app.py

import gradio as gr
import torch
import unicodedata
import re
import numpy as np
from pathlib import Path
from transformers import AutoTokenizer, AutoModel
from sklearn.feature_extraction.text import HashingVectorizer
from sklearn.preprocessing import normalize as sk_normalize
import chromadb
import joblib
import pickle
import scipy.sparse
import textwrap
import os

# --------------------------- CONFIG -----------------------------------
DB_DIR = Path("./chroma_db_greekbertChatbotVol106")
ASSETS_DIR = Path("./assets")
STATIC_PDF_DIR = Path("./static_pdfs")
STATIC_PDF_DIR_NAME = "static_pdfs"

COL_NAME = "dataset14_grbert_charword"
MODEL_NAME = "sentence-transformers/paraphrase-xlm-r-multilingual-v1"
CHUNK_SIZE = 512
ALPHA_BASE = 0.2
ALPHA_LONGQ = 0.5
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

print(f"Running on device: {DEVICE}")

# ----------------------- PRE-/POST HELPERS ----------------------------
def strip_acc(s: str) -> str:
    return ''.join(ch for ch in unicodedata.normalize('NFD', s)
                   if not unicodedata.combining(ch))

STOP = {"σχετικο", "σχετικα", "με", "και"}

def preprocess(txt: str) -> str:
    txt = strip_acc(txt.lower())
    txt = re.sub(r"[^a-zα-ω0-9 ]", " ", txt)
    txt = re.sub(r"\s+", " ", txt).strip()
    return " ".join(w for w in txt.split() if w not in STOP)

def cls_embed(texts, tok, model):
    out = []
    enc = tok(texts, padding=True, truncation=True,
              max_length=CHUNK_SIZE, return_tensors="pt").to(DEVICE)
    with torch.no_grad():
        hs = model(**enc, output_hidden_states=True).hidden_states
        cls = torch.stack(hs[-4:],0).mean(0)[:,0,:]
        cls = torch.nn.functional.normalize(cls, p=2, dim=1)
        out.append(cls.cpu())
    return torch.cat(out).numpy()

# ---------------------- LOAD MODELS & DATA (Μία φορά κατά την εκκίνηση) --------------------
print("⏳ Loading Model and Tokenizer...")
try:
    tok = AutoTokenizer.from_pretrained(MODEL_NAME)
    model = AutoModel.from_pretrained(MODEL_NAME).to(DEVICE).eval()
    print("✓ Model and tokenizer loaded.")
except Exception as e:
    print(f"CRITICAL ERROR loading model/tokenizer: {e}")
    raise

print("⏳ Loading TF-IDF vectorizers and SPARSE matrices...")
try:
    char_vec = joblib.load(ASSETS_DIR / "char_vectorizer.joblib")
    word_vec = joblib.load(ASSETS_DIR / "word_vectorizer.joblib")
    X_char = scipy.sparse.load_npz(ASSETS_DIR / "X_char_sparse.npz")
    X_word = scipy.sparse.load_npz(ASSETS_DIR / "X_word_sparse.npz")
    print("✓ TF-IDF components loaded (sparse matrices).")
    print(f"   → X_char shape: {X_char.shape}, type: {type(X_char)}")
    print(f"   → X_word shape: {X_word.shape}, type: {type(X_word)}")
except Exception as e:
    print(f"CRITICAL ERROR loading TF-IDF components: {e}")
    raise

print("⏳ Loading chunk data (pre_chunks, raw_chunks, ids, metas)...")
try:
    with open(ASSETS_DIR / "pre_chunks.pkl", "rb") as f:
        pre_chunks = pickle.load(f)
    with open(ASSETS_DIR / "raw_chunks.pkl", "rb") as f:
        raw_chunks = pickle.load(f)
    with open(ASSETS_DIR / "ids.pkl", "rb") as f:
        ids = pickle.load(f)
    with open(ASSETS_DIR / "metas.pkl", "rb") as f:
        metas = pickle.load(f)
    print(f"✓ Chunk data loaded. Total chunks from ids: {len(ids):,}")
    if not all([pre_chunks, raw_chunks, ids, metas]):
        print("WARNING: One or more chunk data lists are empty!")
except Exception as e:
    print(f"CRITICAL ERROR loading chunk data: {e}")
    raise

