update hf.js inference

package-lock.json

@@ -9,7 +9,7 @@
       "version": "0.0.0",
       "dependencies": {
         "@huggingface/hub": "^1.1.1",
-        "@huggingface/inference": "^3.6.1",
+        "@huggingface/inference": "^4.0.2",
         "@monaco-editor/react": "^4.7.0",
         "@radix-ui/react-avatar": "^1.1.10",
         "@radix-ui/react-dialog": "^1.1.14",
@@ -920,29 +920,32 @@
       "integrity": "sha512-HK6JTVB/nrgjOnbe77HFSENftfAp67AI4mHMR2x64Os1hvchuTT88M8fKEiyESSvqKFKwW4lQKkHva07p05AXw=="
     },
     "node_modules/@huggingface/inference": {
-      "version": "3.6.1",
-      "resolved": "https://registry.npmjs.org/@huggingface/inference/-/inference-3.6.1.tgz",
-      "integrity": "sha512-EtQlbBqcZycPe+qiTEFI+wNHOMpG0gwNTaZSvYu1juN1p/1dEgqAb2GO31dxLgNev2PzH9d+9nm8GngOsIepJg==",
+      "version": "4.0.2",
+      "resolved": "https://registry.npmjs.org/@huggingface/inference/-/inference-4.0.2.tgz",
+      "integrity": "sha512-XuWb8ocH7lA5kSdXrGnqshtRz3ocSBzEzxcp5xeAXLjgM1ocoIHq+RW8/Ti0xq3MeRGQWgUkYPCgDV/xgs8p4g==",
+      "license": "MIT",
       "dependencies": {
-        "@huggingface/jinja": "^0.3.3",
-        "@huggingface/tasks": "^0.17.8"
+        "@huggingface/jinja": "^0.5.0",
+        "@huggingface/tasks": "^0.19.11"
       },
       "engines": {
         "node": ">=18"
       }
     },
     "node_modules/@huggingface/jinja": {
-      "version": "0.3.3",
-      "resolved": "https://registry.npmjs.org/@huggingface/jinja/-/jinja-0.3.3.tgz",
-      "integrity": "sha512-vQQr2JyWvVFba3Lj9es4q9vCl1sAc74fdgnEMoX8qHrXtswap9ge9uO3ONDzQB0cQ0PUyaKY2N6HaVbTBvSXvw==",
+      "version": "0.5.0",
+      "resolved": "https://registry.npmjs.org/@huggingface/jinja/-/jinja-0.5.0.tgz",
+      "integrity": "sha512-Ptc03/jGRiYRoi0bUYKZ14MkDslsBRT24oxmsvUlfYrvQMldrxCevhPnT+hfX8awKTT8/f/0ZBBWldoeAcMHdQ==",
+      "license": "MIT",
       "engines": {
         "node": ">=18"
       }
     },
     "node_modules/@huggingface/tasks": {
-      "version": "0.17.9",
-      "resolved": "https://registry.npmjs.org/@huggingface/tasks/-/tasks-0.17.9.tgz",
-      "integrity": "sha512-lV6RgCJkqy3p93FFxP9H4SGJmFcHAwr1FO+Zk56q/JWsf7Tdsel1DEo1Xfd3An7ZPWpc2Y9ldRecGo9efDYghg=="
+      "version": "0.19.11",
+      "resolved": "https://registry.npmjs.org/@huggingface/tasks/-/tasks-0.19.11.tgz",
+      "integrity": "sha512-oBhSgVlg7Pp643MsH8BiI3OAXIMJNxdSiMtv4mApRZV8dmAz8oasKhg6CVKIplO7vAO7F6dkmMn4bYM64I2A9w==",
+      "license": "MIT"
     },
     "node_modules/@humanfs/core": {
       "version": "0.19.1",

package.json

@@ -12,7 +12,7 @@
   },
   "dependencies": {
     "@huggingface/hub": "^1.1.1",
-    "@huggingface/inference": "^3.6.1",
+    "@huggingface/inference": "^4.0.2",
     "@monaco-editor/react": "^4.7.0",
     "@radix-ui/react-avatar": "^1.1.10",
     "@radix-ui/react-dialog": "^1.1.14",

server.js

@@ -298,7 +298,6 @@ app.post("/api/ask-ai", async (req, res) => {
       message: `Context is too long. ${selectedProvider.name} allow ${selectedProvider.max_tokens} max tokens.`,
     });
   }
-
   try {
     const chatCompletion = client.chatCompletionStream(
       {
@@ -316,7 +315,9 @@ app.post("/api/ask-ai", async (req, res) => {
         ],
         max_tokens: selectedProvider.max_tokens,
       },
-      billTo ? { billTo } : {}
+      {
+        billTo: "huggingface",
+      }
     );
 
     while (true) {
@@ -465,7 +466,6 @@ ${REPLACE_END}
   const client = new InferenceClient(token);
 
   const selectedProvider = PROVIDERS[selectedModel.autoProvider];
-
   try {
     const response = await client.chatCompletion(
       {