feat: stream persistent

backend/internal/api/generation.go

@@ -0,0 +1,194 @@
+package api
+
+import (
+	"errors"
+	"sync"
+
+	"github.com/google/uuid"
+	"reichard.io/aethera/internal/store"
+	"reichard.io/aethera/internal/types"
+)
+
+var errGenerationActive = errors.New("generation already active")
+
+type generationManager struct {
+	mu          sync.RWMutex
+	generations map[uuid.UUID]*generation
+}
+
+type generation struct {
+	mu          sync.RWMutex
+	subscribers map[chan *MessageChunk]struct{}
+	done        chan struct{}
+	closed      bool
+}
+
+func newGenerationManager() *generationManager {
+	return &generationManager{generations: make(map[uuid.UUID]*generation)}
+}
+
+func (m *generationManager) start(chatID uuid.UUID, prepare func(*generation) error, run func(*generation)) error {
+	m.mu.Lock()
+	if _, found := m.generations[chatID]; found {
+		m.mu.Unlock()
+		return errGenerationActive
+	}
+
+	// Reserve Generation
+	gen := &generation{
+		subscribers: make(map[chan *MessageChunk]struct{}),
+		done:        make(chan struct{}),
+	}
+	m.generations[chatID] = gen
+	m.mu.Unlock()
+
+	// Prepare Generation - This runs while the generation is reserved so a
+	// concurrent request cannot persist duplicate user/assistant messages.
+	if err := prepare(gen); err != nil {
+		gen.close()
+		m.mu.Lock()
+		delete(m.generations, chatID)
+		m.mu.Unlock()
+		return err
+	}
+
+	// Run Generation
+	go func() {
+		defer func() {
+			gen.close()
+			m.mu.Lock()
+			delete(m.generations, chatID)
+			m.mu.Unlock()
+		}()
+		run(gen)
+	}()
+
+	return nil
+}
+
+func (m *generationManager) subscribe(chatID uuid.UUID) (<-chan *MessageChunk, func(), bool) {
+	m.mu.RLock()
+	gen, found := m.generations[chatID]
+	m.mu.RUnlock()
+	if !found {
+		return nil, func() {}, false
+	}
+
+	ch := gen.subscribe()
+	return ch, func() { gen.unsubscribe(ch) }, true
+}
+
+func (g *generation) subscribe() chan *MessageChunk {
+	ch := make(chan *MessageChunk, 64)
+
+	// Add Subscriber
+	g.mu.Lock()
+	if g.closed {
+		close(ch)
+	} else {
+		g.subscribers[ch] = struct{}{}
+	}
+	g.mu.Unlock()
+
+	return ch
+}
+
+func (g *generation) unsubscribe(ch chan *MessageChunk) {
+	// Remove Subscriber
+	g.mu.Lock()
+	if _, found := g.subscribers[ch]; found {
+		delete(g.subscribers, ch)
+		close(ch)
+	}
+	g.mu.Unlock()
+}
+
+func (g *generation) broadcast(chunk *MessageChunk) {
+	g.mu.RLock()
+	defer g.mu.RUnlock()
+
+	// Broadcast Chunk
+	for subscriber := range g.subscribers {
+		select {
+		case subscriber <- cloneMessageChunk(chunk):
+		default:
+		}
+	}
+}
+
+func (g *generation) close() {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+
+	// Close Subscribers
+	if g.closed {
+		return
+	}
+	g.closed = true
+	close(g.done)
+	for subscriber := range g.subscribers {
+		close(subscriber)
+		delete(g.subscribers, subscriber)
+	}
+}
+
+func cloneMessageChunk(chunk *MessageChunk) *MessageChunk {
+	if chunk == nil {
+		return nil
+	}
+
+	// Clone Chunk
+	cloned := &MessageChunk{
+		Chat:             chunk.Chat,
+		UserMessage:      cloneStoreMessage(chunk.UserMessage),
+		AssistantMessage: cloneStoreMessage(chunk.AssistantMessage),
+	}
+	return cloned
+}
+
+func cloneStoreMessage(msg *store.Message) *store.Message {
+	if msg == nil {
+		return nil
+	}
+
+	// Clone Message
+	cloned := *msg
+	if msg.Stats != nil {
+		stats := *msg.Stats
+		cloned.Stats = &stats
+		cloneMessageStatsPointers(msg.Stats, cloned.Stats)
+	}
+	return &cloned
+}
+
+func cloneMessageStatsPointers(src, dst *types.MessageStats) {
+	// Clone Pointer Fields
+	if src.EndTime != nil {
+		v := *src.EndTime
+		dst.EndTime = &v
+	}
+	if src.PromptTokens != nil {
+		v := *src.PromptTokens
+		dst.PromptTokens = &v
+	}
+	if src.GeneratedTokens != nil {
+		v := *src.GeneratedTokens
+		dst.GeneratedTokens = &v
+	}
+	if src.PromptPerSec != nil {
+		v := *src.PromptPerSec
+		dst.PromptPerSec = &v
+	}
+	if src.GeneratedPerSec != nil {
+		v := *src.GeneratedPerSec
+		dst.GeneratedPerSec = &v
+	}
+	if src.TimeToFirstToken != nil {
+		v := *src.TimeToFirstToken
+		dst.TimeToFirstToken = &v
+	}
+	if src.TimeToLastToken != nil {
+		v := *src.TimeToLastToken
+		dst.TimeToLastToken = &v
+	}
+}

