From 11637c52445b12769382b8c5f2169b7d6781d8ad Mon Sep 17 00:00:00 2001
From: Louis <louis@jan.ai>
Date: Tue, 19 Nov 2024 22:10:12 +0700
Subject: [PATCH] fix: correct OpenAI o1 model parameters

---
 .../browser/extensions/engines/helpers/sse.ts    |  4 +++-
 .../resources/models.json                        | 16 ++++++----------
 .../inference-openai-extension/src/index.ts      |  4 ++--
 extensions/model-extension/package.json          |  2 +-
 4 files changed, 12 insertions(+), 14 deletions(-)

diff --git a/core/src/browser/extensions/engines/helpers/sse.ts b/core/src/browser/extensions/engines/helpers/sse.ts
index 9713256b3..bd9945d3c 100644
--- a/core/src/browser/extensions/engines/helpers/sse.ts
+++ b/core/src/browser/extensions/engines/helpers/sse.ts
@@ -45,7 +45,9 @@ export function requestInference(
           subscriber.complete()
           return
         }
-        if (model.parameters?.stream === false) {
+        // There could be overriden stream parameter in the model
+        // that is set in request body (transformed payload)
+        if (requestBody?.stream === false || model.parameters?.stream === false) {
           const data = await response.json()
           if (transformResponse) {
             subscriber.next(transformResponse(data))
diff --git a/extensions/inference-openai-extension/resources/models.json b/extensions/inference-openai-extension/resources/models.json
index 124e123b9..3f41c0a7d 100644
--- a/extensions/inference-openai-extension/resources/models.json
+++ b/extensions/inference-openai-extension/resources/models.json
@@ -97,11 +97,9 @@
     "format": "api",
     "settings": {},
     "parameters": {
-      "max_tokens": 4096,
-      "temperature": 0.7,
-      "top_p": 0.95,
-      "stream": true,
-      "stop": [],
+      "temperature": 1,
+      "top_p": 1,
+      "max_tokens": 32768,
       "frequency_penalty": 0,
       "presence_penalty": 0
     },
@@ -125,11 +123,9 @@
     "format": "api",
     "settings": {},
     "parameters": {
-      "max_tokens": 4096,
-      "temperature": 0.7,
-      "top_p": 0.95,
-      "stream": true,
-      "stop": [],
+      "temperature": 1,
+      "top_p": 1,
+      "max_tokens": 65536,
       "frequency_penalty": 0,
       "presence_penalty": 0
     },
diff --git a/extensions/inference-openai-extension/src/index.ts b/extensions/inference-openai-extension/src/index.ts
index 64880b678..d484c8ae5 100644
--- a/extensions/inference-openai-extension/src/index.ts
+++ b/extensions/inference-openai-extension/src/index.ts
@@ -76,11 +76,11 @@ export default class JanInferenceOpenAIExtension extends RemoteOAIEngine {
   transformPayload = (payload: OpenAIPayloadType): OpenAIPayloadType => {
     // Transform the payload for preview models
     if (this.previewModels.includes(payload.model)) {
-      const { max_tokens, temperature, top_p, stop, ...params } = payload
+      const { max_tokens, stop, ...params } = payload
       return {
         ...params,
         max_completion_tokens: max_tokens,
-        stream: false // o1 only support stream = false
+        stream: false, // o1 only support stream = false
       }
     }
     // Pass through for non-preview models
diff --git a/extensions/model-extension/package.json b/extensions/model-extension/package.json
index bd834454a..ca563ff9f 100644
--- a/extensions/model-extension/package.json
+++ b/extensions/model-extension/package.json
@@ -1,7 +1,7 @@
 {
   "name": "@janhq/model-extension",
   "productName": "Model Management",
-  "version": "1.0.34",
+  "version": "1.0.35",
   "description": "Model Management Extension provides model exploration and seamless downloads",
   "main": "dist/index.js",
   "author": "Jan <service@jan.ai>",