fix: #1569 - Does not apply thread settings when loading model (#1576)

extensions/inference-nitro-extension/src/index.ts

@@ -22,6 +22,7 @@ import {
   joinPath,
   InferenceExtension,
   log,
+  InferenceEngine,
 } from "@janhq/core";
 import { requestInference } from "./helpers/sse";
 import { ulid } from "ulid";
@@ -128,7 +129,7 @@ export default class JanInferenceNitroExtension implements InferenceExtension {
   }
 
   private async onModelInit(model: Model) {
-    if (model.engine !== "nitro") return;
+    if (model.engine !== InferenceEngine.nitro) return;
 
     const modelFullPath = await joinPath(["models", model.id]);
 

extensions/inference-nitro-extension/src/module.ts

@@ -180,7 +180,7 @@ function promptTemplateConverter(promptTemplate) {
  * @returns A Promise that resolves when the model is loaded successfully, or rejects with an error message if the model is not found or fails to load.
  */
 function loadLLMModel(settings): Promise<Response> {
-  log(`[NITRO]::Debug: Loading model with params ${settings}`);
+  log(`[NITRO]::Debug: Loading model with params ${JSON.stringify(settings)}`);
   return fetchRetry(NITRO_HTTP_LOAD_MODEL_URL, {
     method: "POST",
     headers: {

web/hooks/useActiveModel.ts

@@ -1,11 +1,12 @@
 /* eslint-disable @typescript-eslint/no-explicit-any */
 import { EventName, events, Model } from '@janhq/core'
-import { atom, useAtom } from 'jotai'
+import { atom, useAtom, useAtomValue } from 'jotai'
 
 import { toaster } from '@/containers/Toast'
 
 import { useGetDownloadedModels } from './useGetDownloadedModels'
 import { LAST_USED_MODEL_ID } from './useRecommendedModel'
+import { activeThreadAtom } from '@/helpers/atoms/Thread.atom'
 
 export const activeModelAtom = atom<Model | undefined>(undefined)
 
@@ -17,6 +18,7 @@ export const stateModelAtom = atom({
 
 export function useActiveModel() {
   const [activeModel, setActiveModel] = useAtom(activeModelAtom)
+  const activeThread = useAtomValue(activeThreadAtom)
   const [stateModel, setStateModel] = useAtom(stateModelAtom)
   const { downloadedModels } = useGetDownloadedModels()
 
@@ -34,7 +36,7 @@ export function useActiveModel() {
 
     setStateModel({ state: 'start', loading: true, model: modelId })
 
-    const model = downloadedModels.find((e) => e.id === modelId)
+    let model = downloadedModels.find((e) => e.id === modelId)
 
     if (!model) {
       toaster({
@@ -49,6 +51,17 @@ export function useActiveModel() {
       return
     }
 
+    /// Apply thread model settings
+    if (activeThread?.assistants[0]?.model.id === modelId) {
+      model = {
+        ...model,
+        settings: {
+          ...model.settings,
+          ...activeThread.assistants[0].model.settings,
+        },
+      }
+    }
+
     localStorage.setItem(LAST_USED_MODEL_ID, model.id)
     events.emit(EventName.OnModelInit, model)
   }