backend/internal/api/handlers.go

@@ -28,6 +28,7 @@ type API struct {
 	store             store.Store
 	client            *client.Client
 	dataDir           string
+	generationManager *generationManager
 }
 
 func New(s store.Store, dataDir string, logger *logrus.Logger) *API {
@@ -35,6 +36,7 @@ func New(s store.Store, dataDir string, logger *logrus.Logger) *API {
 		store:             s,
 		dataDir:           dataDir,
 		logger:            logger.WithField("service", "api"),
+		generationManager: newGenerationManager(),
 	}
 }
 
@@ -322,13 +324,18 @@ func (a *API) PostChat(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	// Send Message
+	// Start Message
-	responseStarted, err := a.sendMessage(r.Context(), w, chat.ID, genReq.Model, genReq.Prompt)
+	chunk, err := a.startMessageGeneration(chat.ID, genReq.Model, genReq.Prompt)
 	if err != nil {
-		log.WithError(err).WithField("chat_id", chat.ID).Error("failed to send message")
+		log.WithError(err).WithField("chat_id", chat.ID).Error("failed to start message generation")
-		if !responseStarted {
+		http.Error(w, "Failed to start message generation", http.StatusInternalServerError)
-			http.Error(w, "Failed to send message", http.StatusInternalServerError)
+		return
 	}
+
+	w.Header().Set("Content-Type", "application/json")
+	if err := json.NewEncoder(w).Encode(chunk); err != nil {
+		log.WithError(err).Error("failed to encode message generation response")
+		http.Error(w, "Failed to encode message generation response", http.StatusInternalServerError)
 	}
 }
 
@@ -395,6 +402,68 @@ func (a *API) GetChat(w http.ResponseWriter, r *http.Request) {
 	}
 }
 
