diff --git a/.env.example b/.env.example
index 96570a3d..925f1278 100644
--- a/.env.example
+++ b/.env.example
@@ -57,4 +57,6 @@ LIVEKIT_API_SECRET=
 AI_PROVIDER=anthropic
 AI_MODEL=claude-sonnet-4-20250514
 AI_EFFORT=medium
-AI_API_KEY=
\ No newline at end of file
+AI_API_KEY=
+# If blank default is used
+OPENAI_BASE_URL=https://openrouter.ai/api/v1
diff --git a/src/core/systems/ServerAI.js b/src/core/systems/ServerAI.js
index 81da7475..d5a9c765 100644
--- a/src/core/systems/ServerAI.js
+++ b/src/core/systems/ServerAI.js
@@ -178,68 +178,105 @@ export class ServerAI extends System {
 
 class OpenAIClient {
   constructor(apiKey, model, effort) {
-    this.client = new OpenAI({ apiKey })
+    const config = { apiKey }
+    
+    // Check for custom base URL override
+    const baseURL = process.env.OPENAI_BASE_URL
+    if (baseURL) {
+      config.baseURL = baseURL
+    }
+    
+    this.client = new OpenAI(config)
     this.model = model
     this.effort = effort
   }
 
+
   async create(prompt) {
-    const resp = await this.client.responses.create({
+    const resp = await this.client.chat.completions.create({
       model: this.model,
-      reasoning: { effort: this.effort },
-      // max_output_tokens: 8192,
-      instructions: `
-        ${docs}
-        ===============
-        You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.`,
-      input: `Respond with the javascript needed to generate the following:\n\n"${prompt}"`,
+      max_tokens: 8192,
+      messages: [
+        {
+          role: 'system',
+          content: `
+            ${docs}
+            ===============
+            You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.`,
+        },
+        {
+          role: 'user',
+          content: `Respond with the javascript needed to generate the following:\n\n"${prompt}"`,
+        },
+      ],
     })
-    return resp.output_text
+    return resp.choices[0].message.content
   }
 
   async edit(code, prompt) {
-    const resp = await this.client.responses.create({
+    const resp = await this.client.chat.completions.create({
       model: this.model,
-      reasoning: { effort: this.effort },
-      // max_output_tokens: 8192,
-      instructions: `
-        ${docs}
-        ===============
-        You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.
-        Here is the existing script that you will be working with:
-        ===============
-        ${code}`,
-      input: `Please edit the code above to satisfy the following request:\n\n"${prompt}"`,
+      max_tokens: 8192,
+      messages: [
+        {
+          role: 'system',
+          content: `
+            ${docs}
+            ===============
+            You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.
+            Here is the existing script that you will be working with:
+            ===============
+            ${code}`,
+        },
+        {
+          role: 'user',
+          content: `Please edit the code above to satisfy the following request:\n\n"${prompt}"`,
+        },
+      ],
     })
-    return resp.output_text
+    return resp.choices[0].message.content
   }
 
   async fix(code, error) {
-    const resp = await this.client.responses.create({
+    const resp = await this.client.chat.completions.create({
       model: this.model,
-      reasoning: { effort: this.effort },
-      // max_output_tokens: 8192,
-      instructions: `
-        ${docs}
-        ===============
-        You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.
-        Here is the existing script that you will be working with:
-        ===============
-        ${code}`,
-      input: `This code has an error please fix it:\n\n"${JSON.stringify(error, null, 2)}"`,
+      max_tokens: 8192,
+      messages: [
+        {
+          role: 'system',
+          content: `
+            ${docs}
+            ===============
+            You are an artist and code generator. Always respond with raw code only, never use markdown code blocks or any other formatting.
+            Here is the existing script that you will be working with:
+            ===============
+            ${code}`,
+        },
+        {
+          role: 'user',
+          content: `This code has an error please fix it:\n\n"${JSON.stringify(error, null, 2)}"`,
+        },
+      ],
     })
-    return resp.output_text
+    return resp.choices[0].message.content
   }
 
   async classify(prompt) {
-    const resp = await this.client.responses.create({
+    const resp = await this.client.chat.completions.create({
       model: this.model,
-      reasoning: { effort: this.effort },
-      // max_output_tokens: 8192,
-      instructions: `You are a classifier. We will give you a prompt that a user has entered to generate a 3D object and your job is respond with a short name for the object. For example if someone prompts "a cool gamer desk with neon lights" you would respond with something like "Gamer Desk" because it is a short descriptive name that captures the essence of the object.`,
-      input: `Please classify the following prompt:\n\n"${prompt}"`,
+      max_tokens: 8192,
+      messages: [
+        {
+          role: 'system',
+          content: `You are a classifier. We will give you a prompt that a user has entered to generate a 3D object and your job is respond with a short name for the object. For example if someone prompts "a cool gamer desk with neon lights" you would respond with something like "Gamer Desk" because it is a short descriptive name that captures the essence of the object.`,
+        },
+        {
+          role: 'user',
+          content: `Please classify the following prompt:\n\n"${prompt}"`,
+        },
+      ],
     })
-    return resp.output_text
+    return resp.choices[0].message.content
   }
 }
 
@@ -552,4 +589,4 @@ class XAIClient {
     const data = await resp.json()
     return data.choices[0].message.content
   }
-}
+}
\ No newline at end of file