Spaces:
Sleeping
Sleeping
| """ | |
| Summary generation utilities for research papers. | |
| """ | |
| from langchain_groq import ChatGroq | |
| from langchain_core.prompts import ChatPromptTemplate | |
| from langchain_core.output_parsers import StrOutputParser | |
| from src.config import GROQ_API_KEY, LLM_MODEL, LLM_TEMPERATURE | |
| def ensure_groq_key(): | |
| if not GROQ_API_KEY: | |
| raise RuntimeError("GROQ_API_KEY is not configured. Set it in .env or Hugging Face secrets.") | |
| def generate_paper_summary(text: str, max_length: int = 500) -> str: | |
| """Generate a concise summary of the paper.""" | |
| ensure_groq_key() | |
| llm = ChatGroq( | |
| model=LLM_MODEL, | |
| api_key=GROQ_API_KEY, | |
| temperature=LLM_TEMPERATURE | |
| ) | |
| prompt = ChatPromptTemplate.from_template("""You are a research paper analyst. | |
| Generate a clear, concise summary of the following research paper in {max_length} words or less. | |
| Focus on: problem statement, methodology, key findings, and implications. | |
| Paper content: | |
| {text} | |
| Summary:""") | |
| chain = prompt | llm | StrOutputParser() | |
| summary = chain.invoke({"text": text[:3000], "max_length": max_length}) | |
| return summary | |
| def generate_section_summaries(chunks: list[str], num_sections: int = 5) -> list[str]: | |
| """Generate summaries for different sections of the paper.""" | |
| ensure_groq_key() | |
| if len(chunks) < num_sections: | |
| num_sections = len(chunks) | |
| section_size = len(chunks) // num_sections | |
| summaries = [] | |
| llm = ChatGroq( | |
| model=LLM_MODEL, | |
| api_key=GROQ_API_KEY, | |
| temperature=LLM_TEMPERATURE | |
| ) | |
| prompt = ChatPromptTemplate.from_template("""Summarize this section of a research paper in 1-2 sentences: | |
| {section_text}""") | |
| chain = prompt | llm | StrOutputParser() | |
| for i in range(num_sections): | |
| start_idx = i * section_size | |
| end_idx = (i + 1) * section_size if i < num_sections - 1 else len(chunks) | |
| section_text = "\n".join(chunks[start_idx:end_idx]) | |
| summary = chain.invoke({"section_text": section_text[:2000]}) | |
| summaries.append(f"Section {i+1}: {summary}") | |
| return summaries |