+func (a *API) GetChatStream(w http.ResponseWriter, r *http.Request) {
+	log := a.logger.WithField("handler", "GetChatStreamHandler")
+
+	// Parse Chat ID
+	rawChatID := r.PathValue("chatId")
+	if rawChatID == "" {
+		log.Error("missing chat ID parameter")
+		http.Error(w, "Chat ID is required", http.StatusBadRequest)
+		return
+	}
+	chatID, err := uuid.Parse(rawChatID)
+	if err != nil {
+		log.WithError(err).Error("invalid chat ID format")
+		http.Error(w, "Invalid chat ID format", http.StatusBadRequest)
+		return
+	}
+
+	// Subscribe Before Snapshot
+	updates, unsubscribe, active := a.generationManager.subscribe(chatID)
+	defer unsubscribe()
+
+	// Get Chat Snapshot
+	chat, err := a.store.GetChat(chatID)
+	if err != nil {
+		log.WithError(err).WithField("chat_id", chatID).Error("failed to get chat")
+		http.Error(w, "Failed to get chat", http.StatusInternalServerError)
+		return
+	}
+
+	// Set Headers
+	w.Header().Set("Content-Type", "application/x-ndjson")
+	w.Header().Set("Cache-Control", "no-cache")
+	w.Header().Set("Connection", "keep-alive")
+	w.Header().Set("Transfer-Encoding", "chunked")
+	flushWriter := newFlushWriter(w)
+
+	// Send Snapshot
+	if err := json.NewEncoder(flushWriter).Encode(&MessageChunk{Chat: toChat(chat)}); err != nil {
+		log.WithError(err).WithField("chat_id", chatID).Warn("failed to send stream snapshot")
+		return
+	}
+	if !active {
+		return
+	}
+
+	// Forward Updates
+	for {
+		select {
+		case <-r.Context().Done():
+			return
+		case chunk, ok := <-updates:
+			if !ok {
+				return
+			}
+			if err := json.NewEncoder(flushWriter).Encode(chunk); err != nil {
+				log.WithError(err).WithField("chat_id", chatID).Warn("client stream disconnected")
+				return
+			}
+		}
+	}
+}
+
 func (a *API) PostChatMessage(w http.ResponseWriter, r *http.Request) {
 	log := a.logger.WithField("handler", "PostChatMessageHandler")
 
@@ -424,13 +493,22 @@ func (a *API) PostChatMessage(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	// Send Message
+	// Start Message
-	responseStarted, err := a.sendMessage(r.Context(), w, chatID, genReq.Model, genReq.Prompt)
+	chunk, err := a.startMessageGeneration(chatID, genReq.Model, genReq.Prompt)
 	if err != nil {
-		log.WithError(err).WithField("chat_id", chatID).Error("failed to send message")
+		log.WithError(err).WithField("chat_id", chatID).Error("failed to start message generation")
-		if !responseStarted {
+		if errors.Is(err, errGenerationActive) {
-			http.Error(w, "Failed to send message", http.StatusInternalServerError)
+			http.Error(w, "Chat generation already active", http.StatusConflict)
+		} else {
+			http.Error(w, "Failed to start message generation", http.StatusInternalServerError)
 		}
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	if err := json.NewEncoder(w).Encode(chunk); err != nil {
+		log.WithError(err).Error("failed to encode message generation response")
+		http.Error(w, "Failed to encode message generation response", http.StatusInternalServerError)
 	}
 }
 
@@ -456,99 +534,99 @@ func (a *API) getClient() (*client.Client, error) {
 	return a.client, nil
 }
 
-func (a *API) sendMessage(ctx context.Context, w http.ResponseWriter, chatID uuid.UUID, chatModel, userMessage string) (bool, error) {
+func (a *API) startMessageGeneration(chatID uuid.UUID, chatModel, userMessage string) (*MessageChunk, error) {
 	apiClient, err := a.getClient()
 	if err != nil {
-		return false, fmt.Errorf("failed to get client: %w", err)
+		return nil, fmt.Errorf("failed to get client: %w", err)
 	}
 
-	// Detach Request Context
+	var chat *store.Chat
-	ctx, cancel := context.WithTimeout(context.WithoutCancel(ctx), time.Minute*5)
+	var userMsg *store.Message
-	defer cancel()
+	var assistantMsg *store.Message
+	var initialChunk *MessageChunk
 
+	// Start Generation - The manager reserves the chat before messages are
+	// persisted, preventing concurrent completions from creating duplicate rows.
+	if err := a.generationManager.start(chatID, func(_ *generation) error {
 		// Create User Message
-	userMsg := &store.Message{ChatID: chatID, Role: "user", Content: userMessage}
+		userMsg = &store.Message{ChatID: chatID, Role: "user", Content: userMessage}
 		if err := a.store.SaveChatMessage(userMsg); err != nil {
-		return false, fmt.Errorf("failed to add user message to chat: %w", err)
+			return fmt.Errorf("failed to add user message to chat: %w", err)
 		}
 
