Update app.py

app.py

@@ -1,24 +1,22 @@
 #!/usr/bin/env python3
 """
-HF Space API for Article Q&A AI.
-Optimized for CPU / Free Tier.
-Uses tiny-flan-t5 for faster generation.
+Hugging Face Space app: Article Q&A AI
+Simplified, CPU-friendly, public model (google/flan-t5-small)
 """
 
 import os
 import json
-import faiss
 import numpy as np
+import faiss
+import gradio as gr
 from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
-from fastapi import FastAPI
-from pydantic import BaseModel
 
 # ---- Config ----
 CHUNK_SIZE = 500
 CHUNK_OVERLAP = 100
 JSON_FILE = "articles.json"
-TOP_K = 3  # fewer chunks for speed
+TOP_K = 4
 SERVER_PORT = 7860
 
 # ---- Global variables ----
@@ -28,12 +26,10 @@ INDEX_DIM = None
 
 # ---- Models ----
 embed_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
-gen_model_name = "sshleifer/tiny-flan-t5"
+gen_model_name = "google/flan-t5-small"
 tokenizer = AutoTokenizer.from_pretrained(gen_model_name)
 gen_model = AutoModelForSeq2SeqLM.from_pretrained(gen_model_name)
-gen_pipeline = pipeline(
-    "text2text-generation", model=gen_model, tokenizer=tokenizer, device=-1
-)
+gen_pipeline = pipeline("text2text-generation", model=gen_model, tokenizer=tokenizer, device=-1)
 
 # ---- Helpers ----
 def chunk_text(text):
@@ -43,17 +39,19 @@ def chunk_text(text):
         end = min(start + CHUNK_SIZE, len(text))
         chunks.append(text[start:end])
         start = end - CHUNK_OVERLAP
-        if start < 0:
-            start = 0
-        if start >= len(text):
-            break
+        if start < 0: start = 0
+        if start >= len(text): break
     return chunks
 
 def build_index_in_memory():
     print("🚀 Building FAISS index...")
+    print("Current WORKDIR:", os.getcwd())
+    print("Files:", os.listdir("."))
+
     if not os.path.exists(JSON_FILE):
         print("❌ articles.json missing")
         return None, None, None
+
     try:
         with open(JSON_FILE, "r", encoding="utf-8") as f:
             articles = json.load(f)
@@ -68,9 +66,8 @@ def build_index_in_memory():
     embeddings_list, texts, metas = [], [], []
 
     for art_id, art in enumerate(articles):
-        content = art.get("Continut") or art.get("continut") or ""
-        url = art.get("URL") or art.get("url") or ""
-        title = art.get("Titlu") or art.get("titlu") or f"articol_{art_id}"
+        content = art.get("continut") or art.get("Continut") or ""
+        url = art.get("url") or art.get("URL") or ""
         if not content.strip():
             continue
         chunks = chunk_text(content)
@@ -81,7 +78,7 @@ def build_index_in_memory():
             embs = embs.reshape(1, -1)
         embeddings_list.append(embs)
         texts.extend(chunks)
-        metas.extend([{"source": title, "url": url, "chunk_id": i} for i in range(len(chunks))])
+        metas.extend([{"source": art.get("titlu") or art.get("Titlu") or f"articol_{art_id}", "url": url, "chunk_id": i} for i in range(len(chunks))])
 
     if not embeddings_list:
         print("❌ No valid chunks found")
@@ -92,12 +89,14 @@ def build_index_in_memory():
     d = embeddings.shape[1]
     index = faiss.IndexFlatIP(d)
     index.add(embeddings)
+
     metadata = {"texts": texts, "metas": metas}
     print(f"✅ Index built with {len(texts)} chunks")
     return index, metadata, d
 
-def ask_question(question, top_k=TOP_K, max_answer_tokens=64):
+def ask_question(question, top_k=TOP_K, max_answer_tokens=256):
     global INDEX, METADATA, INDEX_DIM
+
     if not question.strip():
         return "⚠️ Please provide a question."
 
@@ -142,17 +141,15 @@ def ask_question(question, top_k=TOP_K, max_answer_tokens=64):
 
     return f"{out} Find out more at {', '.join([u for u in urls if u])}"
 
-# ---- FastAPI ----
-app = FastAPI()
-
-class Question(BaseModel):
-    text: str
-
-@app.post("/ask")
-def ask(q: Question):
-    return {"answer": ask_question(q.text)}
+def main():
+    print("🚀 Starting Article Q&A AI...")
+    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=SERVER_PORT)
 
 if __name__ == "__main__":
-    import uvicorn
-    INDEX, METADATA, INDEX_DIM = build_index_in_memory()
-    uvicorn.run(app, host="0.0.0.0", port=SERVER_PORT)
+    main()