README.md

metadata

title: AI Research Paper Analyst
emoji: 🔬
colorFrom: blue
colorTo: indigo
sdk: gradio
python_version: '3.10'
app_file: app.py
pinned: false
license: mit

🔬 AI Research Paper Analyst

Automated Peer-Review System powered by Multi-Agent AI (CrewAI + GPT-4o).

🚀 Features

• Multi-Agent Pipeline: 7 agents (1 safety gate + 6 analysis) review methodology, novelty, and quality.
• Programmatic Safety: Deterministic guardrails for PII redaction, prompt injection detection, and malicious URL blocking (regex/logic-based — no LLM).
• Structured Output: All agent outputs enforced via Pydantic schemas.
• 15-Point Rubric: Binary quality scoring with automatic enhancement for failed criteria.
• Interactive UI: Gradio interface with 6 tabs, real-time progress tracking, and report export.

🛠️ Agents

1. Safety Guardian — Programmatic gate that blocks unsafe documents (no LLM).
2. Paper Extractor (GPT-4o) — Structures raw text into metadata, methodology, findings.
3. Methodology Critic (GPT-4o-mini) — Evaluates study design, reproducibility, bias risks.
4. Relevance Researcher (GPT-4o-mini) — Checks novelty via Semantic Scholar / OpenAlex.
5. Review Synthesizer (GPT-4o-mini) — Drafts the initial peer-review report.
6. Rubric Evaluator (GPT-4o-mini) — Scores against 15 strict binary criteria.
7. Enhancer (GPT-4o-mini) — Polishes the final output into a publication-ready report.

🔧 Tools

Tool	Purpose
pdf_parser	Extract text from PDF (pdfplumber)
pii_detector	Detect & redact PII (emails, SSNs, phones, credit cards)
injection_scanner	Detect 9 prompt-injection patterns
url_validator	Blocklist check for malicious URLs
citation_search	Search Semantic Scholar / OpenAlex for related papers

📚 How to Use

Prerequisites

• Python 3.10+
• OpenAI API key with GPT-4o access

Setup & Run

# 1. Install dependencies
pip install -r requirements.txt

# 2. Create .env file with your API key
echo "OPENAI_API_KEY=your-key-here" > .env

# 3. Launch the app
python app.py

Open http://localhost:7860 in your browser, upload a PDF, and click "Analyze Paper".

On HuggingFace Spaces

Add OPENAI_API_KEY as a Secret in Space Settings.

📁 Project Structure

├── app.py                    # Main pipeline + Gradio UI
├── agents/                   # 6 CrewAI agent definitions
├── tools/                    # 5 custom tool implementations
├── schemas/models.py         # 8 Pydantic output schemas
├── requirements.txt          # Python dependencies
├── walkthrough.md            # Detailed project walkthrough
└── PROJECT_DOCUMENTATION.md  # Full technical documentation

🔑 Environment Variables

Variable	Required	Purpose
OPENAI_API_KEY	Yes	OpenAI API access (GPT-4o required)

---

Homework 5 — Agentic AI Bootcamp