app.py

import streamlit as st
from crewai import Agent, Task, Crew
import os
from langchain_groq import ChatGroq
from langchain_openai import ChatOpenAI
from fpdf import FPDF
import pandas as pd
import plotly.express as px
import tempfile
import time
import ast
import logging

# Setup logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')

# Title and Application Introduction
st.title("Patent Strategy and Innovation Consultant")
st.sidebar.write(
    "This application uses AI to provide actionable insights and comprehensive analysis for patent-related strategies."
)

# User Input Section
st.sidebar.header("User Inputs")
patent_area = st.text_input("Enter Patent Technology Area", value="Transparent Antennas for Windshields")
stakeholder = st.text_input("Enter Stakeholder", value="Patent Attorneys")

# Initialize LLM
llm = None

# Model Selection
model_choice = st.radio("Select LLM", ["GPT-4o", "llama-3.3-70b"], index=1, horizontal=True)


# API Key Validation and LLM Initialization
groq_api_key = os.getenv("GROQ_API_KEY")
openai_api_key = os.getenv("OPENAI_API_KEY")

#llm = ChatGroq(groq_api_key=os.getenv("GROQ_API_KEY"), model="groq/llama-3.3-70b-versatile")

if model_choice == "llama-3.3-70b":
    if not groq_api_key:
        st.error("Groq API key is missing. Please set the GROQ_API_KEY environment variable.")
        llm = None
    else:
        llm = ChatGroq(groq_api_key=groq_api_key, model="groq/llama-3.3-70b-versatile")
elif model_choice == "GPT-4o":
    if not openai_api_key:
        st.error("OpenAI API key is missing. Please set the OPENAI_API_KEY environment variable.")
        llm = None
    else:
        llm = ChatOpenAI(api_key=openai_api_key, model="gpt-4o")


# Advanced Options
st.sidebar.header("Advanced Options")
enable_advanced_analysis = st.sidebar.checkbox("Enable Advanced Analysis", value=True)
enable_custom_visualization = st.sidebar.checkbox("Enable Custom Visualizations", value=True)

# Agent Customization
st.sidebar.header("Agent Customization")
with st.sidebar.expander("Customize Agent Goals", expanded=False):
    enable_customization = st.checkbox("Enable Custom Goals")
    if enable_customization:
        planner_goal = st.text_area(
            "Planner Goal",
            value=(
                "Conduct comprehensive research on trends in patent filings, technological innovation, and market dynamics within the {topic} sector. "
                "Identify key players, emerging technologies, competitor strategies, and market gaps. "
                "Provide actionable suggestions specifically tailored to the needs of {stakeholder}"
            )
        )
        writer_goal = st.text_area(
            "Writer Goal",
            value=(
                "Craft a professional, logically structured insights report that combines research data and strategic analysis "
                "into a cohesive narrative. The report should summarize trends, highlight strategies, and provide actionable outcomes "
                "specifically tailored to the needs of {stakeholder}."
            )
        )
        analyst_goal = st.text_area(
            "Analyst Goal",
            value=(
                "Perform detailed statistical analysis of patent filings, growth trends, and innovation distribution within the {topic} sector, specifically tailored to the needs of {stakeholder}. "
                "Identify top regions, key assignees/companies, and emerging technologies driving innovation. "
                "Conduct market gap analysis and evaluate competitor positioning to uncover untapped opportunities. "
                "Deliver actionable, data-driven insights to guide strategic decision-making."
                "Provide structured output in a list of dictionaries with 'Category' and 'Values' keys for clear data presentation."
            )
        )
    else:
        planner_goal = "Research trends in patent filings and technological innovation, identify key players, and provide strategic recommendations."
        writer_goal = (
            "Craft a professional, logically structured insights report that combines research data and strategic analysis "
            "into a cohesive narrative. The report should summarize trends, highlight strategies, and provide actionable outcomes "
            "specifically tailored to the needs of {stakeholder}."
        )
        analyst_goal = (
            "Perform detailed statistical analysis of patent filings, growth trends, and innovation distribution within the {topic} sector, specifically tailored to the needs of {stakeholder}. "
            "Identify top regions, key assignees/companies, and emerging technologies driving innovation. "
            "Conduct market gap analysis and evaluate competitor positioning to uncover untapped opportunities. "
            "Deliver actionable, data-driven insights to guide strategic decision-making."
            "Provide structured output in a list of dictionaries with 'Category' and 'Values' keys for clear data presentation."
        )

