guard-rag 1.1.8

Privacy-first, fully offline AI document assistant secured by tiered safety guardrails

🛡️ GuardRAG

Privacy-First, Fully Offline AI Document Assistant

Secured by a Tiered Safety Guardrails System
v1.1.5 — High-Performance, Secure, and Minimalist

Upload any document. Ask anything. Get answers — entirely on your local machine.
No cloud dependencies. No API keys. No data ever leaves your device.

---

💡 Use Cases

GuardRAG is designed for professionals and organizations that handle sensitive data and require the power of LLMs without compromising privacy.

• 🔒 Secure Document Analysis: Chat with confidential contracts, internal financial reports, or legal documents without uploading them to a cloud provider.
• 🏥 Healthcare & Privacy: Analyze medical records or research data locally, ensuring compliance with privacy standards like HIPAA or GDPR through integrated PII detection.
• 💻 Offline Research: Work on-the-go or in air-gapped environments. Once the models are downloaded, no internet connection is required.
• 🛠️ Developer Productivity: Quickly query local documentation or large codebases (via text/PDF) using a streamlined CLI or Web interface.

---

⚙️ Data Sensitivity Tiers

Protect your information using our built-in safety engine:

Level	Protection Scope
🟢 Public	Detects jailbreaks and basic prompt injections.
🔵 Internal	Adds detection for API keys, credentials, and tokens.
🟡 Confidential	Protects SSNs, emails, phone numbers, and credit card info.
🔴 Restricted	Strict protection for medical history, HIPAA/GDPR, and financial data.

---

📥 Installation

Install the package directly from PyPI:

# Recommended stable version
pip install guard-rag==1.1.5

Prerequisites

1. Ollama: Download and install from ollama.com.
2. Model: Pull a model to use locally (e.g., ollama pull gemma3:1b).
3. Windows Users: You must have the Microsoft Visual C++ Redistributable installed to run the AI engine.

---

🚀 Commands & Usage

GuardRAG provides a flexible CLI to interact with your documents.

1. Launch the Web Interface

Simply run the command with no arguments to start the local server and open the UI in your browser:

guard-rag

2. Chat with a Document via CLI

Start an interactive chat session directly in your terminal:

guard-rag --pdf path/to/my_document.pdf

3. Advanced Configuration

Customize the model, server, and safety levels:

guard-rag --pdf report.pdf --model llama3 --sensitivity Confidential --chunk-size 1000

📖 Available CLI Options

Argument	Description	Default
--pdf <file>	Path to the PDF document you want to analyze.	Required for CLI
--model <name>	The Ollama model to use for inference.	gemma3:1b
--ollama-host	The URL of your Ollama server.	http://localhost:11434
--sensitivity	Safety level: Public, Internal, Confidential, Restricted.	Internal
--chunk-size	Size of document chunks for processing.	1000
--no-guardrails	Disable all safety checks (not recommended).	False
--help	Show all available commands and flags.	-

---

🛠️ Quick Example Session

# Start a confidential session with a specific model
guard-rag --pdf Q4_Internal_Report.pdf --model llama3 --sensitivity Confidential

# Chatbot: [Q4_Internal_Report.pdf Loaded] How can I help you?
# You: What were the total earnings mentioned in the summary?
# Chatbot: Based on the document, the total earnings for Q4 were $2.4M...

---

Built with ❤️ by Sowmiyan S

GitHub · PyPI · Documentation