From 20a652e0ebce6fa7235d49da7efb45b662d7ea63 Mon Sep 17 00:00:00 2001
From: Louis <louis@jan.ai>
Date: Sat, 7 Dec 2024 14:35:03 +0700
Subject: [PATCH] chore: fix openai vision models

---
 extensions/inference-openai-extension/package.json          | 2 +-
 extensions/inference-openai-extension/resources/models.json | 4 +++-
 extensions/inference-openai-extension/src/index.ts          | 5 +++++
 3 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/extensions/inference-openai-extension/package.json b/extensions/inference-openai-extension/package.json
index 9700383d63..d5b2a1d7a6 100644
--- a/extensions/inference-openai-extension/package.json
+++ b/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",
diff --git a/extensions/inference-openai-extension/resources/models.json b/extensions/inference-openai-extension/resources/models.json
index a34bc54604..4d19654bc5 100644
--- a/extensions/inference-openai-extension/resources/models.json
+++ b/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,
diff --git a/extensions/inference-openai-extension/src/index.ts b/extensions/inference-openai-extension/src/index.ts
index 18bc4e0aae..2612ed8153 100644
--- a/extensions/inference-openai-extension/src/index.ts
+++ b/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