chore: clean

plugins/inference-plugin/@types/global.d.ts

@@ -1,2 +1,3 @@
 declare const PLUGIN_NAME: string;
 declare const MODULE_PATH: string;
+declare const INFERENCE_URL: string;

plugins/inference-plugin/index.ts

@@ -1,17 +1,23 @@
-import { EventName, InferenceService, NewMessageRequest, PluginService, core, events, store } from "@janhq/core";
+import {
+  EventName,
+  InferenceService,
+  NewMessageRequest,
+  PluginService,
+  events,
+  store,
+  invokePluginFunc,
+} from "@janhq/core";
 import { Observable } from "rxjs";
 
-const inferenceUrl = "http://127.0.0.1:3928/llama/chat_completion";
-
-const initModel = async (product) => core.invokePluginFunc(MODULE_PATH, "initModel", product);
+const initModel = async (product) => invokePluginFunc(MODULE_PATH, "initModel", product);
 
 const stopModel = () => {
-  core.invokePluginFunc(MODULE_PATH, "killSubprocess");
+  invokePluginFunc(MODULE_PATH, "killSubprocess");
 };
 
 function requestInference(recentMessages: any[]): Observable<string> {
   return new Observable((subscriber) => {
-    fetch(inferenceUrl, {
+    fetch(INFERENCE_URL, {
       method: "POST",
       headers: {
         "Content-Type": "application/json",
@@ -22,11 +28,7 @@ function requestInference(recentMessages: any[]): Observable<string> {
         messages: recentMessages,
         stream: true,
         model: "gpt-3.5-turbo",
-        max_tokens: 2048,
-        stop: ["hello"],
-        frequency_penalty: 0,
-        presence_penalty: 0,
-        temperature: 0
+        max_tokens: 500,
       }),
     })
       .then(async (response) => {

plugins/inference-plugin/module.ts

@@ -5,8 +5,8 @@ const fs = require("fs");
 const tcpPortUsed = require("tcp-port-used");
 const { killPortProcess } = require("kill-port-process");
 
-let subprocess = null;
 const PORT = 3928;
+let subprocess = null;
 
 const initModel = (fileName) => {
   return (
@@ -15,9 +15,7 @@ const initModel = (fileName) => {
         reject("Model not found, please download again.");
       }
       if (subprocess) {
-        console.error(
-          "A subprocess is already running. Attempt to kill then reinit."
-        );
+        console.error("A subprocess is already running. Attempt to kill then reinit.");
         killSubprocess();
       }
       resolve(fileName);
@@ -32,26 +30,13 @@ const initModel = (fileName) => {
       // Spawn Nitro subprocess to load model
       .then(() => {
         let binaryFolder = path.join(__dirname, "nitro"); // Current directory by default
-
-        const config = {
-          llama_model_path: "",
-          ctx_len: 2048,
-          ngl: 100,
-          embedding: true // Always enable embedding mode on
-        }
-
-        const modelPath = path.join(app.getPath("userData"), fileName);
-
-        config.llama_model_path = modelPath;
-
         let binaryName;
 
         if (process.platform === "win32") {
           binaryName = "nitro_windows_amd64_cuda.exe";
         } else if (process.platform === "darwin") {
           // Mac OS platform
-          binaryName =
-            process.arch === "arm64" ? "nitro_mac_arm64" : "nitro_mac_amd64";
+          binaryName = process.arch === "arm64" ? "nitro_mac_arm64" : "nitro_mac_intel";
         } else {
           // Linux
           binaryName = "nitro_linux_amd64_cuda"; // For other platforms
@@ -60,7 +45,6 @@ const initModel = (fileName) => {
         const binaryPath = path.join(binaryFolder, binaryName);
 
         // Execute the binary
-
         subprocess = spawn(binaryPath, { cwd: binaryFolder });
 
         // Handle subprocess output
@@ -78,13 +62,24 @@ const initModel = (fileName) => {
         });
       })
       .then(() => tcpPortUsed.waitUntilUsed(PORT, 300, 30000))
-      .then((config) => {
-        const initModel = fetch(`http://127.0.0.1:${PORT}/inferences/llamacpp/loadmodel`, {
+      .then(() => {
+        const llama_model_path = path.join(app.getPath("userData"), fileName);
+
+        const config = {
+          llama_model_path,
+          ctx_len: 2048,
+          ngl: 100,
+          embedding: true, // Always enable embedding mode on
+        };
+
+        // Load model config
+        return fetch(`http://127.0.0.1:${PORT}/inferences/llamacpp/loadmodel`, {
           method: "POST",
           headers: {
             "Content-Type": "application/json",
           },
           body: JSON.stringify(config),
+        });
       })
       .then((res) => {
         if (res.ok) {
@@ -92,7 +87,6 @@ const initModel = (fileName) => {
         }
         throw new Error("Nitro: Model failed to load.");
       })
-      })
       .catch((err) => {
         return { error: err };
       })

plugins/inference-plugin/webpack.config.js

@@ -19,6 +19,7 @@ module.exports = {
     new webpack.DefinePlugin({
       PLUGIN_NAME: JSON.stringify(packageJson.name),
       MODULE_PATH: JSON.stringify(`${packageJson.name}/${packageJson.module}`),
+      INFERENCE_URL: JSON.stringify("http://127.0.0.1:3928/inferences/llamacpp/chat_completion"),
     }),
   ],
   output: {