docs(openai): clarify non-streaming path comment and rationale for system role; align with review suggestions

Author: roomote

src/api/providers/openai.ts

@@ -218,9 +218,14 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				yield this.processUsageMetrics(lastUsage, modelInfo)
 			}
 		} else {
-			// o1 for instance doesnt support streaming, non-1 temp, or system prompt
+			// Note: o1/o3/o4 models do not support streaming, non-1 temperature, or system prompts.
+			// This non-streaming branch is for general OpenAI-compatible providers that DO support system prompts.
+			// The o1/o3/o4 family is handled above in handleO3FamilyMessage(), so we still use a proper "system" role
+			// here for consistency with streaming behavior and provider expectations.
 			const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
 				role: "system",
+				// Use "system" to match streaming behavior; many OpenAI-compatible providers treat system prompts specially
+				// (e.g., safety, formatting, or caching), and tests assert consistency across modes.
 				content: systemPrompt,
 			}