# Agent Definitions
planner = Agent(
    role="Patent Research Consultant",
    goal=planner_goal,
    backstory=(
        "You're tasked with researching {topic} patents and identifying key trends and players. Your work supports the Patent Writer and Data Analyst."
    ),
    allow_delegation=False,
    verbose=True,
    llm=llm
)

writer = Agent(
    role="Patent Insights Writer",
    goal=writer_goal,
    backstory=(
        "Using the research from the Planner and data from the Analyst, craft a professional document summarizing patent insights for {stakeholder}."
    ),
    allow_delegation=False,
    verbose=True,
    llm=llm
)

analyst = Agent(
    role="Patent Data Analyst",
    goal=analyst_goal,
    backstory=(
        "Analyze patent filing data and innovation trends in {topic} to provide statistical insights. Your analysis will guide the Writer's final report."
    ),
    allow_delegation=False,
    verbose=True,
    llm=llm
)

# Task Definitions
plan = Task(
    description=(
        "1. Research recent trends in {topic} patent filings and innovation.\n"
        "2. Identify key players and emerging technologies.\n"
        "3. Provide recommendations for stakeholders on strategic directions.\n"
        "4. Identify key statistics such as top regions, top players, and hot areas of innovation.\n"
        "5. Limit the output to 600 words."
    ),
    expected_output="A research document with structured insights, strategic recommendations, and key statistics.",
    agent=planner
)

write = Task(
    description=(
        "1. Use the Planner's and Analyst's outputs to craft a professional patent insights document.\n"
        "2. Include key findings, visual aids, and actionable strategies.\n"
        "3. Suggest strategic directions and highlight untapped innovation areas.\n"
        "4. Incorporate summarized tables for key statistics and example inventions.\n"
        "5. Limit the document to 600 words."
    ),
    expected_output="A polished, stakeholder-ready patent insights document with actionable recommendations.",
    agent=writer
)

analyse = Task(
    description=(
        "1. Conduct a comprehensive statistical analysis of patent filing trends, innovation hot spots, and future growth projections in the transparent antenna industry.\n"
        "2. Identify and rank the top regions, leading assignees/companies driving innovation.\n"
        "3. Highlight regional innovation trends and the distribution of emerging technologies across different geographies.\n"
        "4. Provide actionable insights and strategic recommendations based on the data.\n"
        "5. Deliver structured output in a list of dictionaries with 'Category' and 'Values' fields:\n"
        "   - 'Values' can be:\n"
        "     a) A dictionary with counts for quantitative data (e.g., {{'Region A': 120, 'Region B': 95}}),\n"
        "     b) A list of key items (technologies, companies, inventors), or\n"
        "     c) Descriptive text for qualitative insights.\n"
        "6. Example Output Format:\n"
        "[\n"
        "  {{'Category': 'Top Regions', 'Values': {{'North America': 120, 'Europe': 95, 'Asia-Pacific': 85}}}},\n"
        "  {{'Category': 'Top Assignees', 'Values': {{'Company A': 40, 'Company B': 35}}}},\n"
        "  {{'Category': 'Emerging Technologies', 'Values': ['Graphene Antennas', '5G Integration']}},\n"
        "  {{'Category': 'Strategic Insights', 'Values': 'Collaborations between automotive and material science industries are accelerating innovation.'}}\n"
        "]\n"
        "7. Ensure that the output is clean, well-structured, and formatted for use in visualizations and tables."
    ),
    expected_output="A structured, well-organized dataset with numeric, list-based, and descriptive insights for comprehensive visual and tabular reporting.",
    agent=analyst
)


crew = Crew(
    agents=[planner, analyst, writer],
    tasks=[plan, analyse, write],
    verbose=True
)

