app.py

import streamlit as st
import requests
from transformers import pipeline
import re
from collections import Counter
import json

# Load summarizer model
@st.cache_resource
def load_summarizer():
    return pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")

summarizer = load_summarizer()

# Summarize user's invention
def generate_summary(text):
    if len(text.split()) < 10:
        return "Please enter at least 10 words for summarization."
    result = summarizer(text, max_length=200, min_length=60, do_sample=False)
    return result[0]['summary_text']

# Improved keyword extraction
def extract_keywords(text, max_keywords=5):
    words = re.findall(r'\b[a-z]{4,}\b', text.lower())
    stopwords = set([
        "about", "above", "after", "again", "against", "being", "have", "still", "does", "with", "this",
        "since", "their", "which", "these", "those", "other", "where", "while", "into", "from", "that",
        "will", "would", "could", "should", "might", "shall", "more", "such", "only", "like", "than",
        "there", "here", "also", "very", "every", "because", "through", "among", "between", "during"
    ])
    filtered = [w for w in words if w not in stopwords]
    common = Counter(filtered).most_common(max_keywords)
    return [w for w, _ in common]

def search_patents(keywords):
    # Fixed query structure for PatentsView API
    query = {
        "_text_any": {
            "patent_title": " ".join(keywords),
            "patent_abstract": " ".join(keywords)
        }
    }
    
    fields = ["patent_number", "patent_title", "patent_abstract"]
    
    url = "https://api.patentsview.org/patents/query"
    
    params = {
        "q": json.dumps(query),
        "f": json.dumps(fields),
        "o": json.dumps({"per_page": 5})
    }
    
    try:
        response = requests.get(url, params=params, timeout=30)
        response.raise_for_status()
        data = response.json()
        
        # Debug info
        st.write(f"API Response Status: {response.status_code}")
        if "patents" not in data:
            st.write("API Response Keys:", list(data.keys()))
            
        return data.get("patents", [])
    except requests.exceptions.Timeout:
        st.error("Patent API request timed out. Please try again.")
        return []
    except requests.exceptions.RequestException as e:
        st.error(f"Patent API request error: {str(e)}")
        return []
    except json.JSONDecodeError as e:
        st.error(f"Failed to parse API response: {str(e)}")
        return []
    except Exception as e:
        st.error(f"Unexpected error: {str(e)}")
        return []

# Alternative search function using simpler query structure
def search_patents_simple(keywords):
    # Try a simpler query structure
    keyword_string = " ".join(keywords)
    
    query = {
        "_text_any": keyword_string
    }
    
    fields = ["patent_number", "patent_title", "patent_abstract"]
    
    url = "https://api.patentsview.org/patents/query"
    
    params = {
        "q": json.dumps(query),
        "f": json.dumps(fields),
        "o": json.dumps({"per_page": 5})
    }
    
    try:
        response = requests.get(url, params=params, timeout=30)
        response.raise_for_status()
        data = response.json()
        return data.get("patents", [])
    except Exception as e:
        st.error(f"Simple search also failed: {str(e)}")
        return []

# Similarity score via keyword overlap
def similarity_score(text1, text2):
    if not text1 or not text2:
        return 0
    set1 = set(text1.lower().split())
    set2 = set(text2.lower().split())
    overlap = set1.intersection(set2)
    return len(overlap) / max(len(set1), len(set2), 1)

# UI
st.title("🧠 Patent Infringement Analyzer (Live Search)")
st.write("This tool summarizes your invention and compares it to real patents using the PatentsView API.")

invention = st.text_area("✍️ Describe your invention (10+ words):", height=250)

if st.button("🔍 Search & Analyze"):
    if len(invention.split()) < 10:
        st.error("Please enter more detail — at least 10 words.")
    else:
        with st.spinner("Summarizing your invention..."):
            summary = generate_summary(invention)
            st.subheader("📄 Invention Summary")
            st.write(summary)
            
            keywords = extract_keywords(invention)
            st.info("🔑 Keywords used for search: " + ", ".join(keywords))
        
        with st.spinner("Searching patents..."):
            # Try the main search first
            patents = search_patents(keywords)
            
            # If that fails, try the simpler approach
            if not patents:
                st.write("Trying alternative search method...")
                patents = search_patents_simple(keywords)
        
        if not patents:
            st.warning("No similar patents found. This could mean:")
            st.write("- Your invention might be novel")
            st.write("- The keywords might be too specific")
            st.write("- The patent database might be temporarily unavailable")
        else:
            st.subheader("🔍 Similar Patents & Overlap Risk")
            for i, p in enumerate(patents):
                # Handle missing fields gracefully
                patent_num = p.get("patent_number", f"Patent {i+1}")
                patent_title = p.get("patent_title", "No title available")
                patent_abstract = p.get("patent_abstract", "No abstract available")
                
                score = similarity_score(summary, patent_abstract)
                risk = "High" if score > 0.25 else "Moderate" if score > 0.15 else "Low"
                
                st.markdown(f"""
**[{patent_num}]** - **{patent_title}**

*Abstract*: {patent_abstract[:500]}{'...' if len(patent_abstract) > 500 else ''}

**Similarity Score**: {score:.3f} | **Estimated Risk**: `{risk}`

---
                """)