Skip to main content

Enterprise GenAI Code Base

Implementation Resources and Examples​

This section provides access to our production-ready code repositories for building enterprise-grade GenAI applications. Our modular architecture allows for flexible deployment and customization.

CAP Open Source Version

CAP (Core AI Platform) is an end-to-end solution to configure and deploy AI agents. This version is free for personal use as well as for research. It provides core functionality for cloud infrastructure management and automation, serving as a starting point for users that seek a secure, flexible, and scalable entry into agentic AI development using AWS and LangGraph.

Architectural Concept​

CAP Architecture

One of the architecture concepts of the CAP Platform is the use of AWS private cloud deployed via Terraform. This allows for a secure and flexible environment to deploy and manage the CAP Platform.

The use of AWS ECS and ECR allows for a scalable and flexible environment to deploy and manage the CAP Platform. The LangGraph agents that either come from a CAP Engine Repository or are developed by the customer using LangGraph Studio are deployed to an ECS cluster.

Agentic functions can be executed by ECS tasks that not only can call one or more LLMs but also any API to either gather information or execute actions.

For applications that need a user interface, the CAP Frontend provides a modern chatbot interface that can be customized to the needs of your application. It runs server-side rendering to minimize exposure and runs in a separate container, deployed to a separate ECS cluster. Each cluster can be accessed via a load balancer and a custom domain name.

CAP Platform Codebase​

The CAP Platform codebase consists of multiple repositories that are designed to be modular and can be extended to include additional functionality. You can choose to use only the repositories that you need.

CAP Architecture

For the CAP ENGINE, you can choose between multiple implementations as a starting point. The engine contains the agent design that needs to be customized for the specific use case in most cases.

Core Repositories​

1. Cloud Infrastructure (cap-cloud-ce)​

Enterprise-grade AWS infrastructure:

  • Multi-account setup
  • Security configurations
  • Network architecture
  • Service deployments
  • Terraform modules

2. Data Ingestion (cap-ingestion-ce)​

Scalable data processing pipeline:

  • Document processing
  • Data transformation
  • Knowledge base integration
  • Vector store management

3. GenAI Engine (cap-engine-lite-ce)​

The cap-engine-lite-ce repository contains the core AI functionality of the CAP Platform. It is a LangGraph implementation that deploys a very basic AI agent. Core AI functionality:

  • LLM integration
  • Prompt management
  • Agent orchestration
  • Response generation

Find more sophisticated implementations in the agent library. Think of the cap-engines as disks in a jukebox that enable you to deploy different agents in the same environment.

4. Frontend Interface (cap-frontend-ce)​

User interface and interaction layer:

  • Next.js based UI
  • Multi-tenant support
  • LangGraph integration
  • Customizable chat interface
  • Docker containerization
  • AWS ECR deployment

LangGraph Platform as Alternative Engine

While the CAP Platform uses AWS private cloud to maximize security and can run with AWS LLMs to avoid any exposure, LangGraph Platform can be used for convenience or training, test, and development purposes. The LangGraph Platform is a powerful tool for building and deploying AI agents. It is a flexible and scalable platform that allows you to build and deploy AI agents using LangGraph.

CAP Architecture

Getting Started​

Each repository includes:

  • Detailed documentation
  • Deployment guides
  • Configuration examples
  • Security best practices

Prerequisites​

  • AWS Account
  • Terraform installed
  • Basic understanding of cloud architecture
  • Familiarity with Python and TypeScript
  • Node.js and npm for frontend development