hf

app.py

@@ -44,7 +44,7 @@ class GLM45Model:
         if reasoning_content and not skip_think:
             reasoning_escaped = html.escape(reasoning_content).replace("\n", "<br>")
             think_html = (
-                    "<details open><summary style='cursor:pointer;font-weight:bold;color:#007acc;'>🤔 Thinking</summary>"
+                    "<details open><summary style='cursor:pointer;font-weight:bold;color:#007acc;'>Thinking</summary>"
                     "<div style='color:#555555;line-height:1.6;padding:15px;border-left:4px solid #007acc;margin:10px 0;background-color:#f0f7ff;border-radius:4px;'>"
                     + reasoning_escaped +
                     "</div></details>"
@@ -61,15 +61,16 @@ class GLM45Model:
     def _build_messages(self, raw_hist, sys_prompt):
         msgs = []
         if sys_prompt.strip():
-            msgs.append({"role": "system", "content": [{"type": "text", "text": sys_prompt.strip()}]})
+            msgs.append({"role": "system", "content": sys_prompt.strip()})
 
         for h in raw_hist:
             if h["role"] == "user":
-                msgs.append({"role": "user", "content": [{"type": "text", "text": h["content"]}]})
+                msgs.append({"role": "user", "content": h["content"]})
             else:
-                clean_content = html.escape(h["content"]).replace("<br>", "\n")
-                if clean_content:
-                    msgs.append({"role": "assistant", "content": [{"type": "text", "text": clean_content}]})
+                msg = {"role": "assistant", "content": h.get("content", "")}
+                if h.get("reasoning_content"):
+                    msg["reasoning_content"] = h.get("reasoning_content")
+                msgs.append(msg)
         return msgs
 
     def stream_generate(self, raw_hist, sys_prompt, thinking_enabled=True, temperature=1.0):
@@ -109,7 +110,11 @@ def chat(msg, raw_hist, sys_prompt, thinking_enabled, temperature):
         raw_hist = []
 
     raw_hist.append({"role": "user", "content": msg.strip()})
-    place = {"role": "assistant", "content": ""}
+    place = {
+        "role": "assistant",
+        "content": "",
+        "reasoning_content": ""
+    }
     raw_hist.append(place)
 
     yield raw_hist, copy.deepcopy(raw_hist), ""
@@ -118,10 +123,15 @@ def chat(msg, raw_hist, sys_prompt, thinking_enabled, temperature):
         for chunk in glm45.stream_generate(raw_hist[:-1], sys_prompt, thinking_enabled, temperature):
             if stop_generation:
                 break
-            place["content"] = chunk
+
+            place["content"] = glm45.accumulated_content
+            place["reasoning_content"] = glm45.accumulated_reasoning
+            place["display_content"] = chunk
+
             yield raw_hist, copy.deepcopy(raw_hist), ""
     except Exception as e:
-        place["content"] = f"<div style='color: red;'>Error: {html.escape(str(e))}</div>"
+        place["content"] = f"Error: {str(e)}"
+        place["display_content"] = f"<div style='color: red;'>Error: {html.escape(str(e))}</div>"
         yield raw_hist, copy.deepcopy(raw_hist), ""
 
 
@@ -132,6 +142,17 @@ def reset():
     return [], [], ""
 
 
+def format_history_for_display(raw_hist):
+    display_hist = []
+    for msg in raw_hist:
+        if msg["role"] == "user":
+            display_hist.append({"role": "user", "content": msg["content"]})
+        else:
+            content = msg.get("display_content", msg.get("content", ""))
+            display_hist.append({"role": "assistant", "content": content})
+    return display_hist
+
+
 demo = gr.Blocks(title="GLM-4.5 API Demo", theme=gr.themes.Soft())
 
 with demo:
@@ -139,8 +160,8 @@ with demo:
         "<div style='text-align:center;font-size:32px;font-weight:bold;margin-bottom:10px;'>GLM-4.5 API Demo</div>"
         "<div style='text-align:center;color:red;font-size:16px;margin-bottom:20px;'>"
         "This demo uses the API version of the service for faster response speeds.<br>"
-        "Only chat functionality is supported. For tool usage, MCP support, and web search, please refer to the API documentation.</div>"
-        "<div style='text-align:center;'><a href='https://huggingface.co/collections/zai-org/glm-45-687c621d34bda8c9e4bf503b'>Model</a> | "
+        "Only chat functionality with 64K token length is supported. For tool usage, MCP support, and web search, please refer to the API documentation.</div>"
+        "<div style='text-align:center;'><a href='https://modelscope.cn/collections/GLM-45-b8693e2a08984f'>Model</a> | "
         "<a href='https://github.com/zai-org/GLM-4.5'>Github</a> | "
         "<a href='http://z.ai/blog/glm-4.5'>Blog</a> | "
         "<a href='https://docs.bigmodel.cn/cn/guide/models/text/glm-4.5'>API Docs</a></div>"
@@ -179,13 +200,20 @@ with demo:
             )
             sys = gr.Textbox(label="System Prompt", lines=6)
 
+
+    def chat_wrapper(msg, raw_hist, sys_prompt, thinking_enabled, temperature):
+        for hist, raw_hist_updated, textbox_value in chat(msg, raw_hist, sys_prompt, thinking_enabled, temperature):
+            display_hist = format_history_for_display(hist)
+            yield display_hist, raw_hist_updated, textbox_value
+
+
     send.click(
-        chat,
+        chat_wrapper,
         inputs=[textbox, raw_history, sys, thinking_toggle, temperature_slider],
         outputs=[chatbox, raw_history, textbox]
     )
     textbox.submit(
-        chat,
+        chat_wrapper,
         inputs=[textbox, raw_history, sys, thinking_toggle, temperature_slider],
         outputs=[chatbox, raw_history, textbox]
     )