A langgraph-based agent system for analyzing research trends in academic conferences. This project provides a modern langchain agent architecture with English documentation and langgraph compatibility for research paper analysis.

• Keyword Generation: Generate research keywords for specific topics using LLMs
• Paper Crawling: Fetch papers from academic conferences (NeurIPS, PoPETs, USENIX, ACL)
• Paper Filtering: Filter papers by topic relevance using keyword matching or LLM assessment
• Paper Summarization: Download PDFs, extract content, and generate summaries in both English and Chinese
• Summary Aggregation: Structure and aggregate summaries into Excel reports
• Langgraph Workflow: Orchestrated workflow with state management and error handling

research_trend_analyzer_light_agent/
├── agent/                    # Langgraph agent components
│   ├── state.py             # Workflow state definition
│   ├── nodes.py             # Workflow nodes implementation
│   └── graph.py             # Langgraph construction
├── configs/                  # Configuration files
│   ├── __init__.py
│   ├── llm_provider.py      # LLM configuration for langchain
│   ├── log_config.py        # Logging configuration
│   └── analysis_scope.json  # Research topics and keywords
├── tools/                    # Langchain tools
│   ├── keywords_generator.py
│   ├── paper_crawler.py
│   ├── paper_filter.py
│   ├── paper_summarizer.py
│   └── summary_aggregator.py
├── utils/                    # Utility functions
│   ├── __init__.py
│   ├── call_llms.py         # Langchain-compatible LLM calling
│   ├── helper_func.py       # File operations and utilities
│   ├── paper_process.py     # PDF processing utilities
│   └── prompts.py          # LLM prompts
├── tests/                   # Unit tests
│   └── test_workflow.py
├── main.py                  # Main entry point (CLI)
├── example_usage.py         # Programmatic usage examples
├── run_tests.py            # Test runner
├── requirements.txt        # Dependencies
└── .env.example           # Environment variables template

1. Clone and setup:

   cd research_trend_analyzer_light_agent
   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

2. Install dependencies:

   pip install -r requirements.txt

3. Configure environment:

   cp .env.example configs/.env
   # Edit .env with your API keys and configurations

# Basic usage
python main.py --conference neurips --year 2023 --topic dp_theory

# Advanced options
python main.py --conference popets --year 2023 --topic usable_privacy \
               --method llm --language EN --output results.json

# Skip existing steps
python main.py --conference neurips --year 2023 --topic AI_audit \
               --skip-keyword-generation --skip-crawling

# Display help
python main.py --help

from agent.graph import run_research_workflow

# Run workflow programmatically
config = {
    "conference": "neurips",
    "year": 2023,
    "topic": "privacy",
    "method": "llm",
    "language": "CH",
    "max_papers": 10  # Optional: limit for testing
}

results = run_research_workflow(config)
print(f"Workflow status: {results['status']}")

from example_usage import example_basic_usage, example_advanced_usage

# Run examples
example_basic_usage()
example_advanced_usage()

• NeurIPS (Neural Information Processing Systems)
• PoPETs (Proceedings on Privacy Enhancing Technologies)
• USENIX Security (USENIX Security Symposium)
• USENIX SOUPS (Symposium on Usable Privacy and Security)
• ACL Long (ACL Anthology Long Papers)
• ACL Findings (ACL Anthology Findings Papers)

Create a .env file with your API keys:

OPENAI_API_KEY=your_openai_api_key
GEMINI_API_KEY=your_gemini_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key

Edit configs/analysis_scope.json to define research topics:

{
  "privacy": {
    "definition": "Privacy and data protection research",
    "keywords": ["privacy", "data protection", "encryption", "differential privacy"]
  },
  "machine_learning": {
    "definition": "Machine learning and AI research",
    "keywords": ["machine learning", "ai", "neural networks", "deep learning"]
  }
}

1. Keyword Generation: Generate keywords for the research topic
2. Paper Crawling: Fetch papers from the specified conference and year
3. Paper Filtering: Filter papers by topic relevance
4. Paper Summarization: Download PDFs and generate summaries
5. Summary Aggregation: Create structured Excel reports

The agent includes comprehensive error handling:

• Graceful failure with detailed error messages
• State preservation for resumable workflows
• Logging for debugging and monitoring

Run the test suite:

python run_tests.py

Or run individual tests:

python -m unittest tests.test_workflow

• langchain: Agent framework and tool integration
• langgraph: Workflow orchestration and state management
• pydantic: Data validation and configuration management
• requests: HTTP requests for paper crawling
• pymupdf: PDF parsing and text extraction
• pandas: Data analysis and Excel export
• tqdm: Progress bars for long-running operations