# PDF Report Generation
def generate_pdf_report(result, charts=None, table_data=None, metadata=None):
    with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_pdf:
        pdf = FPDF()
        pdf.add_page()
        pdf.set_font("Arial", size=12)
        pdf.set_auto_page_break(auto=True, margin=15)

        pdf.set_font("Arial", size=16, style="B")
        pdf.cell(200, 10, txt="Patent Strategy and Innovation Report", ln=True, align="C")
        pdf.ln(10)

        if metadata:
            pdf.set_font("Arial", size=10)
            for key, value in metadata.items():
                pdf.cell(200, 10, txt=f"{key}: {value}", ln=True)

        pdf.set_font("Arial", size=12)
        pdf.multi_cell(0, 10, txt=result)

        if charts:
            for chart_path in charts:
                try:
                    pdf.add_page()
                    pdf.image(chart_path, x=10, y=20, w=180)
                    logging.info(f"Successfully included chart: {chart_path}")
                except Exception as e:
                    logging.error(f"Failed to include chart in PDF: {chart_path}. Error: {e}")

        if table_data:
            pdf.add_page()
            pdf.set_font("Arial", size=10)
            pdf.cell(200, 10, txt="Consolidated Table:", ln=True, align="L")
            for row in table_data:
                pdf.cell(200, 10, txt=str(row), ln=True)

        pdf.output(temp_pdf.name)
        return temp_pdf.name

# Data Validation
def validate_analyst_output(analyst_output):
    if not analyst_output:
        st.warning("No data available for analysis.")
        return None
    if not isinstance(analyst_output, list) or not all(isinstance(item, dict) for item in analyst_output):
        st.warning("Analyst output must be a list of dictionaries.")
        return None
    required_keys = {'Category', 'Values'}
    if not all(required_keys.issubset(item.keys()) for item in analyst_output):
        st.warning(f"Each dictionary must contain keys: {required_keys}")
        return None
    return analyst_output

# Visualization and Table Display
def create_visualizations(analyst_output):
    chart_paths = []
    validated_data = validate_analyst_output(analyst_output)

    if validated_data:
        for item in validated_data:
            category = item["Category"]
            values = item["Values"]

            try:
                # Handle dictionary data
                if isinstance(values, dict):
                    df = pd.DataFrame(list(values.items()), columns=["Label", "Count"])

                    # Choose Pie Chart for fewer categories, else Bar Chart
                    if len(df) <= 5:
                        chart = px.pie(df, names="Label", values="Count", title=f"{category} Distribution")
                    else:
                        chart = px.bar(df, x="Label", y="Count", title=f"{category} Analysis")

                # Handle list data 
                elif isinstance(values, list):
                    # Convert the list into a frequency count without dummy values
                    df = pd.DataFrame(values, columns=["Label"])
                    df = df["Label"].value_counts().reset_index()
                    df.columns = ["Label", "Count"]

                    # Plot as a bar chart or pie chart
                    if len(df) <= 5:
                        chart = px.pie(df, names="Label", values="Count", title=f"{category} Distribution")
                    else:
                        chart = px.bar(df, x="Label", y="Count", title=f"{category} Frequency")

                # Handle text data
                elif isinstance(values, str):
                    st.subheader(f"{category} Insights")
                    st.table(pd.DataFrame({"Insights": [values]}))
                    continue  # No chart for text data

                else:
                    st.warning(f"Unsupported data format for category: {category}")
                    continue

                # Display the chart in Streamlit
                st.plotly_chart(chart)

                # Save the chart for PDF export
                with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as temp_chart:
                    chart.write_image(temp_chart.name)
                    chart_paths.append(temp_chart.name)

            except Exception as e:
                st.error(f"Failed to generate visualization for {category}: {e}")
                logging.error(f"Error in {category} visualization: {e}")

    return chart_paths



def display_table(analyst_output):
    table_data = []
    validated_data = validate_analyst_output(analyst_output)

    if validated_data:
        for item in validated_data:
            category = item["Category"]
            values = item["Values"]

