Update app.py

app.py

@@ -3,13 +3,10 @@ import json
 import faiss
 import pickle
 import numpy as np
-from fastapi import FastAPI, HTTPException
-from pydantic import BaseModel
+import gradio as gr
 from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
 
-app = FastAPI(title="MetaGPT AI - Local Q&A")
-
 # Config
 DATA_DIR = "data"
 INDEX_FILE = os.path.join(DATA_DIR, "index.faiss")
@@ -20,7 +17,7 @@ JSON_FILE = "articles.json"
 
 os.makedirs(DATA_DIR, exist_ok=True)
 
-# Load models
+# Modele
 embed_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
 gen_model_name = "google/flan-t5-small"
 tokenizer = AutoTokenizer.from_pretrained(gen_model_name)
@@ -52,12 +49,10 @@ def load_index():
         metadata = pickle.load(f)
     return index, metadata
 
-# ---- Build / Rebuild index from JSON ----
-@app.post("/build_index")
 def build_index():
     if not os.path.exists(JSON_FILE):
-        raise HTTPException(status_code=404, detail=f"{JSON_FILE} not found")
-    
+        return None, None
+
     with open(JSON_FILE, "r", encoding="utf-8") as f:
         articles = json.load(f)
 
@@ -79,37 +74,34 @@ def build_index():
     metadata = {"texts": texts, "metas": metas}
 
     save_index(index, metadata)
-    return {"status": "ok", "total_chunks": len(texts)}
-
-# ---- Ask endpoint ----
-class AskRequest(BaseModel):
-    question: str
-    top_k: int = 4
-    max_answer_tokens: int = 256
+    return index, metadata
 
-@app.post("/ask")
-def ask(req: AskRequest):
+def ask_question(question, top_k=4, max_answer_tokens=256):
     index, metadata = load_index()
     if index is None:
-        raise HTTPException(status_code=404, detail="No index found. Call /build_index first.")
-    
-    q_emb = embed_model.encode([req.question], convert_to_numpy=True).astype("float32")
+        index, metadata = build_index()
+        if index is None:
+            return "Error: articles.json not found."
+
+    q_emb = embed_model.encode([question], convert_to_numpy=True).astype("float32")
     faiss.normalize_L2(q_emb)
-    D, I = index.search(q_emb, req.top_k)
+    D, I = index.search(q_emb, top_k)
 
     retrieved = [metadata["texts"][i] for i in I[0]]
     urls = [metadata["metas"][i]["url"] for i in I[0] if "url" in metadata["metas"][i]]
 
     context = "\n\n".join(retrieved)
-    prompt = f"Context:\n{context}\n\nQuestion: {req.question}\nAnswer:"
-    out = gen_pipeline(prompt, max_length=req.max_answer_tokens, do_sample=False)[0]["generated_text"]
-
-    return {
-        "answer": f"{out} Find out more at {', '.join(urls)}",
-        "sources": [metadata["metas"][i] for i in I[0]]
-    }
-
-# ---- Health check ----
-@app.get("/health")
-def health():
-    return {"status": "ok"}
+    prompt = f"Context:\n{context}\n\nQuestion: {question}\nAnswer:"
+    out = gen_pipeline(prompt, max_length=max_answer_tokens, do_sample=False)[0]["generated_text"]
+
+    return f"{out} Find out more at {', '.join(urls)}"
+
+# Gradio UI
+iface = gr.Interface(
+    fn=ask_question,
+    inputs=[gr.Textbox(label="Întrebare")],
+    outputs=[gr.Textbox(label="Răspuns")],
+    live=False,
+)
+
+iface.launch(server_name="0.0.0.0", server_port=7860)