Merge pull request #4338 from janhq/main

Merge bug fixes from 0.5.12 into dev

extensions/conversational-extension/src/index.ts

@@ -40,7 +40,7 @@ export default class CortexConversationalExtension extends ConversationalExtensi
   async listThreads(): Promise<Thread[]> {
     return this.queue.add(() =>
       ky
-        .get(`${API_URL}/v1/threads`)
+        .get(`${API_URL}/v1/threads?limit=-1`)
         .json<ThreadList>()
         .then((e) => e.data)
     ) as Promise<Thread[]>
@@ -133,7 +133,7 @@ export default class CortexConversationalExtension extends ConversationalExtensi
   async listMessages(threadId: string): Promise<ThreadMessage[]> {
     return this.queue.add(() =>
       ky
-        .get(`${API_URL}/v1/threads/${threadId}/messages?order=asc`)
+        .get(`${API_URL}/v1/threads/${threadId}/messages?order=asc&limit=-1`)
         .json<MessageList>()
         .then((e) => e.data)
     ) as Promise<ThreadMessage[]>
@@ -147,7 +147,9 @@ export default class CortexConversationalExtension extends ConversationalExtensi
    */
   async getThreadAssistant(threadId: string): Promise<ThreadAssistantInfo> {
     return this.queue.add(() =>
-      ky.get(`${API_URL}/v1/assistants/${threadId}`).json<ThreadAssistantInfo>()
+      ky
+        .get(`${API_URL}/v1/assistants/${threadId}?limit=-1`)
+        .json<ThreadAssistantInfo>()
     ) as Promise<ThreadAssistantInfo>
   }
   /**
@@ -188,7 +190,7 @@ export default class CortexConversationalExtension extends ConversationalExtensi
    * Do health check on cortex.cpp
    * @returns
    */
-  healthz(): Promise<void> {
+  async healthz(): Promise<void> {
     return ky
       .get(`${API_URL}/healthz`, {
         retry: { limit: 20, delay: () => 500, methods: ['get'] },

extensions/inference-cortex-extension/bin/version.txt

@@ -1 +1 @@
-1.0.5-rc2
+1.0.6

extensions/inference-openai-extension/resources/models.json

@@ -85,6 +85,63 @@
     },
     "engine": "openai"
   },
+  {
+    "sources": [
+      {
+        "url": "https://openai.com"
+      }
+    ],
+    "id": "gpt-4o-mini",
+    "object": "model",
+    "name": "OpenAI GPT 4o-mini",
+    "version": "1.1",
+    "description": "GPT-4o mini (“o” for “omni”) is a fast, affordable small model for focused tasks.",
+    "format": "api",
+    "settings": {
+      "vision_model": true
+    },
+    "parameters": {
+      "max_tokens": 16384,
+      "temperature": 0.7,
+      "top_p": 0.95,
+      "stream": true,
+      "stop": [],
+      "frequency_penalty": 0,
+      "presence_penalty": 0
+    },
+    "metadata": {
+      "author": "OpenAI",
+      "tags": ["General"]
+    },
+    "engine": "openai"
+  },
+  {
+    "sources": [
+      {
+        "url": "https://openai.com"
+      }
+    ],
+    "id": "o1",
+    "object": "model",
+    "name": "OpenAI o1",
+    "version": "1.0",
+    "description": "OpenAI o1 is a new model with complex reasoning",
+    "format": "api",
+    "settings": {},
+    "parameters": {
+      "max_tokens": 100000,
+      "temperature": 1,
+      "top_p": 1,
+      "stream": true,
+      "frequency_penalty": 0,
+      "presence_penalty": 0
+    },
+    "metadata": {
+      "author": "OpenAI",
+      "tags": ["General"]
+    },
+    "engine": "openai"
+  },
   {
     "sources": [
       {

extensions/model-extension/src/cortex.ts

@@ -53,7 +53,7 @@ export class CortexAPI implements ICortexAPI {
    */
   getModels(): Promise<Model[]> {
     return this.queue
-      .add(() => ky.get(`${API_URL}/v1/models`).json<ModelList>())
+      .add(() => ky.get(`${API_URL}/v1/models?limit=-1`).json<ModelList>())
       .then((e) =>
         typeof e === 'object' ? e.data.map((e) => this.transformModel(e)) : []
       )

joi/src/core/TextArea/index.tsx

@@ -34,7 +34,11 @@ const TextArea = forwardRef<HTMLTextAreaElement, TextAreaProps>(
     return (
       <div className="textarea__wrapper">
         <textarea
-          className={twMerge('textarea', className)}
+          className={twMerge(
+            'textarea',
+            className,
+            autoResize && 'resize-none'
+          )}
           ref={autoResize ? textareaRef : ref}
           {...props}
         />

web/containers/ModelDropdown/index.tsx

@@ -183,10 +183,7 @@ const ModelDropdown = ({
     if (!activeThread) return
     const modelId = activeAssistant?.model?.id
 