            # Error handling to prevent crashes
            try:
                # Handle dictionary data (Table View)
                if isinstance(values, dict):
                    df = pd.DataFrame(list(values.items()), columns=["Label", "Count"])
                    st.subheader(f"{category} (Table View)")
                    st.dataframe(df)
                    table_data.extend(df.to_dict(orient="records"))

                # Handle list data (List View)
                elif isinstance(values, list):
                    df = pd.DataFrame(values, columns=["Items"])
                    st.subheader(f"{category} (List View)")
                    st.dataframe(df)
                    table_data.extend(df.to_dict(orient="records"))

                # Handle text data (Summary View)
                elif isinstance(values, str):
                    st.subheader(f"{category} (Summary)")
                    st.table(pd.DataFrame({"Insights": [values]}))
                    table_data.append({"Category": category, "Values": values})

                else:
                    st.warning(f"Unsupported data format for category: {category}")

            except Exception as e:
                logging.error(f"Error processing {category}: {e}")
                st.error(f"Failed to display {category} as a table due to an error.")

    return table_data



def parse_analyst_output(raw_output):
    structured_data = []
    current_category = None
    current_values = []

    # Split raw output by line
    lines = raw_output.split('\n')

    for line in lines:
        line = line.strip()

        # Detect the start of a new category
        if line.startswith("Category:"):
            # Save the previous category and its values
            if current_category and current_values:
                structured_data.append({
                    "Category": current_category,
                    "Values": current_values if len(current_values) > 1 else current_values[0]
                })
            # Start processing the new category
            current_category = line.replace("Category:", "").strip()
            current_values = []

        # Skip 'Values:' header
        elif line.startswith("Values:"):
            continue

        # Process the values under the current category
        elif line and current_category:
            try:
                # Attempt to convert the line into Python data (dict/list)
                parsed_value = ast.literal_eval(line)
                current_values.append(parsed_value)
            except (ValueError, SyntaxError):
                # If parsing fails, treat it as plain text
                current_values.append(line)

    # Save the last processed category
    if current_category and current_values:
        structured_data.append({
            "Category": current_category,
            "Values": current_values if len(current_values) > 1 else current_values[0]
        })

    return structured_data


# Main Execution Block
if st.button("Generate Patent Insights"):
    with st.spinner('Processing...'):
        try:
            start_time = time.time()
            results = crew.kickoff(inputs={"topic": patent_area, "stakeholder": stakeholder})
            elapsed_time = time.time() - start_time

            writer_output = getattr(results.tasks_output[2], "raw", "No details available.")
            if writer_output:
                st.markdown("### Final Report")
                st.write(writer_output)
            else:
                st.warning("No final report available.")

            with st.expander("Explore Detailed Insights"):
                tab1, tab2 = st.tabs(["Planner's Insights", "Analyst's Analysis"])

                with tab1:
                    planner_output = getattr(results.tasks_output[0], "raw", "No details available.")
                    st.write(planner_output)

                with tab2:
                    analyst_output = getattr(results.tasks_output[1], "raw", "No details available.")
                    st.write(analyst_output)
                    # Convert raw text to structured data
                    if isinstance(analyst_output, str):
                        analyst_output = parse_analyst_output(analyst_output)
                    st.subheader("Structured Analyst Output")
                    st.write(analyst_output)
                    

                    charts = []
                    if enable_advanced_analysis:
                        charts = create_visualizations(analyst_output)

                    table_data = display_table(analyst_output)

            st.success(f"Analysis completed in {elapsed_time:.2f} seconds.")
            pdf_path = generate_pdf_report(writer_output, charts=charts, table_data=table_data, metadata={"Technology Area": patent_area, "Stakeholder": stakeholder})
            with open(pdf_path, "rb") as report_file:
                st.download_button("Download Report", data=report_file, file_name="Patent_Strategy_Report.pdf")

        except Exception as e:
            logging.error(f"An error occurred during execution: {e}")
            st.error(f"An error occurred during execution: {e}")