Agentic Auditor 🕵️‍♂️

Agentic Auditor is an AI-driven contract reviewer that runs a Drafter–Critic agent loop to identify risks and unfair clauses in legal documents quickly and with reduced hallucination.

---

Key features

• Drafter & Critic agent loop to reduce hallucination
• PDF ingestion and chunking
• Local policy retrieval via Qdrant vector DB
• JSON report + visual risk log output
• Self-healing: auto-downloads data and rebuilds DB if missing

---

Architecture / Tech stack

• Frontend: Streamlit
• Orchestration: LangGraph + Groq (GPT OSS)
• Vector DB / Memory: Qdrant
• Ingestion: Unstructured + Poppler
• Python 3.10+

---

Quickstart

1. Clone the repo:

git clone https://github.com/Akshad135/agentic-auditor
cd agentic-auditor

2. Create and activate a virtual environment

macOS / Linux:

python -m venv .venv
source .venv/bin/activate

Windows (PowerShell):

python -m venv .venv
.venv\Scripts\Activate.ps1

3. Install dependencies:

pip install -r requirements.txt

4. (Optional, GPU users) Restore CUDA PyTorch after dependency install (Poppler tends to reinstall torch without CUDA):

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121 --force-reinstall

---

Prerequisites

• Python 3.10+
• Poppler (required for PDF processing)

Install Poppler:

• macOS:

brew install poppler

• Ubuntu / Debian:

sudo apt-get install poppler-utils

• Windows: Download a Poppler binary for Windows like this and add its bin/ folder to your PATH. Restart your terminal/IDE after updating PATH.

---

Configuration

Copy the example environment file and add your Groq API key:

cp .env.example .env
# Edit .env and set GROQ_API_KEY and other variables as needed

---

Running the app

Start the Streamlit dashboard:

streamlit run app.py

Upload a PDF or paste text in the UI. The Drafter and Critic will debate assessments in the sidebar and the app will generate a structured JSON report and a visual risk log.

---

Useful scripts

• System checks (GPU, Poppler, API)

python scripts/check.py

• Rebuild vector DB and playbook

python development/setup_db.py

• Run CLI audit

python development/run_audit.py

---

Project layout

app.py                     # Main Streamlit dashboard
src/agents/                # LangGraph agent logic (Drafter / Critic)
data/                      # Vector DB, raw PDFs, and artifacts
development/               # Scripts: build DB, CLI, data download
requirements.txt
.env.example

---

Notes

• Designed to auto-bootstrap: on first run it will attempt to download training data, derive policies, and build the vector index if missing.
• Reports are produced as structured JSON plus an interactive risk log for review.

---