diff --git a/src/modules/llms/server/llm.server.streaming.ts b/src/modules/llms/server/llm.server.streaming.ts
index 421d64efb..1d056e6ae 100644
--- a/src/modules/llms/server/llm.server.streaming.ts
+++ b/src/modules/llms/server/llm.server.streaming.ts
@@ -217,8 +217,7 @@ function createEventStreamTransformer(muxingFormat: MuxingFormat, vendorTextPars
 
       // Send initial packet indicating the start of the stream
       const startPacket: ChatStreamingPreambleStartSchema = { type: 'start' };
-      const preambleStart = JSON.stringify(startPacket) + '\n';
-      controller.enqueue(textEncoder.encode(preambleStart));
+      controller.enqueue(textEncoder.encode(JSON.stringify(startPacket)));
 
       // only used for debugging
       let debugLastMs: number | null = null;
@@ -307,8 +306,8 @@ function createStreamParserAnthropicMessages(): AIStreamParser {
         responseMessage = anthropicWireMessagesResponseSchema.parse(message);
         // hack: prepend the model name to the first packet
         if (firstMessage) {
-          const preambleModel: ChatStreamingPreambleModelSchema = { model: responseMessage.model };
-          text = JSON.stringify(preambleModel) + '\n';
+          const firstPacket: ChatStreamingPreambleModelSchema = { model: responseMessage.model };
+          text = JSON.stringify(firstPacket);
         }
         break;
 
@@ -422,8 +421,8 @@ function createStreamParserGemini(modelName: string): AIStreamParser {
     // hack: prepend the model name to the first packet
     if (!hasBegun) {
       hasBegun = true;
-      const preambleModel: ChatStreamingPreambleModelSchema = { model: modelName };
-      text = JSON.stringify(preambleModel) + '\n' + text;
+      const firstPacket: ChatStreamingPreambleModelSchema = { model: modelName };
+      text = JSON.stringify(firstPacket) + text;
     }
 
     return { text, close: false };
@@ -458,8 +457,8 @@ function createStreamParserOllama(): AIStreamParser {
     // hack: prepend the model name to the first packet
     if (!hasBegun && chunk.model) {
       hasBegun = true;
-      const preambleModel: ChatStreamingPreambleModelSchema = { model: chunk.model };
-      text = JSON.stringify(preambleModel) + '\n' + text;
+      const firstPacket: ChatStreamingPreambleModelSchema = { model: chunk.model };
+      text = JSON.stringify(firstPacket) + text;
     }
 
     return { text, close: chunk.done };
@@ -499,8 +498,8 @@ function createStreamParserOpenAI(): AIStreamParser {
     // hack: prepend the model name to the first packet
     if (!hasBegun) {
       hasBegun = true;
-      const preambleModel: ChatStreamingPreambleModelSchema = { model: json.model };
-      text = JSON.stringify(preambleModel) + '\n' + text;
+      const firstPacket: ChatStreamingPreambleModelSchema = { model: json.model };
+      text = JSON.stringify(firstPacket) + text;
     }
 
     // [LocalAI] workaround: LocalAI doesn't send the [DONE] event, but similarly to OpenAI, it sends a "finish_reason" delta update
diff --git a/src/modules/llms/vendors/unifiedStreamingClient.ts b/src/modules/llms/vendors/unifiedStreamingClient.ts
index c94e6250c..07094baa6 100644
--- a/src/modules/llms/vendors/unifiedStreamingClient.ts
+++ b/src/modules/llms/vendors/unifiedStreamingClient.ts
@@ -94,10 +94,10 @@ export async function unifiedStreamingClient<TSourceSetup = unknown, TLLMOptions
     while ((!parsedPreambleStart || !parsedPreableModel) && incrementalText.startsWith('{')) {
 
       // extract a complete JSON object, if present
-      const endOfJson = incrementalText.indexOf('}\n');
+      const endOfJson = incrementalText.indexOf('}');
       if (endOfJson === -1) break;
       const jsonString = incrementalText.substring(0, endOfJson + 1);
-      incrementalText = incrementalText.substring(endOfJson + 2);
+      incrementalText = incrementalText.substring(endOfJson + 1);
 
       // first packet: preamble to let the Vercel edge function go over time
       if (!parsedPreambleStart) {