ai: Hide reasoning status after complete.

jarvis.py

@@ -218,14 +218,12 @@ async def fetch_response_stream_async(host, key, model, msgs, cfg, sid, stop_eve
                             try:
                                 j = json.loads(data)
                                 if isinstance(j, dict) and j.get("choices"):
-                                    ch = j["choices"][0]
-                                    delta = ch.get("delta", {})
-                                    reasoning = delta.get("reasoning")
-                                    content = delta.get("content")
-                                    if reasoning:
-                                        yield reasoning
-                                    if content:
-                                        yield content
+                                    for ch in j["choices"]:
+                                        delta = ch.get("delta", {})
+                                        if "reasoning" in delta and delta["reasoning"] is not None and delta["reasoning"] != "":
+                                            yield ("reasoning", delta["reasoning"])
+                                        if "content" in delta and delta["content"] is not None and delta["content"] != "":
+                                            yield ("content", delta["content"])
                             except:
                                 continue
         except:
@@ -236,7 +234,7 @@ async def fetch_response_stream_async(host, key, model, msgs, cfg, sid, stop_eve
 async def chat_with_model_async(history, user_input, model_display, sess, custom_prompt):
     ensure_stop_event(sess)
     if not get_available_items(LINUX_SERVER_PROVIDER_KEYS, LINUX_SERVER_PROVIDER_KEYS_MARKED) or not get_available_items(LINUX_SERVER_HOSTS, LINUX_SERVER_HOSTS_ATTEMPTS):
-        yield RESPONSES["RESPONSE_3"]
+        yield ("content", RESPONSES["RESPONSE_3"])
         return
     if not hasattr(sess, "session_id") or not sess.session_id:
         sess.session_id = str(uuid.uuid4())
@@ -270,11 +268,11 @@ async def chat_with_model_async(history, user_input, model_display, sess, custom
                 if not got_any:
                     got_any = True
                     sess.active_candidate = (h, k)
-                full_text += chunk
+                full_text += chunk[1]
                 yield chunk
             if got_any and full_text:
                 return
-    yield RESPONSES["RESPONSE_2"]
+    yield ("content", RESPONSES["RESPONSE_2"])
 
 async def respond_async(multi, history, model_display, sess, custom_prompt):
     ensure_stop_event(sess)
@@ -293,15 +291,26 @@ async def respond_async(multi, history, model_display, sess, custom_prompt):
     yield history, gr.update(interactive=False, submit_btn=False, stop_btn=True), sess
     queue = asyncio.Queue()
     async def background():
-        full = ""
-        async for chunk in chat_with_model_async(history, inp, model_display, sess, custom_prompt):
-            full += chunk
-            await queue.put(chunk)
+        display_text = ""
+        content_started = False
+        async for typ, chunk in chat_with_model_async(history, inp, model_display, sess, custom_prompt):
+            if typ == "reasoning":
+                if not content_started:
+                    display_text += chunk
+                    await queue.put(("set", display_text))
+            else:
+                if not content_started:
+                    content_started = True
+                    display_text = chunk
+                    await queue.put(("replace", display_text))
+                else:
+                    display_text += chunk
+                    await queue.put(("append", display_text))
         await queue.put(None)
-        return full
+        return display_text
     bg_task = asyncio.create_task(background())
     stop_task = asyncio.create_task(sess.stop_event.wait())
-    first_meaningful_chunk_found = False
+    first_update = False
     try:
         while True:
             done, _ = await asyncio.wait({stop_task, asyncio.create_task(queue.get())}, return_when=asyncio.FIRST_COMPLETED)
@@ -312,15 +321,11 @@ async def respond_async(multi, history, model_display, sess, custom_prompt):
                 sess.stop_event.clear()
                 return
             for d in done:
-                chunk = d.result()
-                if chunk is None:
+                result = d.result()
+                if result is None:
                     raise StopAsyncIteration
-                if not first_meaningful_chunk_found:
-                    if chunk.strip():
-                        history[-1][1] = chunk
-                        first_meaningful_chunk_found = True
-                else:
-                    history[-1][1] += chunk
+                action, text = result
+                history[-1][1] = text
                 yield history, gr.update(interactive=False, submit_btn=False, stop_btn=True), sess
     except StopAsyncIteration:
         pass