print("⏳ Connecting to ChromaDB...")
try:
    client = chromadb.PersistentClient(path=str(DB_DIR.resolve()))
    col = client.get_collection(COL_NAME)
    print(f"✓ Connected to ChromaDB. Collection '{COL_NAME}' count: {col.count()}")
    if col.count() == 0:
        print("WARNING: ChromaDB collection is empty or not found correctly!")
except Exception as e:
    print(f"CRITICAL ERROR connecting to ChromaDB or getting collection: {e}")
    print(f"Attempted DB path for PersistentClient: {str(DB_DIR.resolve())}")
    print("Ensure the ChromaDB directory structure is correct in your Hugging Face Space repository.")
    raise

# ---------------------- HYBRID SEARCH (Κύρια Λογική) ---------------------------------
def hybrid_search_gradio(query, k=5):
    if not query.strip():
        return "Παρακαλώ εισάγετε μια ερώτηση."

    if not ids:
        return "Σφάλμα: Τα δεδομένα αναζήτησης (ids) δεν έχουν φορτωθεί. Επικοινωνήστε με τον διαχειριστή."

    q_pre = preprocess(query)
    words = q_pre.split()
    alpha = ALPHA_LONGQ if len(words) > 30 else ALPHA_BASE

    exact_ids_set = {ids[i] for i, t in enumerate(pre_chunks) if q_pre in t}

    q_emb_np = cls_embed([q_pre], tok, model)
    q_emb_list = q_emb_np.tolist()

    try:
        sem_results = col.query(
            query_embeddings=q_emb_list,
            n_results=min(k * 30, len(ids)),
            include=["distances", "metadatas", "documents"]
        )
    except Exception as e:
        print(f"ERROR during ChromaDB query: {e}")
        return "Σφάλμα κατά την σημασιολογική αναζήτηση."

    sem_sims = {doc_id: 1 - dist for doc_id, dist in zip(sem_results["ids"][0], sem_results["distances"][0])}

    q_char_sparse = char_vec.transform([q_pre])
    q_char_normalized = sk_normalize(q_char_sparse)
    char_sim_scores = (q_char_normalized @ X_char.T).toarray().flatten()

    q_word_sparse = word_vec.transform([q_pre])
    q_word_normalized = sk_normalize(q_word_sparse)
    word_sim_scores = (q_word_normalized @ X_word.T).toarray().flatten()

    lex_sims = {}
    for idx, (c_score, w_score) in enumerate(zip(char_sim_scores, word_sim_scores)):
        if c_score > 0 or w_score > 0:
            if idx < len(ids):
                lex_sims[ids[idx]] = 0.85 * c_score + 0.15 * w_score
            else:
                print(f"Warning: Lexical score index {idx} out of bounds for ids list (len: {len(ids)}).")

    all_chunk_ids_set = set(sem_sims.keys()) | set(lex_sims.keys()) | exact_ids_set
    scored = []
    for chunk_id_key in all_chunk_ids_set:
        s = alpha * sem_sims.get(chunk_id_key, 0.0) + \
            (1 - alpha) * lex_sims.get(chunk_id_key, 0.0)
        if chunk_id_key in exact_ids_set:
            s = 1.0
        scored.append((chunk_id_key, s))

    scored.sort(key=lambda x: x[1], reverse=True)

    hits_output = []
    seen_doc_main_ids = set()
    for chunk_id_val, score_val in scored:
        try:
            idx_in_lists = ids.index(chunk_id_val)
        except ValueError:
            print(f"Warning: chunk_id '{chunk_id_val}' from search results not found in global 'ids' list. Skipping.")
            continue

        doc_meta = metas[idx_in_lists]
        doc_main_id = doc_meta['id']

        if doc_main_id in seen_doc_main_ids:
            continue

        original_url_from_meta = doc_meta.get('url', '#')
        