-    let model = downloadedModels.find((model) => model.id === modelId)
-    if (!model) {
-      model = undefined
-    }
+    const model = downloadedModels.find((model) => model.id === modelId)
     setSelectedModel(model)
   }, [
     recommendedModel,
@@ -378,14 +375,14 @@ const ModelDropdown = ({
                 !selectedModel && 'text-[hsla(var(--text-tertiary))]'
               )}
             >
-              {selectedModel?.name || 'Select Model'}
+              {selectedModel?.name || 'Select a model'}
             </span>
           </Badge>
         ) : (
           <Input
             value={selectedModel?.name || ''}
             className="cursor-pointer"
-            placeholder="Select Model"
+            placeholder="Select a model"
             disabled={disabled}
             readOnly
             suffixIcon={

web/containers/Providers/ModelHandler.tsx

@@ -18,7 +18,7 @@ import {
   extractInferenceParams,
   ModelExtension,
 } from '@janhq/core'
-import { useAtomValue, useSetAtom } from 'jotai'
+import { useAtom, useAtomValue, useSetAtom } from 'jotai'
 import { ulid } from 'ulidx'
 
 import { activeModelAtom, stateModelAtom } from '@/hooks/useActiveModel'
@@ -32,6 +32,7 @@ import {
   updateMessageAtom,
   tokenSpeedAtom,
   deleteMessageAtom,
+  subscribedGeneratingMessageAtom,
 } from '@/helpers/atoms/ChatMessage.atom'
 import { downloadedModelsAtom } from '@/helpers/atoms/Model.atom'
 import {
@@ -40,6 +41,7 @@ import {
   isGeneratingResponseAtom,
   updateThreadAtom,
   getActiveThreadModelParamsAtom,
+  activeThreadAtom,
 } from '@/helpers/atoms/Thread.atom'
 
 const maxWordForThreadTitle = 10
@@ -54,6 +56,10 @@ export default function ModelHandler() {
   const activeModel = useAtomValue(activeModelAtom)
   const setActiveModel = useSetAtom(activeModelAtom)
   const setStateModel = useSetAtom(stateModelAtom)
+  const [subscribedGeneratingMessage, setSubscribedGeneratingMessage] = useAtom(
+    subscribedGeneratingMessageAtom
+  )
+  const activeThread = useAtomValue(activeThreadAtom)
 
   const updateThreadWaiting = useSetAtom(updateThreadWaitingForResponseAtom)
   const threads = useAtomValue(threadsAtom)
@@ -62,11 +68,17 @@ export default function ModelHandler() {
   const setIsGeneratingResponse = useSetAtom(isGeneratingResponseAtom)
   const updateThread = useSetAtom(updateThreadAtom)
   const messagesRef = useRef(messages)
+  const messageGenerationSubscriber = useRef(subscribedGeneratingMessage)
   const activeModelRef = useRef(activeModel)
+  const activeThreadRef = useRef(activeThread)
   const activeModelParams = useAtomValue(getActiveThreadModelParamsAtom)
   const activeModelParamsRef = useRef(activeModelParams)
   const setTokenSpeed = useSetAtom(tokenSpeedAtom)
 
+  useEffect(() => {
+    activeThreadRef.current = activeThread
+  }, [activeThread])
+
   useEffect(() => {
     threadsRef.current = threads
   }, [threads])
@@ -87,6 +99,10 @@ export default function ModelHandler() {
     activeModelParamsRef.current = activeModelParams
   }, [activeModelParams])
 