 		// Get Chat History - Fetch before creating the in-progress assistant message so the
 		// LLM request does not include an empty assistant response prefill.
-	chat, err := a.store.GetChat(chatID)
+		chat, err = a.store.GetChat(chatID)
 		if err != nil {
-		return false, fmt.Errorf("failed to get chat: %w", err)
+			return fmt.Errorf("failed to get chat: %w", err)
 		}
 
-	// Add Assistant Response - TODO: Ensure InProgress Flag?
+		// Add Assistant Response
-	assistantMsg := &store.Message{ChatID: chatID, Role: "assistant"}
+		assistantMsg = &store.Message{ChatID: chatID, Role: "assistant", Status: store.MessageStatusStreaming}
 		if err := a.store.SaveChatMessage(assistantMsg); err != nil {
-		return false, fmt.Errorf("failed to add assistant message to chat: %w", err)
+			return fmt.Errorf("failed to add assistant message to chat: %w", err)
 		}
 
-	// Set Headers
+		// Create Initial Chunk
-	w.Header().Set("Content-Type", "application/x-ndjson")
+		initialChunk = &MessageChunk{
-	w.Header().Set("Cache-Control", "no-cache")
-	w.Header().Set("Connection", "keep-alive")
-	w.Header().Set("Transfer-Encoding", "chunked")
-
-	// Create Flush Writer
-	flushWriter := newFlushWriter(w)
-
-	// Send Initial Chunk - User Message & Chat
-	if err := json.NewEncoder(flushWriter).Encode(&MessageChunk{
 			Chat:             toChatNoMessages(chat),
 			UserMessage:      userMsg,
-	}); err != nil {
+			AssistantMessage: assistantMsg,
-		return false, fmt.Errorf("failed to send initial chunk: %w", err)
 		}
-	responseStarted := true
+		return nil
-	streamToClient := true
+	}, func(gen *generation) {
+		a.runMessageGeneration(apiClient, chat, assistantMsg, chatModel, gen)
+	}); err != nil {
+		return nil, err
+	}
+
+	return initialChunk, nil
+}
+
+func (a *API) runMessageGeneration(apiClient *client.Client, chat *store.Chat, assistantMsg *store.Message, chatModel string, gen *generation) {
+	// Create Generation Context
+	ctx, cancel := context.WithTimeout(context.Background(), time.Minute*5)
+	defer cancel()
 
 	// Send Message
 	if _, err := apiClient.SendMessage(ctx, chat.Messages, chatModel, func(m *client.MessageChunk) error {
-		var apiMsgChunk MessageChunk
 		messageChanged := false
 
 		if m.Stats != nil {
 			messageChanged = true
 			assistantMsg.Stats = m.Stats
 		}
-
 		if m.Message != nil {
 			messageChanged = true
 			assistantMsg.Content += *m.Message
 		}
-
 		if m.Thinking != nil {
 			messageChanged = true
 			assistantMsg.Thinking += *m.Thinking
 		}
 
-		// Save Assistant Progress - Persist each streamed update so partial content
+		// Save And Broadcast Progress
-		// survives client disconnects or upstream stream failures.
 		if messageChanged {
 			if err := a.store.SaveChatMessage(assistantMsg); err != nil {
 				return fmt.Errorf("failed to save assistant progress: %w", err)
 			}
-			apiMsgChunk.AssistantMessage = assistantMsg
+			gen.broadcast(&MessageChunk{AssistantMessage: assistantMsg})
-		}
-
-		// Send Progress Chunk - If the browser disconnects, keep the detached
-		// generation running and continue saving streamed content to the store.
-		if streamToClient {
-			if err := json.NewEncoder(flushWriter).Encode(apiMsgChunk); err != nil {
-				streamToClient = false
-				a.logger.WithError(err).WithField("chat_id", chat.ID).Warn("client stream disconnected")
-			}
 		}
 
