Description
Agno has built-in guardrails for PII detection, prompt injection, and OpenAI moderation, but no integration with AWS Bedrock Guardrails. Bedrock Guardrails is widely used in enterprise for content filtering with configurable policies (hate, violence, PII, denied topics, word filters).
Proposed Change
Add a BedrockGuardrail class in agno/guardrails/bedrock.py that uses the Bedrock ApplyGuardrail API. Supports both input (pre_hook) and output (post_hook) checking, fail-open and fail-closed modes, and standard AWS credential resolution.
Why
Bedrock Guardrails is the most enterprise-adopted content filtering service for AI agents. Agno already has Bedrock support for models, embedders, and rerankers — guardrails is the missing piece.
Description
Agno has built-in guardrails for PII detection, prompt injection, and OpenAI moderation, but no integration with AWS Bedrock Guardrails. Bedrock Guardrails is widely used in enterprise for content filtering with configurable policies (hate, violence, PII, denied topics, word filters).
Proposed Change
Add a BedrockGuardrail class in agno/guardrails/bedrock.py that uses the Bedrock ApplyGuardrail API. Supports both input (pre_hook) and output (post_hook) checking, fail-open and fail-closed modes, and standard AWS credential resolution.
Why
Bedrock Guardrails is the most enterprise-adopted content filtering service for AI agents. Agno already has Bedrock support for models, embedders, and rerankers — guardrails is the missing piece.