import streamlit as st import os from gpt_researcher import GPTResearcher import asyncio import nest_asyncio from contextlib import redirect_stdout import io from fpdf import FPDF from datetime import datetime import uuid import importlib.metadata # ---------- version helpers (as in your original) ---------- def get_version(package_name, module=None): try: if module and hasattr(module, '__version__'): version = module.__version__ else: version = importlib.metadata.version(package_name) print(f"{package_name} version: {version}") except AttributeError: print(f"{package_name} does not have a __version__ attribute.") except importlib.metadata.PackageNotFoundError: print(f"{package_name} is not installed.") import asyncio as _asyncio get_version('streamlit', st) get_version('gpt_researcher') get_version('nest_asyncio', nest_asyncio) get_version('fpdf') print("\nStandard Library Modules:") for lib in ['os', 'asyncio', 'contextlib', 'io', 'datetime', 'uuid']: print(f"{lib} is part of the Python Standard Library and does not have a separate version number.") # ---------- Streamlit / asyncio setup ---------- st.set_page_config(layout="wide") nest_asyncio.apply() # ---------- PDF helper (same logic, safe path) ---------- class PDF(FPDF): def header(self): self.set_font("Arial", "B", 12) self.cell(0, 10, "Research Report", 0, 1, "C") def footer(self): self.set_y(-15) self.set_font("Arial", "I", 8) self.cell(0, 10, f"Page {self.page_no()}", 0, 0, "C") def create_pdf(report_text: str) -> str: """Create the PDF and return a unique file path under /tmp.""" os.makedirs("/tmp", exist_ok=True) pdf_path = f"/tmp/report_{uuid.uuid4().hex}.pdf" pdf = PDF() pdf.add_page() pdf.set_auto_page_break(auto=True, margin=15) pdf.set_font("Arial", size=12) for line in report_text.split('\n'): # FPDF is latin-1; degrade gracefully pdf.multi_cell(0, 10, line.encode('latin-1', 'replace').decode('latin-1')) pdf.output(pdf_path, 'F') return pdf_path # ---------- async run (unchanged behavior: capture stdout, return at end) ---------- async def get_report(query: str, report_type: str, sources: list, report_source: str): f = io.StringIO() unique_key = str(uuid.uuid4()) # kept from your original (not strictly used) with redirect_stdout(f): if report_source == 'local': # Use a guaranteed-writable dir in containers os.environ['DOC_PATH'] = '/tmp/uploads' os.makedirs(os.environ['DOC_PATH'], exist_ok=True) researcher = GPTResearcher(query=query, report_type=report_type, report_source=report_source) else: researcher = GPTResearcher(query=query, report_type=report_type, source_urls=sources) await researcher.conduct_research() # Your original polling to avoid infinite loop max_attempts = 30 attempts = 0 while attempts < max_attempts: logs = f.getvalue() if "Finalized research step" in logs: break await asyncio.sleep(1) attempts += 1 report = await researcher.write_report() return report, f.getvalue() # ---------- UI (same layout/wording as your original) ---------- st.title("GPT Researcher") st.markdown(""" GPT Researcher is an autonomous agent designed for comprehensive online research tasks. It pulls information from the web or uploaded documents to create detailed, factual, research reports. """) with st.expander("Why Use GPT Researcher?", expanded=False): st.markdown(""" - **Objective and Unbiased**: GPT Researcher focuses on delivering accurate and factual information without bias. - **Time-Efficient**: It significantly reduces the time required for manual research tasks. - **Up-to-Date Information**: Unlike traditional LLMs, GPT Researcher avoids outdated information and minimizes the risk of hallucinations. - **Comprehensive Reports**: Capable of producing long, detailed research reports (2,000+ words). - **Reduced Misinformation**: By considering a wide range of sources, it minimizes the risks associated with limited or biased information. For more details, visit the [GPT Researcher GitHub repository](https://github.com/assafelovic/gpt-researcher). """) # Keep your style tweaks st.markdown( """ """, unsafe_allow_html=True, ) st.markdown('

Enter your research query:

', unsafe_allow_html=True) default_query = "Why is the Stock Price of Nvidia Soaring?" # Same input — just hide label for accessibility warning user_query = st.text_input("", default_query, help="Type your research question or topic.", label_visibility="collapsed") if user_query: current_date = datetime.now().strftime("%B %Y") final_query = f"{user_query} Current Date is {current_date}" else: final_query = "" st.sidebar.title("Research Settings") with st.sidebar.expander("How to Use", expanded=False): st.markdown(""" ### How to Use 1. **Select Research Type**: Choose between Web Research and Document Research. 2. **Enter Research Query**: Type in your research question or topic. 3. **Choose Report Type**: Select the format of the report you want (research report, resource list, or article outline). 4. **Provide Sources or Upload Files**: For Web Research, you can enter URLs. For Document Research, upload the necessary files. 5. **Run Research**: Click the "Run Research" button to start. The logs will update after the run, and the final report will be displayed and available for download as a PDF. """) with st.sidebar: research_type = st.selectbox("Select research type:", ["Web Research", "Document Research"], help="Choose between web-based research or research from local documents.") report_type = st.selectbox("Select report type:", ["research_report", "resource_list", "article_outline"], help="Choose the format of the final report.") if research_type == "Web Research": sources_input = st.text_area("Enter your sources (optional, comma-separated URLs):", help="Provide a list of URLs to use as sources, separated by commas.") sources = [url.strip() for url in sources_input.split(',') if url.strip()] else: uploaded_files = st.file_uploader("Upload files for local research:", accept_multiple_files=True, help="Upload documents for the research.") sources = [] if uploaded_files: # Save uploads to the writable /tmp/uploads os.makedirs("/tmp/uploads", exist_ok=True) for uploaded_file in uploaded_files: file_path = os.path.join("/tmp/uploads", uploaded_file.name) with open(file_path, "wb") as f: f.write(uploaded_file.getbuffer()) if st.button("Run Research"): if not user_query: st.warning("Please enter a research query.") else: # Same as your original os.environ['RETRIEVER'] = 'tavily' report_source = 'local' if research_type == "Document Research" else 'web' with st.spinner("Running research..."): report, logs = asyncio.run(get_report(final_query, report_type, sources, report_source)) st.session_state.report = report st.session_state.logs = logs # ---------- Results (report + PDF + final logs, same as your original) ---------- if 'report' in st.session_state: st.markdown("### Research Report") st.markdown(st.session_state.report) # Create PDF safely under /tmp and offer download try: pdf_path = create_pdf(st.session_state.report) with open(pdf_path, "rb") as pdf_file: st.download_button( label="Download report as PDF", data=pdf_file, file_name="report.pdf", mime="application/pdf" ) except Exception as e: st.warning(f"Could not generate PDF: {e}") st.markdown("### Agent Logs") if 'logs' in st.session_state: st.text_area( "Logs will appear here during the research process:", value=st.session_state.logs, height=200, key=f"logs_{uuid.uuid4()}" # same trick you had originally ) else: st.text_area( "Logs will appear here during the research process", height=200, key=f"logs_{uuid.uuid4()}" ) # Hide Streamlit's default footer and menu hide_streamlit_style = """ """ st.markdown(hide_streamlit_style, unsafe_allow_html=True)