 		return nil
 	}); err != nil {
-		return responseStarted, fmt.Errorf("failed to generate text stream: %w", err)
+		assistantMsg.Status = store.MessageStatusFailed
+		if saveErr := a.store.SaveChatMessage(assistantMsg); saveErr != nil {
+			a.logger.WithError(saveErr).WithField("chat_id", chat.ID).Error("failed to save failed assistant message")
+		}
+		gen.broadcast(&MessageChunk{AssistantMessage: assistantMsg})
+		a.logger.WithError(err).WithField("chat_id", chat.ID).Error("failed to generate text stream")
+		return
 	}
 
 	// Summarize & Update Chat Title
 	if chat.Title == "" {
+		var err error
 		chat.Title, err = apiClient.CreateTitle(ctx, chat.Messages[0].Content, chatModel)
 		if err != nil {
 			a.logger.WithError(err).WithField("chat_id", chat.ID).Error("failed to create chat title")
@@ -557,20 +635,11 @@ func (a *API) sendMessage(ctx context.Context, w http.ResponseWriter, chatID uui
 		}
 	}
 
-	// Update Assistant Message
+	// Complete Assistant Message
+	assistantMsg.Status = store.MessageStatusComplete
 	if err := a.store.SaveChatMessage(assistantMsg); err != nil {
-		return responseStarted, fmt.Errorf("failed to save assistant message to chat: %w", err)
+		a.logger.WithError(err).WithField("chat_id", chat.ID).Error("failed to save assistant message")
+		return
 	}
-
+	gen.broadcast(&MessageChunk{Chat: toChatNoMessages(chat), AssistantMessage: assistantMsg})
-	// Send Final Chunk
-	if streamToClient {
-		if err := json.NewEncoder(flushWriter).Encode(&MessageChunk{
-			Chat:             toChatNoMessages(chat),
-			AssistantMessage: assistantMsg,
-		}); err != nil {
-			return responseStarted, fmt.Errorf("failed to send final chunk: %w", err)
-		}
-	}
-
-	return responseStarted, nil
 }

backend/internal/server/server.go

@@ -48,6 +48,7 @@ func StartServer(settingsStore store.Store, dataDir, staticDir, listenAddress st
 	mux.HandleFunc("GET /api/chats", api.GetChats)
 	mux.HandleFunc("POST /api/chats", api.PostChat)
 	mux.HandleFunc("GET /api/chats/{chatId}", api.GetChat)
+	mux.HandleFunc("GET /api/chats/{chatId}/stream", api.GetChatStream)
 	mux.HandleFunc("POST /api/chats/{chatId}", api.PostChatMessage)
 	mux.HandleFunc("DELETE /api/chats/{chatId}", api.DeleteChat)
 

backend/internal/store/types.go

@@ -28,6 +28,14 @@ type Chat struct {
 	Messages []*Message `json:"messages"`
 }
 
+type MessageStatus string
+
+const (
+	MessageStatusStreaming MessageStatus = "streaming"
+	MessageStatusComplete  MessageStatus = "complete"
+	MessageStatusFailed    MessageStatus = "failed"
+)
+
 type Message struct {
 	baseModel
 
@@ -35,5 +43,6 @@ type Message struct {
 	Role     string              `json:"role"`
 	Thinking string              `json:"thinking"`
 	Content  string              `json:"content"`
+	Status   MessageStatus       `json:"status,omitempty"`
 	Stats    *types.MessageStats `json:"stats,omitempty"`
 }

frontend/src/client.ts

@@ -90,13 +90,37 @@ export async function sendMessage(
   requestData: GenerateTextRequest,
   onChunk: (chunk: MessageChunk) => void,
 ) {
+  const initialChunk = await startMessage(chatId, requestData);
+  onChunk(initialChunk);
+
+  if (!initialChunk.chat) return;
+  return streamChatUpdates(initialChunk.chat.id, onChunk);
+}
+
+export async function startMessage(
+  chatId: string,
+  requestData: GenerateTextRequest,
+): Promise<MessageChunk> {
   const url = chatId ? `/api/chats/${chatId}` : '/api/chats';
   const response = await fetch(url, {
     method: 'POST',
     headers: { 'Content-Type': 'application/json' },
     body: JSON.stringify(requestData),
   });
+  const data = await response.json().catch(() => ({}));
 
