fix(openai): transition to max_completion_tokens for newer models

Newer OpenAI models (o1, o3, gpt-5) have deprecated 'max_tokens' in favor of
'max_completion_tokens'. The provider now automatically maps this parameter
to ensure compatibility and avoid 400 errors.

src/providers/openai.rs

@@ -54,7 +54,15 @@ impl super::Provider for OpenAIProvider {
 
     async fn chat_completion(&self, request: UnifiedRequest) -> Result<ProviderResponse, AppError> {
         let messages_json = helpers::messages_to_openai_json(&request.messages).await?;
-        let body = helpers::build_openai_body(&request, messages_json, false);
+        let mut body = helpers::build_openai_body(&request, messages_json, false);
+
+        // Transition: Newer OpenAI models (o1, o3, gpt-5) require max_completion_tokens
+        // instead of the legacy max_tokens parameter.
+        if request.model.starts_with("o1-") || request.model.starts_with("o3-") || request.model.contains("gpt-5") {
+            if let Some(max_tokens) = body.as_object_mut().and_then(|obj| obj.remove("max_tokens")) {
+                body["max_completion_tokens"] = max_tokens;
+            }
+        }
 
         let response = self
             .client
@@ -301,6 +309,13 @@ impl super::Provider for OpenAIProvider {
         // Standard OpenAI cleanup
         if let Some(obj) = body.as_object_mut() {
             obj.remove("stream_options");
+
+            // Transition: Newer OpenAI models (o1, o3, gpt-5) require max_completion_tokens
+            if request.model.starts_with("o1-") || request.model.starts_with("o3-") || request.model.contains("gpt-5") {
+                if let Some(max_tokens) = obj.remove("max_tokens") {
+                    obj.insert("max_completion_tokens".to_string(), max_tokens);
+                }
+            }
         }
 
         let url = format!("{}/chat/completions", self.config.base_url);