We take security seriously. If you discover a security vulnerability in ragnarok-ai, please report it responsibly.

Do NOT open a public GitHub issue for security vulnerabilities.

Instead, please use GitHub Security Advisories to report vulnerabilities privately.

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

1. Acknowledgment: We will confirm receipt of your report within 48 hours
2. Assessment: We will investigate and assess the severity
3. Communication: We will keep you informed of our progress
4. Fix: We will develop and test a fix
5. Release: We will release the fix and credit you (unless you prefer anonymity)
6. Disclosure: We will publish a security advisory

• Keep ragnarok-ai updated to the latest version
• Run Ollama locally — ragnarok-ai is designed for local-first operation
• Don't expose services — Keep Ollama and vector stores on localhost
• Review configurations — Check ragnarok.yaml for sensitive settings

• No hardcoded secrets — Use environment variables
• Validate inputs — Sanitize all user inputs
• Dependencies — Keep dependencies updated, review security advisories
• Code review — All PRs require review before merge

• ragnarok-ai core library (src/ragnarok_ai/)
• CLI tool
• Official adapters (Ollama, Qdrant, LangChain)
• Configuration handling
• Data processing pipelines

• Third-party dependencies (report to their maintainers)
• Ollama security issues (report to Ollama team)
• Vector store security (report to respective projects)
• User misconfiguration

ragnarok-ai is designed to run 100% locally. This means:

• No data is sent to external APIs by default
• LLM inference happens on your machine via Ollama
• Vector stores run locally

The tool reads and writes files for:

• Knowledge base documents
• Test sets
• Evaluation reports
• Checkpoints

Ensure appropriate file permissions in production environments.

When using LLM-as-judge for metrics (faithfulness, relevance), be aware that:

• Malicious content in documents could attempt prompt injection
• We implement basic sanitization, but no system is foolproof
• Review generated test sets before using in CI/CD

Thank you for helping keep ragnarok-ai secure!