Feat: Added the Export BibTex Citation tool

analysis.py

@@ -21,26 +21,6 @@ Provide a comprehensive technical analysis by structuring your response in Markd
 (Propose 2-3 concrete, technically-grounded hypotheses for extending this research. For each, describe a potential experiment or technical extension.)
 """
 
-# # The master prompt that asks for everything at once.
-# COMPREHENSIVE_ANALYSIS_PROMPT = """
-# Provide a comprehensive technical analysis of the document for a knowledgeable audience (e.g., graduate students, researchers). Structure your response in Markdown with the following sections, in this exact order:
-
-# ## 1. Abstract Summary
-# (A concise summary of the paper's core contributions, methods, and key results, similar to a conference abstract.)
-
-# ## 2. Core Architecture and Methodology
-# (Deconstruct the system's architecture and the flow of data or logic. Use bullet points to detail key components and algorithms. Be technically precise.)
-
-# ## 3. Quantitative Results & Critical Analysis
-# (Present the main quantitative results in a list or responsive format (NO WIDE TABLES). Provide a brief but critical analysis of what these results mean.)
-
-# ## 4. Positioning in the Field
-# (Situate this work by comparing it to 1-2 key alternative approaches mentioned in the paper, highlighting its unique technical differentiators.)
-
-# ## 5. Proposed Future Research Directions
-# (Propose 2-3 concrete, technically-grounded hypotheses and experimental ideas for extending this research based on the paper's conclusion or limitations.)
-# """
-
 def run_analysis_on_single_paper(documents):
     """
     This simplified version creates an index and runs a single, comprehensive query against it.

app.py

@@ -2,11 +2,11 @@ import gradio as gr
 import os
 import re
 import hashlib
-from llama_index.core import Settings
+from llama_index.core import Settings, Document
 from llama_index.readers.file import PDFReader
 from llama_index.embeddings.mistralai import MistralAIEmbedding
-from utils import get_llm, download_pdf_from_url
-from agents import create_scout_agent
+from utils import get_llm, download_pdf_from_url, format_to_bibtex
+from agents import create_scout_agent, create_specialist_agent, CITATION_EXTRACTOR_PROMPT
 from analysis import run_analysis_on_single_paper
 
 # --- Orchestrator Functions for Gradio ---
@@ -33,6 +33,29 @@ def pdf_analysis_flow(pdf_file, progress=gr.Progress()):
         print(f"An error occurred in pdf_analysis_flow: {e}")
         return f"An error occurred: {e}"
 
+def export_bibtex_flow(documents, file_obj):
+    """Workflow for the 'Export Citation' button."""
+    if not documents:
+        raise gr.Error("Please analyze a paper first.")
+    
+    filename = os.path.basename(file_obj.name)
+    print(f"--- BibTeX Export: Starting citation extraction for {filename} ---")
+    
+    first_page_text = documents[0].text
+
+    # We only need the LLM for this, no other tools.
+    Settings.llm = get_llm()
+    extractor_agent = create_specialist_agent(CITATION_EXTRACTOR_PROMPT, Settings.llm, [])
+    
+    # Give the agent the text and ask it to perform its task
+    response = extractor_agent.chat(f"Extract bibliographic data from this text: {first_page_text[:4000]}")
+    
+    print(f"--- BibTeX Export: Agent responded with: {response.response} ---")
+    
+    # Format the extracted JSON into a BibTeX string
+    bibtex_string = format_to_bibtex(response.response, filename)
+    
+    return bibtex_string
 
 def scout_agent_flow(topic_query, progress=gr.Progress()):
     """This function now runs the scout agent and directly returns its summary."""

utils.py

@@ -1,6 +1,5 @@
-# utils.py (Corrected and Simplified)
-
 import os
+import json
 import requests
 from io import BytesIO
 from llama_index.llms.mistralai import MistralAI
@@ -24,4 +23,27 @@ def download_pdf_from_url(url: str):
         return BytesIO(response.content)
     except requests.exceptions.RequestException as e:
         print(f"Error downloading {url}: {e}")
-        return None
+        return None
+
+def format_to_bibtex(citation_json_str: str, arxiv_id: str) -> str:
+    """Formats a JSON string of citation data into a BibTeX entry."""
+    try:
+        data = json.loads(citation_json_str)
+        title = data.get("title", "No Title Found")
+        authors = " and ".join(data.get("authors", ["N/A"]))
+        year = data.get("year", "N/A")
+        
+        # Create a simple citation key, e.g., "bouzenia2024"
+        first_author_lastname = authors.split(' ')[-1].lower() if ' ' in authors else authors.lower()
+        key = f"{first_author_lastname}{year}"
+        
+        bibtex_entry = f"""@article{{{key},
+            title   = {{{title}}},
+            author  = {{{authors}}},
+            year    = {{{year}}},
+            journal = {{arXiv preprint arXiv:{arxiv_id}}}
+        }}"""
+        return bibtex_entry
+    except (json.JSONDecodeError, KeyError) as e:
+        print(f"Error formatting BibTeX: {e}")
+        return "Could not generate BibTeX citation. The required data could not be extracted."