From d7a387d911f5f8ed9f1b281ab3681cbf545aa407 Mon Sep 17 00:00:00 2001
From: Mohamed Boudra <boudra.moha@gmail.com>
Date: Sun, 19 Oct 2025 17:22:43 +0200
Subject: [PATCH] ``` refactor: stream text segments for TTS preparation

Refactor LLM streaming to emit complete text segments before tool calls for future TTS integration. Moves from step-based callbacks to chunk-based streaming with text buffering and segment flushing at tool boundaries.
```
---
 .../src/server/agent/llm-openai.ts            | 64 ++++++++++++-------
 .../src/server/agent/orchestrator.ts          | 14 +++-
 2 files changed, 53 insertions(+), 25 deletions(-)

diff --git a/packages/voice-assistant/src/server/agent/llm-openai.ts b/packages/voice-assistant/src/server/agent/llm-openai.ts
index bcdd93f0f..5591dc49b 100644
--- a/packages/voice-assistant/src/server/agent/llm-openai.ts
+++ b/packages/voice-assistant/src/server/agent/llm-openai.ts
@@ -131,6 +131,7 @@ export interface Message {
 export interface StreamLLMParams {
   messages: Message[];
   onChunk?: (chunk: string) => void;
+  onTextSegment?: (segment: string) => void;
   onToolCall?: (toolName: string, args: any) => void;
   onToolResult?: (toolName: string, result: any) => void;
   onFinish?: (fullText: string) => void;
@@ -150,39 +151,54 @@ export async function streamLLM(params: StreamLLMParams): Promise<string> {
     model: openrouter("anthropic/claude-haiku-4.5"),
     messages: params.messages,
     tools: terminalTools,
-    stopWhen: stepCountIs(10),
-    onStepFinish: async (event) => {
-      // Called after each step (tool call or text generation)
-      if (event.toolCalls && event.toolCalls.length > 0) {
-        for (const toolCall of event.toolCalls) {
-          if (params.onToolCall) {
-            params.onToolCall(toolCall.toolName, toolCall.input);
-          }
-        }
-      }
-
-      if (event.toolResults && event.toolResults.length > 0) {
-        for (const toolResult of event.toolResults) {
-          if (params.onToolResult) {
-            params.onToolResult(toolResult.toolName, toolResult.output);
-          }
-        }
-      }
+    onChunk: (chuk) => {
+      // if (params.onChunk) {
+      //   params.onChunk(chunk);
+      // }
     },
+    stopWhen: stepCountIs(10),
   });
 
   let fullText = "";
+  let textBuffer = "";
 
-  result.fullStream;
+  function flushTextBuffer() {
+    if (textBuffer.length > 0 && params.onTextSegment) {
+      params.onTextSegment(textBuffer);
+    }
+    textBuffer = "";
+  }
 
-  // Stream text chunks to the caller
-  for await (const chunk of result.textStream) {
-    fullText += chunk;
-    if (params.onChunk) {
-      params.onChunk(chunk);
+  // Stream all events from fullStream
+  for await (const part of result.fullStream) {
+    if (part.type === "text-delta") {
+      // Accumulate text in buffer
+      textBuffer += part.text;
+      fullText += part.text;
+
+      // Stream individual deltas to caller
+      if (params.onChunk) {
+        params.onChunk(part.text);
+      }
+    } else if (part.type === "tool-call") {
+      // Flush accumulated text as a segment before tool call
+      flushTextBuffer();
+
+      // Emit tool call event
+      if (params.onToolCall) {
+        params.onToolCall(part.toolName, part.input);
+      }
+    } else if (part.type === "tool-result") {
+      // Emit tool result event
+      if (params.onToolResult) {
+        params.onToolResult(part.toolName, part.output);
+      }
     }
   }
 
+  // Flush any remaining text at the end
+  flushTextBuffer();
+
   if (params.onFinish) {
     params.onFinish(fullText);
   }
diff --git a/packages/voice-assistant/src/server/agent/orchestrator.ts b/packages/voice-assistant/src/server/agent/orchestrator.ts
index 36fcdccd6..09991ffd6 100644
--- a/packages/voice-assistant/src/server/agent/orchestrator.ts
+++ b/packages/voice-assistant/src/server/agent/orchestrator.ts
@@ -77,6 +77,18 @@ export async function processUserMessage(params: {
           });
         }
       },
+      onTextSegment: (segment) => {
+        // Broadcast complete text segments (for future TTS integration)
+        if (params.wsServer) {
+          params.wsServer.broadcastActivityLog({
+            id: uuidv4(),
+            timestamp: new Date(),
+            type: "system",
+            content: `Text segment: ${segment}`,
+            metadata: { segment, readyForTTS: true },
+          });
+        }
+      },
       onToolCall: (toolName, args) => {
         // Broadcast tool call to WebSocket
         if (params.wsServer) {
@@ -126,7 +138,7 @@ export async function processUserMessage(params: {
         id: uuidv4(),
         timestamp: new Date(),
         type: "error",
-        content: `Error: ${error.message}`,
+        content: `Error: ${error instanceof Error ? error.message : String(error)}`,
       });
     }
     throw error;