Enable streaming for tool call logs

server/src/llm/streaming.ts

@@ -1,7 +1,7 @@
 import { performance } from "node:perf_hooks";
 import { prisma } from "../db.js";
 import { anthropicClient, openaiClient, xaiClient } from "./providers.js";
-import { buildToolLogMessageData, runToolAwareOpenAIChat, type ToolExecutionEvent } from "./chat-tools.js";
+import { buildToolLogMessageData, runToolAwareOpenAIChatStream, type ToolExecutionEvent } from "./chat-tools.js";
 import type { MultiplexRequest, Provider } from "./types.js";
 
 export type StreamEvent =
@@ -57,31 +57,33 @@ export async function* runMultiplexStream(req: MultiplexRequest): AsyncGenerator
   try {
     if (req.provider === "openai" || req.provider === "xai") {
       const client = req.provider === "openai" ? openaiClient() : xaiClient();
-      const toolEvents: ToolExecutionEvent[] = [];
-      const r = await runToolAwareOpenAIChat({
+      for await (const ev of runToolAwareOpenAIChatStream({
         client,
         model: req.model,
         messages: req.messages,
         temperature: req.temperature,
         maxTokens: req.maxTokens,
-        onToolEvent: (event) => {
-          toolEvents.push(event);
-        },
         logContext: {
           provider: req.provider,
           model: req.model,
           chatId,
         },
-      });
-      raw = r.raw;
-      text = r.text;
-      usage = r.usage;
-      toolMessages = toolEvents.map((event) => buildToolLogMessageData(chatId, event));
-      for (const event of toolEvents) {
-        yield { type: "tool_call", event };
-      }
-      if (text) {
-        yield { type: "delta", text };
+      })) {
+        if (ev.type === "delta") {
+          text += ev.text;
+          yield { type: "delta", text: ev.text };
+          continue;
+        }
+
+        if (ev.type === "tool_call") {
+          toolMessages.push(buildToolLogMessageData(chatId, ev.event));
+          yield { type: "tool_call", event: ev.event };
+          continue;
+        }
+
+        raw = ev.result.raw;
+        usage = ev.result.usage;
+        text = ev.result.text;
       }
     } else if (req.provider === "anthropic") {
       const client = anthropicClient();