NEWS · CODING · #3021
EDB Postgres AI builds in-database SemKB for agentic AI using Amazon Bedrock AgentCore
EDB demonstrates building a semantic knowledge base (SemKB) inside an operational Postgres using the AIDB extension, COMMENT statements and foreign keys for join paths, with governance enforced by Postgres GRANTs and row-level security. The SemKB is exposed to external agents running on Amazon Bedrock AgentCore via the Model Context Protocol (MCP) behind an ALB; the demo runs a Strands Agents agent planning with Anthropic Claude and records OpenTelemetry traces in the aidb_otel schema.
KEY POINTS
- EDB demonstrates building a semantic knowledge base (SemKB) inside an operational Postgres using the AIDB extension, COMMENT statements and foreign keys for join paths, with governance enforced by Postgres GRANTs and row-level security.
- The SemKB is exposed to external agents running on Amazon Bedrock AgentCore via the Model Context Protocol (MCP) behind an ALB; the demo runs a Strands Agents agent planning with Anthropic Claude and records OpenTelemetry traces in the aidb_otel schema.
- Putting the semantic layer inside Postgres with enforced access controls and exposing it via MCP provides a practical, auditable pattern for agentic systems to query live data while preserving governance.
WHY IT MATTERS
Putting the semantic layer inside Postgres with enforced access controls and exposing it via MCP provides a practical, auditable pattern for agentic systems to query live data while preserving governance.