | null;
 	readonly diff_status?: ChatDiffStatus;
 	readonly created_at: string;
 	readonly updated_at: string;