+  if (!response.ok) {
+    throw new Error(readError(data) || `HTTP ${response.status}`);
+  }
+
+  return data;
+}
+
+export async function streamChatUpdates(
+  chatId: string,
+  onChunk: (chunk: MessageChunk) => void,
+) {
+  const response = await fetch(`/api/chats/${chatId}/stream`);
   return streamMessage(response, onChunk);
 }
 
@@ -133,6 +157,15 @@ export async function deleteChat(chatId: string): Promise<void> {
   }
 }
 
+function readError(data: unknown): string {
+  if (typeof data !== 'object' || data === null || !('error' in data)) {
+    return '';
+  }
+
+  const error = data.error;
+  return typeof error === 'string' ? error : '';
+}
+
 async function streamMessage(
   response: Response,
   onChunk: (chunk: MessageChunk) => void,

frontend/src/components/chatManager.ts

@@ -6,6 +6,7 @@ import {
   getSettings,
   getModels,
   sendMessage,
+  streamChatUpdates,
   getChatMessages,
   listChats,
   deleteChat,
@@ -41,6 +42,7 @@ Alpine.data('chatManager', () => ({
   selectedChatID: null as string | null,
   chatListOpen: false,
   loading: false,
+  activeStreamChatID: null as string | null,
 
   async init() {
     // Acquire Data
@@ -109,66 +111,27 @@ Alpine.data('chatManager', () => ({
       this.selectedChatID = IN_PROGRESS_UUID;
     }
 
-    // New User Message
+    // Add Optimistic User Message
-    let userMessage: Message = {
+    const currentChat: Chat = this.chats.find(
+      (c) => c.id === this.selectedChatID,
+    )!;
+    currentChat.messages.push({
       id: IN_PROGRESS_UUID,
       chat_id: this.selectedChatID,
       role: 'user',
       thinking: '',
       content: message,
       created_at: new Date().toISOString(),
-    };
+    });
-
-    // Get Chat
-    let currentChat: Chat = this.chats.find(
-      (c) => c.id === this.selectedChatID,
-    )!;
-
-    // Add User Message
-    currentChat.messages.push(userMessage);
     currentChat.message_count += 1;
 
-    // Assistant Message Placeholder
-    let assistantMessage: Message | undefined;
-
     try {
       await sendMessage(
         this.selectedChatID === IN_PROGRESS_UUID ? '' : this.selectedChatID,
         { model: this.selectedModel, prompt: message },
         (chunk: MessageChunk) => {
-          // Handle Chat
+          if (chunk.chat) this.activeStreamChatID = chunk.chat.id;
-          if (chunk.chat) {
+          this.applyMessageChunk(chunk);
-            Object.assign(currentChat, {
-              ...chunk.chat,
-              messages: currentChat.messages,
-            });
-            this.selectedChatID = chunk.chat.id;
-            this.updateHash(chunk.chat.id);
-          }
-
-          // Handle User Message
-          if (chunk.user_message) {
-            Object.assign(userMessage, chunk.user_message);
-          }
-
-          // Handle Assistant Message
-          if (chunk.assistant_message) {
-            if (!assistantMessage) {
-              assistantMessage = chunk.assistant_message;
-              currentChat.messages.push(assistantMessage);
-            } else {
-              const index = currentChat.messages.findIndex(
-                (m) => m.id === assistantMessage!.id,
-              );
-              if (index !== -1) {
-                currentChat.messages[index] = {
-                  ...assistantMessage,
-                  ...chunk.assistant_message,
-                };
-                currentChat.messages = [...currentChat.messages];
-              }
-            }
-          }
         },
       );
     } catch (err) {
@@ -176,9 +139,56 @@ Alpine.data('chatManager', () => ({
       this.error = parseError(err);
     } finally {
       this.loading = false;
+      this.activeStreamChatID = null;
     }
   },
 
