add kia ai chat tools16

src/services/kiaai/chat.py

@@ -122,6 +122,10 @@ class KiaAIService:
         Core SSE parser — yields one AIMessageChunk per content delta token.
         Final chunk: AIMessageChunk(content="", usage_metadata={...}).
         Raises RuntimeError on KIA server errors (triggers retry in stream()).
+
+        Uses aiter_bytes(chunk_size=32) instead of aiter_lines() so every
+        byte received from the network is processed immediately without
+        waiting for httpx's internal 64 KB read buffer to fill up.
         """
         usage: dict = {}
 
@@ -139,7 +143,15 @@ class KiaAIService:
                 resp.raise_for_status()
 
                 line_count = 0
-                async for line in resp.aiter_lines():
+                buf = ""
+                done = False
+
+                async for raw in resp.aiter_bytes(chunk_size=32):
+                    buf += raw.decode("utf-8", errors="replace")
+
+                    while "\n" in buf:
+                        line, buf = buf.split("\n", 1)
+                        line = line.rstrip("\r")
                         line_count += 1
                         if line_count <= 3:
                             print(f"[KiaAI] raw[{line_count}]: {line!r}")
@@ -165,6 +177,7 @@ class KiaAIService:
 
                         if raw_data.strip() == "[DONE]":
                             print("[KiaAI] [DONE]")
+                            done = True
                             break
 
                         try:
@@ -190,6 +203,9 @@ class KiaAIService:
                         if data.get("usage"):
                             usage = data["usage"]
 
+                    if done:
+                        break
+
         # Final chunk carries usage metadata
         print(f"[KiaAI] Stream done | total_lines={line_count} | usage={usage}")
         yield AIMessageChunk(