+  useEffect(() => {
+    messageGenerationSubscriber.current = subscribedGeneratingMessage
+  }, [subscribedGeneratingMessage])
+
   const onNewMessageResponse = useCallback(
     async (message: ThreadMessage) => {
       if (message.type === MessageRequestType.Thread) {
@@ -179,12 +195,19 @@ export default function ModelHandler() {
 
   const updateThreadMessage = useCallback(
     (message: ThreadMessage) => {
-      updateMessage(
-        message.id,
-        message.thread_id,
-        message.content,
-        message.status
-      )
+      if (
+        messageGenerationSubscriber.current &&
+        message.thread_id === activeThreadRef.current?.id &&
+        !messageGenerationSubscriber.current!.thread_id
+      ) {
+        updateMessage(
+          message.id,
+          message.thread_id,
+          message.content,
+          message.status
+        )
+      }
+
       if (message.status === MessageStatus.Pending) {
         if (message.content.length) {
           setIsGeneratingResponse(false)
@@ -244,6 +267,7 @@ export default function ModelHandler() {
       const metadata = {
         ...thread.metadata,
         ...(messageContent && { lastMessage: messageContent }),
+        updated_at: Date.now(),
       }
 
       updateThread({
@@ -302,15 +326,10 @@ export default function ModelHandler() {
 
   const generateThreadTitle = (message: ThreadMessage, thread: Thread) => {
     // If this is the first ever prompt in the thread
-    if (
-      (thread.title ?? thread.metadata?.title)?.trim() !== defaultThreadTitle
-    ) {
+    if ((thread.title ?? thread.metadata?.title)?.trim() !== defaultThreadTitle)
       return
-    }
 
-    if (!activeModelRef.current) {
-      return
-    }
+    if (!activeModelRef.current) return
 
     // Check model engine; we don't want to generate a title when it's not a local engine. remote model using first promp
     if (!isLocalEngine(activeModelRef.current?.engine as InferenceEngine)) {
@@ -332,6 +351,7 @@ export default function ModelHandler() {
             ...updatedThread,
           })
         })
+        .catch(console.error)
     }
 
     // This is the first time message comes in on a new thread

web/helpers/atoms/ChatMessage.atom.ts

@@ -35,6 +35,13 @@ export const chatMessages = atom(
   }
 )
 
+/**
+ * Store subscribed generating message thread
+ */
+export const subscribedGeneratingMessageAtom = atom<{
+  thread_id?: string
+}>({})
+
 /**
  * Stores the status of the messages load for each thread
  */
@@ -175,6 +182,17 @@ export const updateMessageAtom = atom(
       // Update thread last message
       if (text.length)
         set(updateThreadStateLastMessageAtom, conversationId, text)
+    } else {
+      set(addNewMessageAtom, {
+        id,
+        thread_id: conversationId,
+        content: text,
+        status,
+        role: ChatCompletionRole.Assistant,
+        created_at: Date.now() / 1000,
+        completed_at: Date.now() / 1000,
+        object: 'thread.message',
+      })
     }
   }
 )

web/helpers/atoms/Setting.atom.ts

@@ -20,12 +20,7 @@ export const CHAT_WIDTH = 'chatWidth'
 export const themesOptionsAtom = atomWithStorage<
   { name: string; value: string }[]
 >(THEME_OPTIONS, [], undefined, { getOnInit: true })
-export const janThemesPathAtom = atomWithStorage<string | undefined>(
-  THEME_PATH,
-  undefined,
-  undefined,
-  { getOnInit: true }
-)
+
 export const selectedThemeIdAtom = atomWithStorage<string>(
   THEME,
   '',

web/helpers/atoms/Thread.atom.ts

@@ -125,6 +125,26 @@ export const waitingToSendMessage = atom<boolean | undefined>(undefined)
  */
 export const isGeneratingResponseAtom = atom<boolean | undefined>(undefined)
 
+/**
+ * Create a new thread and add it to the thread list
+ */
+export const createNewThreadAtom = atom(null, (get, set, newThread: Thread) => {
+  // create thread state for this new thread
+  const currentState = { ...get(threadStatesAtom) }
+
+  const threadState: ThreadState = {
+    hasMore: false,
+    waitingForResponse: false,
+    lastMessage: undefined,
+  }
+  currentState[newThread.id] = threadState
+  set(threadStatesAtom, currentState)
+
+  // add the new thread on top of the thread list to the state
+  const threads = get(threadsAtom)
+  set(threadsAtom, [newThread, ...threads])
+})
+
 /**
  * Remove a thread state from the atom
  */
@@ -180,12 +200,12 @@ export const updateThreadAtom = atom(
     )
 
     // sort new threads based on updated at
-    threads.sort((thread1, thread2) => {
-      const aDate = new Date(thread1.updated ?? 0)
-      const bDate = new Date(thread2.updated ?? 0)
-      return bDate.getTime() - aDate.getTime()
+    threads.sort((a, b) => {
+      return ((a.metadata?.updated_at as number) ?? 0) >
+        ((b.metadata?.updated_at as number) ?? 0)
+        ? -1
+        : 1
     })
-
     set(threadsAtom, threads)
   }
 )