Cloudable.AI is a multi-tenant, vector search platform that integrates with AWS Bedrock for embeddings and LLM capabilities, using PostgreSQL with pgvector for similarity search.

The system consists of:

• API Gateway: Handles API requests for document uploads, KB sync, KB queries, chat, and customer status
• Lambda Functions: Process API requests and interact with AWS services
• PostgreSQL RDS: Stores vector embeddings and customer data using pgvector extension
• S3: Stores uploaded documents
• AWS Bedrock: Generates embeddings and powers LLM responses
• Terraform: IaC for deploying all resources

• Document Management: Upload documents and generate vector embeddings
• Knowledge Base: Query a vector database with natural language
• Chat Interface: Interact with LLM with knowledge base augmentation
• Customer Status: Track customer progress through implementation stages
• Multi-tenancy: Complete isolation between tenants
• RBAC: Role-based access control for different user types

.
├── infras/
│   ├── core/              # Core infrastructure Terraform and setup scripts
│   ├── lambdas/           # Lambda function code
│   │   ├── kb_manager/    # Main Lambda for KB operations
│   │   └── ...
│   ├── sql/               # SQL scripts for database setup
│   └── terraform/         # Additional Terraform configurations
├── test_files/            # Files used for testing
└── *.sh                   # Various utility and test scripts

1. Configure AWS credentials
2. Run the deployment script:

   ./deploy_terraform.sh

3. Set up the database with pgvector:

   ./setup_pgvector.sh

Run the end-to-end test to verify functionality:

./test_e2e_pipeline.sh

This tests:

• Health check
• Document upload
• KB synchronization
• KB querying
• Chat functionality
• Customer status retrieval
• Error handling
• Multi-tenant isolation

• /api/health - Health check
• /api/upload-url - Get presigned URL for S3 upload
• /api/kb/sync - Trigger knowledge base sync for a document
• /api/kb/query - Query the knowledge base
• /api/chat - Chat with or without knowledge base context
• /api/customer-status - Get customer implementation status

All endpoints require authentication via the X-User-ID header.

The application is configured via environment variables in the Lambda function:

• BUCKET_[TENANT] - S3 bucket for each tenant
• RDS_CLUSTER_ARN - ARN of the RDS cluster
• RDS_SECRET_ARN - ARN of the RDS credentials in Secrets Manager
• CLAUDE_MODEL_ARN - ARN of the Claude model in Bedrock
• REGION - AWS region (eu-west-1)

Proprietary software. All rights reserved.