Guardian
Verified Safeby FirebirdSolutions
Overview
An AI safety system designed to detect mental health crises and other harmful behaviors, prevent AI hallucination of fake crisis resources, and provide verified, region-specific support.
Installation
python -m guardian_llm.cli --interactiveEnvironment Variables
- ANTHROPIC_API_KEY
Security Notes
The project emphasizes privacy and on-device deployment, reducing network exposure for user data. It's designed to prevent AI hallucination of crisis resources, which is a core safety feature. The use of `trust_remote_code=True` for Hugging Face models (like Qwen) is standard but implies trust in the remote model's codebase. The `export.py` script uses `subprocess.run` to call external `llama.cpp` tools for GGUF conversion, which introduces dependency on the security of those external tools. No hardcoded API keys were found; API keys are expected from environment variables.
Similar Servers
bifrost
A high-performance AI gateway with a unified interface for multiple LLM providers, offering real-time monitoring and configuration.
archestra
A platform for securely deploying, orchestrating, and managing Model Context Protocol (MCP) servers and their AI-powered tools within an enterprise environment.
ai-counsel
Enables multi-model AI deliberations to reach true consensus through iterative debate and refine positions across multiple rounds, supporting evidence-based decision-making with historical memory.
cortexgraph
A Model Context Protocol (MCP) server providing AI assistants with ephemeral, local short-term memory, temporal decay, reinforcement, and automatic promotion to long-term storage.