Acknowledgments
Rhesis includes and depends on the open-source projects below, and uses the standards and reports listed at the end. Thanks to their developers, maintainers, and authors.
Evaluation Frameworks
garak - LLM Vulnerability Scanner
Rhesis uses garak for LLM security testing and vulnerability detection probes.
- Website: garak.ai
- Repository: github.com/NVIDIA/garak
- License: Apache License 2.0
- Copyright: Copyright (c) 2023 Leon Derczynski, Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES
ragas - RAG Assessment Framework
Rhesis uses ragas for Retrieval-Augmented Generation (RAG) evaluation metrics.
- Repository: github.com/explodinggradients/ragas
- License: Apache License 2.0
- Copyright: Copyright 2023 Vibrant Labs
deepeval - LLM Evaluation Framework
Rhesis uses deepeval for comprehensive LLM evaluation metrics including single-turn and conversational assessments.
- Website: confident-ai.com
- Repository: github.com/confident-ai/deepeval
- License: Apache License 2.0
- Copyright: Copyright 2024 Confident AI Inc.
deepteam - LLM Red Teaming Framework
Rhesis uses deepteam for red teaming and adversarial testing of LLM systems.
- Repository: github.com/confident-ai/deepteam
- License: Apache License 2.0
- Copyright: Copyright 2024 Confident AI Inc.
Infrastructure
OpenTelemetry Python
Rhesis uses OpenTelemetry for distributed tracing and observability.
- Website: opentelemetry.io
- Repository: github.com/open-telemetry/opentelemetry-python
- License: Apache License 2.0
- Copyright: Copyright The OpenTelemetry Authors (CNCF)
Transformers - Hugging Face
Rhesis optionally uses the Transformers library for machine learning model integration.
- Website: huggingface.co/transformers
- Repository: github.com/huggingface/transformers
- License: Apache License 2.0
- Copyright: Copyright 2018 The Hugging Face team
Standards and Reports
OWASP Top 10 - LLM and Agentic AI
Rhesis generates security test sets from the published OWASP Top 10 reports. The report is downloaded at generation time and its risk-section text is used as context for the generator, so the tests follow the report rather than a hardcoded attack list. The seeded OWASP judge metrics quote the reports in their rubrics.
- Website: owasp.org
- Project: genai.owasp.org
- Reports: OWASP Top 10 for Large Language Model Applications v2025, OWASP Top 10 for Agentic AI
- License: Creative Commons Attribution-ShareAlike 4.0 International
- Copyright: Copyright © 2003-2025 The OWASP® Foundation, Inc.
OWASP® is a registered trademark of the OWASP Foundation. The OWASP Foundation is not affiliated with any technology company, and its material is used here under the license above — Rhesis is neither affiliated with nor endorsed by the OWASP Foundation.
License Information
The full text of the Apache License, Version 2.0 is available at: apache.org/licenses/LICENSE-2.0
The full text of the Creative Commons Attribution-ShareAlike 4.0 International license is available at: creativecommons.org/licenses/by-sa/4.0
For complete attribution details, see the NOTICE file in the repository root.