EmbedGuard

Overview

EmbedGuard is a Python-based observability tool for RAG systems. It aims to detect adversarial embedding attacks and provide provenance attestation for retrieved content across multiple layers. The project is in an early community stage with no released version or star history.

Best for

Best for
Security researchers and RAG developers evaluating defenses against embedding-level attacks.

Use cases

• Detect poisoned or manipulated embeddings in RAG retrieval
• Verify the provenance of chunks before they reach the LLM
• Monitor RAG pipelines for cross-layer attack indicators

Notes

EmbedGuard is a Python-based observability tool for RAG systems. It aims to detect adversarial embedding attacks and provide provenance attestation for retrieved content across multiple layers. The project is in an early community stage with no released version or star history.

0 stars on GitHub. Last updated 2026-09-29. Licensed MIT.

Use cases

• Detect poisoned or manipulated embeddings in RAG retrieval
• Verify the provenance of chunks before they reach the LLM
• Monitor RAG pipelines for cross-layer attack indicators

Pros

• Addresses a specific and underexplored RAG security surface
• Open source under a community vendor with no vendor lock-in
• Python makes it easy to integrate with existing ML workflows

Cons

• No stars or release history to indicate maturity or adoption
• No documentation, benchmarks, or usage examples available
• Scope is limited to adversarial embedding attacks, not general observability

Indexed from awesome-llmops and enriched against its public facts.