Merge pull request #4246 from janhq/fix/openai-vision-models

chore: fix openai vision models

extensions/inference-openai-extension/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@janhq/inference-openai-extension",
   "productName": "OpenAI Inference Engine",
-  "version": "1.0.4",
+  "version": "1.0.5",
   "description": "This extension enables OpenAI chat completion API calls",
   "main": "dist/index.js",
   "module": "dist/module.js",

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

@@ -67,7 +67,9 @@
     "version": "1.1",
     "description": "OpenAI GPT 4o is a new flagship model with fast speed and high quality",
     "format": "api",
-    "settings": {},
+    "settings": {
+      "vision_model": true
+    },
     "parameters": {
       "max_tokens": 4096,
       "temperature": 0.7,

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

@@ -74,6 +74,11 @@ export default class JanInferenceOpenAIExtension extends RemoteOAIEngine {
    * @returns
    */
   transformPayload = (payload: OpenAIPayloadType): OpenAIPayloadType => {
+    // Remove empty stop words
+    if (payload.stop?.length === 0) {
+      const { stop, ...params } = payload
+      payload = params
+    }
     // Transform the payload for preview models
     if (this.previewModels.includes(payload.model)) {
       const { max_tokens, stop, ...params } = payload