chat

README.md

@@ -1,25 +1,12 @@
 ---
-title: OpenRouter Chat
-short_description: A Gradio interface over the OpenRouter chat API.
-emoji: 🤖
+title: harpertoken gpt-oss-120b
+emoji: 👋
 colorFrom: gray
-colorTo: gray
+colorTo: indigo
 sdk: gradio
-sdk_version: 6.14.0
+sdk_version: 5.49.1
 app_file: app.py
-pinned: false
----
-
-# OpenRouter Chat
-
-A Gradio interface over the [OpenRouter](https://openrouter.ai) chat API.
-
-All generation is done by OpenRouter. This Space does not load a model from
-`harpertoken`; it only forwards your message and renders the reply.
-
-## Setup
-
-1. Create an API key at [openrouter.ai](https://openrouter.ai)
-2. Go to **Settings → Variables and secrets**
-3. Add a secret named `OPENROUTER_API_KEY` with that key
-4. Rebuild the Space
+pinned: true
+license: mit
+short_description: 120b gpt model with cerebras speeds
+---

app.py

@@ -2,92 +2,75 @@ import gradio as gr
 import os
 import re
 from dotenv import load_dotenv
-from openai import OpenAI
+from cerebras.cloud.sdk import Cerebras
 
+# Load environment variables
 load_dotenv()
 
-client = OpenAI(
-    base_url="https://openrouter.ai/api/v1",
-    api_key=os.environ.get("OPENROUTER_API_KEY"),
+client = Cerebras(
+    api_key=os.environ.get("CEREBRAS_API_KEY")
 )
 
-def clean_response(text: str) -> str:
+def clean_response(text):
     lines = []
     for line in text.splitlines():
+        # Remove bullets (*, #, •, -) at line start
         line = re.sub(r"^\s*[*#•\-]+\s*", "", line)
+        # Remove markdown bold
         line = re.sub(r"\*\*(.*?)\*\*", r"\1", line)
-        # Only strip double underscores around a phrase. A single token like
-        # __init__ is a Python dunder, which shares the exact shape of
-        # __underline__ and cannot be told apart by position alone.
-        line = re.sub(r"(?<!\w)__([^_\n]*?\s[^_\n]*?)__(?!\w)", r"\1", line)
+        line = re.sub(r"__(.*?)__", r"\1", line)
         lines.append(line)
     return "\n".join(lines)
 
-MAX_SENT_TURNS = 20
-
-
-def chat(user_message: str, history: list):
-    history = history or []
-
-    # Cap what is sent, not what is shown: unbounded history eventually exceeds
-    # the model's context window, and the failure then looks like a model error.
-    recent = history[-(MAX_SENT_TURNS * 2):]
-    messages = [{"role": "system", "content": "You are a helpful assistant."}]
-    for msg in recent:
-        messages.append({"role": msg["role"], "content": msg["content"]})
-    messages.append({"role": "user", "content": user_message})
-
-    history = history + [{"role": "user", "content": user_message}]
-
+def chat(user_message, history):
     try:
-        # openrouter/free automatically picks a working free model
         stream = client.chat.completions.create(
-            model="openrouter/free",
-            messages=messages,
+            messages=[
+                {"role": "system", "content": "You are a helpful assistant."},
+                {"role": "user", "content": user_message}
+            ],
+            model="gpt-oss-120b",
             stream=True,
-            temperature=0.7,
-            max_tokens=4096,
+            max_completion_tokens=65536,
+            temperature=1,
+            top_p=1,
+            reasoning_effort="medium"
         )
 
         response_text = ""
-        history = history + [{"role": "assistant", "content": ""}]
         for chunk in stream:
-            if chunk.choices and chunk.choices[0].delta.content:
+            if chunk.choices[0].delta.content:
                 response_text += chunk.choices[0].delta.content
-                history[-1]["content"] = response_text
-                yield history
 
-        history[-1]["content"] = clean_response(response_text)
-        yield history
+        response_text = clean_response(response_text)
+
+        history = history or []
+        history.append((user_message, response_text))
+        return history
 
     except Exception as e:
-        history.append({
-            "role": "assistant",
-            "content": f"Error: {str(e)}"
-        })
-        yield history
+        history = history or []
+        history.append((user_message, f"Error: {str(e)}"))
+        return history
+
+with gr.Blocks() as demo:
+    gr.Markdown("## Chat UI (Long responses stay open)")
 
-with gr.Blocks(title="OpenRouter Chat") as demo:
-    gr.Markdown("## OpenRouter Chat")
-    
     chatbot = gr.Chatbot(
         label="Conversation",
-        height=600
+        height=600,
+        show_copy_button=True
     )
-    
+
     with gr.Row():
         msg = gr.Textbox(
             label="Ask something",
             placeholder="Type your prompt and press Enter",
-            scale=4,
-            container=False
+            scale=4
         )
-        clear = gr.Button("Clear", scale=1)
+        clear = gr.Button("Clear")
 
-    msg.submit(chat, [msg, chatbot], chatbot).then(
-        lambda: "", None, msg
-    )
+    msg.submit(chat, [msg, chatbot], chatbot)
     clear.click(lambda: [], None, chatbot)
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()

requirements.txt

@@ -1,2 +1,3 @@
-openai
+gradio==4.44.0
+cerebras-cloud-sdk
 python-dotenv