Update Space

requirements.txt

@@ -0,0 +1,5 @@
+torch
+transformers
+flax
+duckduckgo-search
+flask

yuuvx_server.py

@@ -0,0 +1,50 @@
+from flask import Flask, request, jsonify
+from duckduckgo_search import DDGS
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+app = Flask(__name__)
+
+MODEL_NAME = "flax-community/gpt2-medium-indonesian"
+WEB_RESULTS_TO_USE = 2
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model.to(device)
+
+def web_search_snippets(query,max_results=WEB_RESULTS_TO_USE):
+    snippets=[]
+    try:
+        with DDGS() as ddgs:
+            results=list(ddgs.text(query,max_results=max_results))
+        for r in results:
+            body=r.get("body","").strip()
+            title=r.get("title","").strip()
+            if body:
+                snippets.append(f"{title}. {body}" if title else body)
+    except:
+        snippets.append("(gagal ambil web)")
+    if not snippets:
+        snippets=["(tidak ada info web)"]
+    return snippets
+
+def generate_reply(user_input):
+    web_snippets = web_search_snippets(user_input)
+    prompt=f"Aku YuuVx Ai, asisten bahasa Indonesia. Gunakan info web jika relevan.\n{''.join(web_snippets)}\nPertanyaan: {user_input}\nJawab singkat, jelas.\n"
+    input_ids = tokenizer.encode(prompt+tokenizer.eos_token,return_tensors="pt").to(device)
+    out = model.generate(input_ids,max_new_tokens=200,do_sample=True,temperature=0.7,top_p=0.9,pad_token_id=tokenizer.eos_token_id)
+    full = tokenizer.decode(out[0],skip_special_tokens=True)
+    answer = full.split(prompt)[-1].strip() if prompt in full else full
+    return answer
+
+@app.route("/ask",methods=["POST"])
+def ask():
+    data=request.json
+    user_input=data.get("question","")
+    reply=generate_reply(user_input)
+    return jsonify({"reply":reply})
+
+if __name__=="__main__":
+    app.run(host="0.0.0.0",port=5000)
+