rajeshkumarsomu1983/amazon-bedrock-ai-layer-demo

How rajeshkumarsomu1983/amazon-bedrock-ai-layer-demo works

VIRA360 is a Bedrock AI-layer demo with four project-record capabilities: summarize threads, extract action items, search project history, and answer questions grounded in records. The Node workflow uses local seeded data for summarization and extraction, and managed Knowledge Base retrieval or a local Titan-vector index for search and Q&A. A parallel AgentCore workflow exposes those capabilities through a Strands agent, Gateway tools or local tools, and optional conversation memory. The graph distinguishes these paths; AgentCore resource provisioning and local-index persistence details are only partly evidenced by the excerpts.

Main entry points: bedrock.js Bedrock Wrappers · main.py Strands Runtime · retrieve.js Retrieval Switch · core.py Shared Tool Logic

Components 24

Node Use Cases

Summarize Thread src/summarize.js
Extract Actions src/extract.js
Semantic Search src/search.js
Grounded Assistant src/assistant.js

Node Data And Retrieval

Model Configuration src/config.js
Bedrock Wrappers src/bedrock.js
Seeded Project Records src/data/project-docs.js
Local Index Ingest src/ingest.js
Retrieval Switch src/retrieve.js
Local Vector Index src/retrieve.js

AgentCore Runtime

Agent Invocation agent/invoke.py
Strands Runtime agent/runtime/main.py
Agent Prompts agent/common/prompts.py

AgentCore Tools

Gateway Lambda Tools agent/lambda_tools/handler.py
Local Agent Tools agent/runtime/local_tools.py
Shared Tool Logic agent/common/core.py
Agent Project Records agent/data

Other components

Demo User
Amazon Bedrock
Managed Knowledge Base
Agent User
AgentCore Memory
AgentCore Gateway
S3 Document Fallback

Connections 34

Demo User → Summarize Thread: runs
Demo User → Extract Actions: runs
Demo User → Semantic Search: runs
Demo User → Grounded Assistant: runs
Summarize Thread → Seeded Project Records: reads thread src/summarize.js
Extract Actions → Seeded Project Records: reads thread src/extract.js
Summarize Thread → Bedrock Wrappers: requests summary src/summarize.js
Extract Actions → Bedrock Wrappers: requests extraction src/extract.js
Semantic Search → Retrieval Switch: retrieves passages src/search.js
Grounded Assistant → Retrieval Switch: gets context src/assistant.js
Grounded Assistant → Bedrock Wrappers: generates answer src/assistant.js
Bedrock Wrappers → Amazon Bedrock: calls models src/bedrock.js
Bedrock Wrappers → Model Configuration: reads embedding model src/bedrock.js
Retrieval Switch → Model Configuration: reads region and KB src/retrieve.js
Retrieval Switch → Bedrock Wrappers: embeds query src/retrieve.js
Retrieval Switch → Managed Knowledge Base: retrieves passages src/retrieve.js
Retrieval Switch → Local Vector Index: reads vectors src/retrieve.js
Local Index Ingest → Seeded Project Records: embeds records src/ingest.js
Local Index Ingest → Bedrock Wrappers: creates embeddings src/ingest.js
Local Index Ingest → Local Vector Index: writes vectors src/ingest.js
Agent User → Agent Invocation: invokes
Agent Invocation → Strands Runtime: calls runtime README.md
Strands Runtime → Agent Prompts: uses system prompt agent/runtime/main.py
Strands Runtime → AgentCore Gateway: lists tools agent/runtime/main.py
Strands Runtime → AgentCore Memory: loads session context agent/runtime/main.py
Strands Runtime → Local Agent Tools: falls back to agent/runtime/main.py
Strands Runtime → Amazon Bedrock: runs model
AgentCore Gateway → Gateway Lambda Tools: dispatches tools README.md
Gateway Lambda Tools → Shared Tool Logic: dispatches tool call agent/lambda_tools/handler.py
Local Agent Tools → Shared Tool Logic: delegates calls agent/runtime/local_tools.py
Shared Tool Logic → Amazon Bedrock: calls models and retrieval agent/common/core.py
Shared Tool Logic → Agent Project Records: reads bundled records agent/common/core.py
Shared Tool Logic → S3 Document Fallback: falls back to agent/common/core.py
Shared Tool Logic → Managed Knowledge Base: searches and answers agent/common/core.py

Drawn by GitDiagram on November 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·