(WIP)randomly generate api-key hash each session

core/src/browser/extensions/engines/AIEngine.ts

@@ -110,6 +110,7 @@ export interface sessionInfo {
   port: number // llama-server output port (corrected from portid)
   modelName: string, //name of the model
   modelPath: string // path of the loaded model
+  api_key: string
 }
 
 // 4. /unload

extensions/llamacpp-extension/src/index.ts

@@ -23,6 +23,7 @@ import {
 } from '@janhq/core'
 
 import { invoke } from '@tauri-apps/api/core'
+import { createHmac } from 'crypto'
 
 type LlamacppConfig = {
   n_gpu_layers: number;
@@ -83,10 +84,11 @@ export default class llamacpp_extension extends AIEngine {
   readonly providerId: string = 'llamacpp'
 
   private config: LlamacppConfig
-  private downloadManager
+  private downloadManager: any
   private activeSessions: Map<string, sessionInfo> = new Map()
   private modelsBasePath!: string
   private enginesBasePath!: string
+  private apiSecret: string = "Jan"
 
   override async onLoad(): Promise<void> {
     super.onLoad() // Calls registerEngine() from AIEngine
@@ -128,6 +130,11 @@ export default class llamacpp_extension extends AIEngine {
     this.config[key] = value
   }
 
+  private generateApiKey(modelId: string): string {
+    const hash = createHmac('sha256', this.apiSecret).update(modelId).digest("base64")
+    return hash
+  }
+
   // Implement the required LocalProvider interface methods
   override async list(): Promise<modelInfo[]> {
     const modelsDir = await joinPath([this.modelsBasePath, this.provider])
@@ -289,6 +296,9 @@ export default class llamacpp_extension extends AIEngine {
 
     // disable llama-server webui
     args.push('--no-webui')
+    // update key for security; TODO: (qnixsynapse) Make it more secure
+    const api_key = this.generateApiKey(opts.modelPath)
+    args.push(`--api-key ${api_key}`)
 
     // model option is required
     // TODO: llama.cpp extension lookup model path based on modelId
@@ -456,7 +466,7 @@ export default class llamacpp_extension extends AIEngine {
     const url = `${baseUrl}/chat/completions`
     const headers = {
       'Content-Type': 'application/json',
-      'Authorization': `Bearer test-k`,
+      'Authorization': `Bearer ${sessionInfo.api_key}`,
     }
 
     const body = JSON.stringify(opts)

src-tauri/src/core/utils/extensions/inference_llamacpp_extension/server.rs

@@ -42,6 +42,7 @@ pub struct SessionInfo {
     pub session_id: String,       // opaque handle for unload/chat
     pub port: u16,                // llama-server output port
     pub model_path: String,       // path of the loaded model
+    pub api_key: String,
 }
 
 #[derive(serde::Serialize, serde::Deserialize)]
@@ -85,7 +86,8 @@ pub async fn load_llama_model(
     // Configure the command to run the server
     let mut command = Command::new(server_path);
 
-    let model_path = args[0].replace("-m", "");
+    let model_path = args[2].replace("-m", "");
+    let api_key = args[1].replace("--api-key", "")
     command.args(args);
 
     // Optional: Redirect stdio if needed (e.g., for logging within Jan)
@@ -110,6 +112,7 @@ pub async fn load_llama_model(
         session_id: pid,  // Use PID as session ID
         port,
         model_path,
+        api_key,
     };
 
     Ok(session_info)