import gradio as gr
import time
from datetime import datetime
from sentence_transformers import SentenceTransformer
from qdrant_client import QdrantClient
from qdrant_client.models import Filter, FieldCondition, MatchValue
import os
from rapidfuzz import fuzz
from pyairtable import Table
from pyairtable import Api
import re
import unicodedata

# Setup Qdrant Client
qdrant_client = QdrantClient(
    url=os.environ.get("Qdrant_url"),
    api_key=os.environ.get("Qdrant_api"),
    timeout=30.0
)

# Airtable Config
AIRTABLE_API_KEY = os.environ.get("airtable_api")
BASE_ID = os.environ.get("airtable_baseid")
TABLE_NAME = "Feedback_search"
api = Api(AIRTABLE_API_KEY)
table = api.table(BASE_ID, TABLE_NAME)

# Preload Models
model = SentenceTransformer("BAAI/bge-m3")
collection_name = "product_bge-m3"
threshold = 0.45

# Utils
def is_non_thai(text):
    return re.match(r'^[A-Za-z0-9&\-\s]+$', text) is not None

def normalize(text: str) -> str:
    if is_non_thai(text):
        return text.strip()
    text = unicodedata.normalize("NFC", text)
    return text.replace("เแ", "แ").replace("เเ", "แ").strip().lower()

# Global state
latest_query_result = {"query": "", "result": "", "raw_query": "", "time": ""}

# Search Function
def search_product(query):
    yield gr.update(value="🔄 กำลังค้นหา..."), ""

    start_time = time.time()
    latest_query_result["raw_query"] = query

    corrected_query = normalize(query)
    query_embed = model.encode(corrected_query)

    try:
        result = qdrant_client.query_points(
            collection_name=collection_name,
            query=query_embed.tolist(),
            with_payload=True,
            query_filter=Filter(must=[FieldCondition(key="type", match=MatchValue(value="product"))]),
            limit=50
        ).points
    except Exception as e:
        yield gr.update(value="❌ Qdrant error"), f"<p>❌ Qdrant error: {str(e)}</p>"
        return

    if len(result) > 0:
      topk = 50  # ดึงมา rerank แค่ 50 อันดับแรกจาก Qdrant
      result = result[:topk]

      scored = []
      for r in result:
          name = r.payload.get("name", "")

          # ถ้า query สั้นเกินไป ให้ fuzzy_score = 0 เพื่อกันเพี้ยน
          if len(corrected_query) >= 3 and name:
            fuzzy_score = fuzz.partial_ratio(corrected_query, name) / 100.0
          else:
            fuzzy_score = 0.0
          # รวม hybrid score
          if fuzzy_score < 0.5:
            hybrid_score = r.score
          else:
            hybrid_score = 0.7 * r.score + 0.3 * fuzzy_score
          r.payload["score"] = hybrid_score  # เก็บลง payload ใช้เทียบ treshold ตอนเเสดงผล
          r.payload["fuzzy_score"] = fuzzy_score # เก็บไว้เผื่อ debug
          r.payload['semantic_score'] = r.score # เก็บไว้เผื่อ debug
          scored.append((r, hybrid_score))

      # เรียงตาม hybrid score แล้วกรองผลลัพธ์ที่ hybrid score ต่ำเกิน
      scored = sorted(scored, key=lambda x: x[1], reverse=True)
      result = [r[0] for r in scored]

    elapsed = time.time() - start_time
    html_output = f"<p>⏱ <strong>{elapsed:.2f} วินาที</strong></p>"
    if corrected_query != query:
        html_output += f"<p>🔧 แก้คำค้นจาก: <code>{query}</code> → <code>{corrected_query}</code></p>"
    html_output += '<div style="display: grid; grid-template-columns: repeat(auto-fill, minmax(220px, 1fr)); gap: 20px;">'
    result_summary, found = "", False

    for res in result:
        if res.payload["score"] >= threshold:
            found = True
            name = res.payload.get("name", "ไม่ทราบชื่อสินค้า")
            score = f"{res.payload['score']:.4f}"
            img_url = res.payload.get("imageUrl", "")
            price = res.payload.get("price", "ไม่ระบุ")
            brand = res.payload.get("brand", "")

            html_output += f"""
            <div style="border: 1px solid #ddd; border-radius: 8px; padding: 10px; text-align: center; box-shadow: 1px 1px 5px rgba(0,0,0,0.1); background: #fff;">
                <img src="{img_url}" style="width: 100%; max-height: 150px; object-fit: contain; border-radius: 4px;">
                <div style="margin-top: 10px;">
                    <div style="font-weight: bold; font-size: 14px;">{name}</div>
                    <div style="color: gray; font-size: 12px;">{brand}</div>
                    <div style="color: green; margin: 4px 0;">฿{price}</div>
                    <div style="font-size: 12px; color: #555;">score: {score}</div>
                </div>
            </div>
            """
            result_summary += f"{name} (score: {score}) | "

    html_output += "</div>"

    if not found:
        html_output += '<div style="text-align: center; font-size: 18px; color: #a00; padding: 30px;">❌ ไม่พบสินค้าที่เกี่ยวข้องกับคำค้นนี้</div>'

    latest_query_result.update({
        "query": corrected_query,
        "result": result_summary.strip(),
        "time": elapsed,
    })

    yield gr.update(value="✅ ค้นหาเสร็จแล้ว!"), html_output

# Feedback Function
def log_feedback(feedback):
    try:
        now = datetime.now().strftime("%Y-%m-%d")
        table.create({
            "model": "BGE M3",
            "timestamp": now,
            "raw_query": latest_query_result["raw_query"],
            "query": latest_query_result["query"],
            "result": latest_query_result["result"],
            "time(second)": latest_query_result["time"],
            "feedback": feedback
        })
        return "✅ Feedback saved to Airtable!"
    except Exception as e:
        return f"❌ Failed to save feedback: {str(e)}"

# Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("## 🔎 Product Semantic Search (BGE M3 + Qdrant)")

    query_input = gr.Textbox(label="พิมพ์คำค้นหา")
    result_output = gr.HTML(label="📋 ผลลัพธ์")
    status_output = gr.Textbox(label="🕒 สถานะ", interactive=False)

    with gr.Row():
        match_btn = gr.Button("✅ ตรง")
        not_match_btn = gr.Button("❌ ไม่ตรง")

    feedback_status = gr.Textbox(label="📬 สถานะ Feedback")

    query_input.submit(
        search_product,
        inputs=[query_input],
        outputs=[status_output, result_output]
    )
    match_btn.click(fn=lambda: log_feedback("match"), outputs=feedback_status)
    not_match_btn.click(fn=lambda: log_feedback("not_match"), outputs=feedback_status)

demo.launch(share=True)