Switch from GPT-3.5 to GPT-4o Mini

Author: ariya

.github/workflows/test-openai.yml

@@ -21,7 +21,7 @@ jobs:
         env:
           LLM_API_BASE_URL: 'https://api.openai.com/v1'
           LLM_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          LLM_CHAT_MODEL: 'gpt-3.5-turbo-0125'
+          LLM_CHAT_MODEL: 'gpt-4o-mini'
           LLM_STREAMING: 'no'
 
       - run: cat output.txt

ask-llm.clj

@@ -32,7 +32,7 @@
 
 (defn chat [messages]
   (let [body {:messages messages
-              :model (or LLM-CHAT-MODEL "gpt-3.5-turbo")
+              :model (or LLM-CHAT-MODEL "gpt-4o-mini")
               :stop ["<|im_end|>" "<|end|>" "<|eot_id|>"]
               :max_tokens 200
               :temperature 0}

ask-llm.js

@@ -35,7 +35,7 @@ const LLM_DEBUG = process.env.LLM_DEBUG;
 const chat = async (messages, handler) => {
     const url = `${LLM_API_BASE_URL}/chat/completions`;
     const auth = LLM_API_KEY ? { 'Authorization': `Bearer ${LLM_API_KEY}` } : {};
-    const model = LLM_CHAT_MODEL || 'gpt-3.5-turbo';
+    const model = LLM_CHAT_MODEL || 'gpt-4o-mini';
     const stop = ['<|im_end|>', '<|end|>', '<|eot_id|>'];
     const max_tokens = 200;
     const temperature = 0;
@@ -149,4 +149,4 @@ const SYSTEM_PROMPT = 'Answer the question politely and concisely.';
     }
 
     qa();
-})();
+})();

ask-llm.py

@@ -24,7 +24,7 @@ async def chat(messages, handler=None):
     if auth_header:
         headers["Authorization"] = auth_header
 
-    model = LLM_CHAT_MODEL or "gpt-3.5-turbo"
+    model = LLM_CHAT_MODEL or "gpt-4o-mini"
     stop = ["<|im_end|>", "<|end|>", "<|eot_id|>"]
     max_tokens = 200
     temperature = 0