+  applyMessageChunk(chunk: MessageChunk) {
+    // Handle Chat
+    if (chunk.chat) {
+      let chat = this.chats.find((c) => c.id === chunk.chat!.id);
+      if (!chat) chat = this.chats.find((c) => c.id === IN_PROGRESS_UUID);
+      if (!chat) {
+        chat = { ...chunk.chat, messages: chunk.chat.messages || [] };
+        this.chats.unshift(chat);
+      } else {
+        Object.assign(chat, chunk.chat);
+        chat.messages = chunk.chat.messages?.length
+          ? chunk.chat.messages
+          : chat.messages;
+      }
+      this.selectedChatID = chunk.chat.id;
+      this.updateHash(chunk.chat.id);
+    }
+
+    const chatID = chunk.chat?.id || this.selectedChatID;
+    const currentChat = this.chats.find((c) => c.id === chatID);
+    if (!currentChat) return;
+
+    // Handle Messages
+    if (chunk.user_message) this.upsertMessage(currentChat, chunk.user_message);
+    if (chunk.assistant_message)
+      this.upsertMessage(currentChat, chunk.assistant_message);
+  },
+
+  upsertMessage(chat: Chat, message: Message) {
+    // Upsert Message
+    const existingIndex = chat.messages.findIndex(
+      (m) =>
+        m.id === message.id ||
+        (m.id === IN_PROGRESS_UUID && m.role === message.role),
+    );
+    if (existingIndex === -1) {
+      chat.messages.push(message);
+    } else {
+      chat.messages[existingIndex] = {
+        ...chat.messages[existingIndex],
+        ...message,
+      };
+    }
+    chat.messages = [...chat.messages];
+  },
+
   updateHash(chatID: string | null) {
     const newRoute = CHAT_ROUTE + (chatID ? '/' + chatID : '');
     window.history.pushState(null, '', newRoute);
@@ -202,13 +212,46 @@ Alpine.data('chatManager', () => ({
         (c) => c.id == this.selectedChatID,
       );
 
-      this.chats[chatIndex].messages = response.messages || [];
+      if (chatIndex === -1) return;
+      this.chats[chatIndex] = {
+        ...this.chats[chatIndex],
+        ...response,
+        messages: response.messages || [],
+      };
+      await this.reconnectChatStream(response);
     } catch (err) {
       console.error('Error loading chat messages:', err);
       this.error = 'Failed to load messages';
     }
   },
 
+  async reconnectChatStream(chat: Chat) {
+    const latestMessage = chat.messages[chat.messages.length - 1];
+    if (
+      !latestMessage ||
+      latestMessage.role !== 'assistant' ||
+      latestMessage.status !== 'streaming' ||
+      this.activeStreamChatID === chat.id
+    ) {
+      return;
+    }
+
+    // Reconnect Stream
+    this.loading = true;
+    this.activeStreamChatID = chat.id;
+    try {
+      await streamChatUpdates(chat.id, (chunk: MessageChunk) =>
+        this.applyMessageChunk(chunk),
+      );
+    } catch (err) {
+      console.error('Error reconnecting chat stream:', err);
+      this.error = parseError(err);
+    } finally {
+      this.loading = false;
+      this.activeStreamChatID = null;
+    }
+  },
+
   get models(): Model[] {
     if (!this.settings.text_generation_selector) return this._models;
     return applyFilter(this._models, this.settings.text_generation_selector);

frontend/src/types/index.ts

@@ -7,6 +7,8 @@ export interface Chat {
   messages: Message[];
 }
 
+export type MessageStatus = 'streaming' | 'complete' | 'failed';
+
 export interface Message {
   id: string;
   chat_id: string;
@@ -14,6 +16,7 @@ export interface Message {
   role: 'user' | 'assistant';
   thinking: string;
   content: string;
+  status?: MessageStatus;
   stats?: MessageStats;
 }