Update app.py

app.py

@@ -1,27 +1,79 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
-import torch
+from huggingface_hub import InferenceClient
+import os
 
 # ============================================================
-# ১. মডেল লোড (আপনার mx-llms/BLM)
+# ১. মডেল সেটআপ
 # ============================================================
-model = AutoModelForCausalLM.from_pretrained(
-    "mx-llms/BLM",
-    trust_remote_code=True,
-    device_map="auto",
-    torch_dtype=torch.float16,
-)
-tokenizer = AutoTokenizer.from_pretrained("mx-llms/BLM", trust_remote_code=True)
+model_id = os.getenv("MODEL_ID", "mahin1234/lychee-gpt")
+
+def respond(
+    message,
+    history: list[dict[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    hf_token: gr.OAuthToken | None,
+):
+    """
+    Lychee GPT - Hugging Face Inference API ব্যবহার করে স্ট্রিমিং রেসপন্স
+    """
+    # টোকেন নির্বাচন (লগইন করলে OAuth, নইলে Secret)
+    if hf_token is not None:
+        token = hf_token.token
+    else:
+        token = os.getenv("HF_TOKEN")
+        if token is None:
+            yield "⚠️ Please login or set HF_TOKEN in Space Secrets."
+            return
+
+    client = InferenceClient(token=token, model=model_id)
 
-def respond(message, history):
-    messages = [{"role": "user", "content": message}]
-    prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-    inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
-    with torch.no_grad():
-        outputs = model.generate(**inputs, max_new_tokens=100)
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+    messages = [{"role": "system", "content": system_message}]
+    messages.extend(history)
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+    try:
+        for msg in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            token_text = msg.choices[0].delta.content or ""
+            response += token_text
+            yield response
+    except Exception as e:
+        yield f"❌ Error: {str(e)}. Check logs or model availability."
 
 # ============================================================
 # ২. Gradio UI
 # ============================================================
-gr.ChatInterface(fn=respond, title="BLM AI Assistant").launch()
+chatbot = gr.ChatInterface(
+    respond,
+    additional_inputs=[
+        gr.Textbox(
+            value="You are Lychee GPT, a helpful AI assistant created by MX LLMS and MD Mushfiqur Rahim.",
+            label="System message"
+        ),
+        gr.Slider(minimum=1, maximum=4096, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p"),
+    ],
+    title="🍈 Lychee GPT",
+    description="Created by MD Mushfiqur Rahim & MX LLMS. Powered by Hugging Face Inference API.",
+)
+
+# ============================================================
+# ৩. ডেমো (Login Button সহ)
+# ============================================================
+with gr.Blocks(title="Lychee GPT") as demo:
+    with gr.Sidebar():
+        gr.LoginButton()
+    chatbot.render()
+
+if __name__ == "__main__":
+    demo.launch()