        # *** ΕΝΑΡΞΗ ΤΡΟΠΟΠΟΙΗΜΕΝΟΥ/ΝΕΟΥ ΚΩΔΙΚΑ ΓΙΑ PDF DEBUGGING ***
        pdf_serve_url = "#" 
        pdf_filename_display = "N/A"
        pdf_filename_extracted = None # Αρχικοποίηση

        if original_url_from_meta and original_url_from_meta != '#':
            pdf_filename_extracted = os.path.basename(original_url_from_meta)
            print(f"--- Debug: Original URL: {original_url_from_meta}, Initial Extracted filename: {pdf_filename_extracted}")

            # --- ΠΡΟΣΩΡΙΝΟΣ ΚΩΔΙΚΑΣ ΓΙΑ ΔΟΚΙΜΗ ASCII FILENAME (Μπορείτε να τον ενεργοποιήσετε αφαιρώντας τα σχόλια) ---
            # TARGET_ORIGINAL_FILENAME_FOR_TEST = "6ΑΤΘ469Β7Η-963.pdf" # Το αρχικό ελληνικό όνομα που είχατε μετονομάσει
            # ASCII_TEST_FILENAME = "testfileGR.pdf"                 # Το νέο ASCII όνομα που βάλατε στο static_pdfs
            #
            # if pdf_filename_extracted == TARGET_ORIGINAL_FILENAME_FOR_TEST:
            #     print(f"--- INFO: ASCII Filename Test Active ---")
            #     print(f"--- Original filename was: {pdf_filename_extracted}")
            #     print(f"--- Temporarily using: {ASCII_TEST_FILENAME} for linking and checking existence.")
            #     pdf_filename_extracted = ASCII_TEST_FILENAME 
            # --- ΤΕΛΟΣ ΠΡΟΣΩΡΙΝΟΥ ΚΩΔΙΚΑ ASCII ---

            if pdf_filename_extracted and pdf_filename_extracted.lower().endswith(".pdf"):
                potential_pdf_path_on_server = STATIC_PDF_DIR / pdf_filename_extracted
                
                print(f"--- Debug: Final pdf_filename_extracted to check: {pdf_filename_extracted}")
                print(f"--- Debug: Checking for PDF at server path: {potential_pdf_path_on_server.resolve()}")

                if potential_pdf_path_on_server.exists() and potential_pdf_path_on_server.is_file():
                    print(f"--- Debug: Path.exists() and Path.is_file() are TRUE for {potential_pdf_path_on_server.resolve()}. Attempting to open...")
                    try:
                        # Προσπάθεια ανοίγματος του αρχείου σε binary read mode και ανάγνωσης ενός byte
                        with open(potential_pdf_path_on_server, "rb") as f_test_access:
                            f_test_access.read(1) 
                        print(f"--- Debug: Successfully opened and read a byte from: {potential_pdf_path_on_server.resolve()}")
                        
                        pdf_serve_url = f"/file/{STATIC_PDF_DIR_NAME}/{pdf_filename_extracted}" 
                        pdf_filename_display = pdf_filename_extracted

                    except Exception as e_file_access:
                        print(f"!!! CRITICAL ERROR trying to open/read file {potential_pdf_path_on_server.resolve()}: {e_file_access}")
                        pdf_filename_display = "Error accessing file content" # Ενημέρωση για εμφάνιση
                else:
                    print(f"--- Debug: Path.exists() or Path.is_file() is FALSE for {potential_pdf_path_on_server.resolve()}")
                    pdf_filename_display = "File not found by script"
            else: 
                if not pdf_filename_extracted: # Αν το pdf_filename_extracted κατέληξε κενό
                     print(f"--- Debug: pdf_filename_extracted is empty or None after os.path.basename or ASCII test.")
                else: # Αν δεν έχει επέκταση .pdf
                     print(f"--- Debug: Extracted filename '{pdf_filename_extracted}' does not end with .pdf")
                pdf_filename_display = "Not a valid PDF link"
        else: # original_url_from_meta ήταν κενό ή '#'
            print(f"--- Debug: No valid original_url_from_meta found. URL was: '{original_url_from_meta}'")
            pdf_filename_display = "No source URL"
        
