Update app.py

app.py

@@ -1,86 +1,87 @@
 import json
-import re
-import random
 import html
-import requests
-from bs4 import BeautifulSoup
+import re
 import gradio as gr
-from llama_cpp import Llama
-from llama_cpp_agent import LlamaCppAgent
-from llama_cpp_agent.providers import LlamaCppPythonProvider
-from llama_cpp_agent.chat_history import BasicChatHistory
-from llama_cpp_agent.chat_history.messages import Roles
+from bs4 import BeautifulSoup
+import requests
 from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
 
-# Download your GGUF model from the HF repo
-:contentReference[oaicite:2]{index=2}
-:contentReference[oaicite:3]{index=3}
+# Download the model from your HF repo (replace with your actual repo/model filename)
+hf_hub_download(
+    repo_id="Bton/llama3-pr-Q4_K_M.gguf",
+    filename="unsloth.Q4_K_M.gguf",
+    local_dir="."
+)
+
+# Load model
+llm = Llama(
+    model_path="./your-model.Q4_K_M.gguf",
+    n_ctx=2048,
+    n_threads=4
+)
+
+# Scrape product details
+def scrape_amazon_product(url):
+    headers = {
+        "User-Agent": "Mozilla/5.0",
+        "Accept-Language": "en-US,en;q=0.9",
+    }
+    response = requests.get(url, headers=headers)
+    soup = BeautifulSoup(response.content, "html.parser")
+
+    title = soup.find(id="productTitle")
+    price = soup.select_one(".a-offscreen")
+    rating = next((e.get_text() for e in soup.select("span.a-icon-alt") if "out of 5 stars" in e.text), None)
+    review_elem = soup.select_one("div.reviewText.review-text-content")
+
+    about_section = soup.find("div", id="feature-bullets")
+    bullets = about_section.select("ul.a-unordered-list li span") if about_section else []
+    about = " ".join(b.get_text(strip=True) for b in bullets if b.get_text(strip=True))
 
-# Scraping logic (adjusted from your script):
-:contentReference[oaicite:4]{index=4}
-    :contentReference[oaicite:5]{index=5}
     return {
-        "product_title": title_text,
-        "price": price_text,
-        "rating": rating_text,
-        "first_review": review_text,
-        :contentReference[oaicite:6]{index=6}
+        "product_title": title.get_text(strip=True) if title else None,
+        "price": price.get_text(strip=True) if price else None,
+        "rating": rating,
+        "first_review": review_elem.get_text(strip=True) if review_elem else None,
+        "about_this_item": html.escape(about)
     }
 
-# Inference function
-:contentReference[oaicite:7]{index=7}
-    :contentReference[oaicite:8]{index=8}
+# Generate review
+def generate_review(url):
+    product = scrape_amazon_product(url)
 
-    # Format prompt
-    prompt = json.dumps({
-        "product_title": prod["product_title"],
-        "price": prod["price"],
-        "rating": prod["rating"],
-        "first_review": prod["first_review"],
-        "about_this_item": prod["about_this_item"],
-    }, ensure_ascii=False)
-    full_prompt = (
+    prompt = (
         "Write a product review based on the following product details.\n"
-        f"Your input is:\n{prompt}"
+        f"Your input is:\n{json.dumps(product)}"
     )
 
-    # Llama.cpp setup
-    llm = Llama(model_path=MODEL_PATH, n_ctx=2048)
-    provider = LlamaCppPythonProvider(llm)
-    agent = LlamaCppAgent(provider, system_prompt="", predefined_messages_formatter_type="CHATML")
-    settings = provider.get_provider_default_settings()
-    settings.stream = False
-
-    # Run inference
-    stream = agent.get_chat_response(
-        full_prompt,
-        llm_sampling_settings=settings,
-        chat_history=BasicChatHistory(),
-        returns_streaming_generator=False
-    )
-    full_text = "".join(token for token in stream)
+    response = llm(prompt, stop=["<|im_end|>", "<|end_of_text|>"])
+    text = response["choices"][0]["text"]
 
-    # Extract JSON from response
-    match = re.search(r'\{.*"title".*"review".*\}', full_text, re.DOTALL)
-    if match:
-        review = json.loads(match.group())
-    else:
-        review = {"title": "", "review": full_text.strip()}
+    try:
+        match = re.search(r'\{.*"title".*"review".*\}', text, re.DOTALL)
+        if match:
+            review_json = json.loads(match.group())
+        else:
+            review_json = {"title": "Review Generated", "review": text.strip()}
+    except:
+        review_json = {"title": "Review Generated", "review": text.strip()}
 
-    return prod["product_title"], review["title"], review["review"]
+    return product["product_title"], review_json["title"], review_json["review"]
 
 # Gradio UI
-:contentReference[oaicite:9]{index=9}
-    fn=review_generation_ui,
-    :contentReference[oaicite:10]{index=10}
+demo = gr.Interface(
+    fn=generate_review,
+    inputs=gr.Textbox(label="Amazon URL"),
     outputs=[
-        :contentReference[oaicite:11]{index=11}
-        :contentReference[oaicite:12]{index=12}
-        :contentReference[oaicite:13]{index=13}
+        gr.Textbox(label="Product Title"),
+        gr.Textbox(label="Review Title"),
+        gr.Textbox(label="Review Body", lines=5)
     ],
-    :contentReference[oaicite:14]{index=14}
-    :contentReference[oaicite:15]{index=15}
+    title="Amazon Review Bot (GGUF)",
+    description="Enter an Amazon product URL to generate a review using your GGUF model."
 )
 
-:contentReference[oaicite:16]{index=16}
+if __name__ == "__main__":
     demo.launch()