        # *** ΤΕΛΟΣ ΤΡΟΠΟΠΟΙΗΜΕΝΟΥ/ΝΕΟΥ ΚΩΔΙΚΑ ΓΙΑ PDF DEBUGGING ***

        hits_output.append({
            "score": score_val,
            "title": doc_meta.get('title', 'N/A'),
            "snippet": raw_chunks[idx_in_lists][:500] + " ...",
            "original_url_meta": original_url_from_meta, 
            "pdf_serve_url": pdf_serve_url, 
            "pdf_filename_display": pdf_filename_display 
        })
        seen_doc_main_ids.add(doc_main_id)
        if len(hits_output) >= k:
            break

    if not hits_output:
        return "Δεν βρέθηκαν σχετικά αποτελέσματα."

    output_md = f"Βρέθηκαν **{len(hits_output)}** σχετικά αποτελέσματα:\n\n"
    for hit in hits_output:
        output_md += f"### {hit['title']} (Score: {hit['score']:.3f})\n"
        snippet_wrapped = textwrap.fill(hit['snippet'].replace("\n", " "), width=100)
        output_md += f"**Απόσπασμα:** {snippet_wrapped}\n"

        if hit['pdf_serve_url'] and hit['pdf_serve_url'] != '#':
            output_md += f"**Πηγή (PDF):** <a href='{hit['pdf_serve_url']}' target='_blank'>{hit['pdf_filename_display']}</a>\n"
        elif hit['original_url_meta'] and hit['original_url_meta'] != '#':
            output_md += f"**Πηγή (αρχικό):** [{hit['original_url_meta']}]({hit['original_url_meta']})\n"
        output_md += "---\n"



 # ΠΡΟΣΩΡΙΝΗ ΠΡΟΣΘΗΚΗ ΓΙΑ ΔΟΚΙΜΗ TXT ΑΡΧΕΙΟΥ
    output_md += "\n\n---\n**Δοκιμαστικός Σύνδεσμος Κειμένου:** <a href='/file/static_pdfs/test_text_file.txt' target='_blank'>Άνοιγμα test_text_file.txt</a>\n"
    return output_md
    

# ---------------------- GRADIO INTERFACE -----------------------------------
print("🚀 Launching Gradio Interface...")
iface = gr.Interface(
    fn=hybrid_search_gradio,
    inputs=gr.Textbox(lines=3, placeholder="Γράψε την ερώτησή σου εδώ...", label="Ερώτηση προς τον βοηθό:"),
    outputs=gr.Markdown(label="Απαντήσεις από τα έγγραφα:", rtl=False, sanitize_html=False),
    title="🏛️ Ελληνικό Chatbot Υβριδικής Αναζήτησης (v1.0.9)", # Νέα έκδοση για παρακολούθηση
    description="Πληκτρολογήστε την ερώτησή σας για αναζήτηση στα διαθέσιμα έγγραφα. Η αναζήτηση συνδυάζει σημασιολογική ομοιότητα (GreekBERT) και ομοιότητα λέξεων/χαρακτήρων (TF-IDF).\nΧρησιμοποιεί το μοντέλο: sentence-transformers/paraphrase-xlm-r-multilingual-v1.\nΤα PDF ανοίγουν σε νέα καρτέλα.",
    allow_flagging="never",
    examples=[
        ["Ποια είναι τα μέτρα για τον κορονοϊό;", 5],
        ["Πληροφορίες για άδεια ειδικού σκοπού", 3],
        ["Τι προβλέπεται για τις μετακινήσεις εκτός νομού;", 5]
    ],
)

if __name__ == '__main__':
    # Παραλλαγή 2
    # STATIC_PDF_DIR ορίζεται στην αρχή του αρχείου ως Path("./static_pdfs")
    iface.launch(allowed_paths=[str(STATIC_PDF_DIR.resolve())])