Verbis Graph - GraphRAG Knowledge Retrieval Engine

VERSION: 1.0.0 (Initial Release) RELEASE DATE: December 2025 PRODUCT SKU: VG-GRAPHRAG-FREE-DEMO-V1 LICENSE: Apache License 2.0 PRICING: FREE (AWS infrastructure costs apply)

================================================================================

1. EXECUTIVE SUMMARY ================================================================================

We are pleased to announce the initial release of Verbis Graph Engine Demo on AWS Marketplace. This free container-based product brings GraphRAG (Graph-enhanced Retrieval Augmented Generation) technology to developers and enterprises evaluating next-generation AI knowledge retrieval solutions.

Unlike traditional vector-only RAG systems, Verbis Graph combines semantic vector search with knowledge graph traversal, enabling retrieval of information that spans multiple documents, entities, and relationships. This hybrid approach is ideal for complex queries requiring reasoning beyond simple similarity matching.

This initial release establishes the foundation for GraphRAG-powered knowledge retrieval.

o Proprietary Knowledge Graph Retrieval: Combines dense vector embeddings with structured knowledge graph traversal for better accuracy compared to vector-only RAG systems o Multi-Document Reasoning: Retrieves and synthesizes information across document boundaries, capturing entity relationships that span your entire knowledge base o Workspace-Scoped Isolation: Each user or project gets an isolated workspace with independent document staging, GraphRAG indexing, and chat sessions o Async Indexing with Locks: Per-workspace locks prevent indexing conflicts, ensuring data integrity during concurrent operations

o RESTful API: Integrated FastAPI backend with OpenAPI (Swagger) documentation available at /docs, accessible via a dedicated tab in the Streamlit web interface. o Streaming Chat: Server-Sent Events (SSE) for real-time streaming responses from the GraphRAG retriever o CORS Enabled: Cross-Origin Resource Sharing configured for browser-based clients and single-page applications o Health Check Endpoint: GET /_stcore/health for container orchestration and load balancer integration

o Multi-Format Support: Upload and process PDF, TXT, DOCX, and CSV files for knowledge extraction o Document Staging: Stage documents before indexing to review and validate content o Data Retention: Application data is retained within the container runtime and associated workspace for the lifetime of the container and user account. Data is not guaranteed to persist across container termination or redeployment. External durable storage is not included in the free demo.

o LiteLLM Gateway: Unified interface supporting 100+ LLM providers including OpenAI, Anthropic Claude, and AWS Bedrock o AWS Bedrock Native: First-class support for Amazon Bedrock models with configurable region and model selection o Model Flexibility: Switch between Claude, GPT-4, Llama, Mistral, and other models via environment variables

o Bearer Token Authentication: Secure API access with SERVICE_AUTH_TOKEN for service-level auth plus user bearer tokens for workspace access o Streamlit GUI: Helper interface for visual document management and chat testing

o Metering Hooks: Built-in hooks for AWS Marketplace metering (ingestion, indexing, chat, translation) - disabled in free demo o Container Compliance: Meets AWS Marketplace container requirements for security scanning and deployment

ECR Image URI: 709825985650.dkr.ecr.us-east-1.amazonaws.com/prodigy-ai-solutions/verbis-graph-engine-free:verbis-demo Port: 8080/tcp (HTTP) Health Check: GET /_stcore/health Logging: stdout/stderr (CloudWatch compatible) Base Image: Python 3.11 (Debian-based)

vCPU: 1 vCPU 2 vCPU Memory: 2 GiB 4 GiB Storage: 1 GB Depends on workload

o Amazon Elastic Container Service (Amazon ECS) - Managed container orchestration on AWS o Amazon ECS Anywhere - Run on your on-premises or hybrid infrastructure while managed by ECS o AWS Fargate - Serverless container execution (ECS launch type) o Docker-compatible runtimes - Any Docker-compatible environment for evaluation

SERVICE_AUTH_TOKEN (Required): Service-level auth token (X-Service-Token header)

o AWS Account: Active AWS account with appropriate permissions o IAM Permissions: ecr:GetAuthorizationToken, ecr:BatchCheckLayerAvailability, ecr:GetDownloadUrlForLayer, ecr:BatchGetImage o Networking: VPC with subnets and security group allowing inbound traffic on port 8080/tcp

For local evaluation, run the container directly with Docker:

docker run -p 8080:8080
709825985650.dkr.ecr.us-east-1.amazonaws.com/prodigy-ai-solutions/verbis-graph-engine-free:verbis-demo

1. Subscribe to the product: Click 'Continue to Subscribe' on the AWS Marketplace listing page
2. Accept terms: Review and accept the Apache License 2.0 terms
3. Create ECS cluster: Use existing cluster or create new via ECS console
4. Create task definition: Define container with ECR image URI, port mapping (8080)
5. Configure networking: Assign VPC, subnets, and security group with port 8080 open
6. Launch service: Create ECS service with desired task count
7. Verify deployment: Access /_stcore/health endpoint to confirm container is running

For on-premises or hybrid deployments using ECS Anywhere:

1. Register external instances: Install ECS agent and SSM agent on your on-premises servers
2. Create ECS Anywhere cluster: Configure cluster with EXTERNAL capacity providers
3. Deploy task: Use the same task definition with your registered external instances

Note: ECS Anywhere could incur additional costs (eg $0.01025 per instance-hour).

Key API endpoints available in this release:

Method Endpoint Description

GET /_stcore/health Health check endpoint POST /api/auth/register Register new user account POST /api/auth/login Authenticate and obtain bearer token POST /api/docs/upload Upload documents to workspace POST /api/docs/index Trigger GraphRAG indexing POST /api/chat Query GraphRAG (streaming SSE) GET /docs OpenAPI/Swagger documentation

Note: Some authentication and user management endpoints are provided for demonstration purposes only in the free demo.

IMPORTANT: This is a free demonstration version intended for evaluation purposes only.

o Single-Tenant: One container instance per deployment; not designed for multi-tenant production workloads o No SLA: Best-effort support only; no uptime guarantees or service credits o Demo Data Only: Not suitable for PII, regulated data (HIPAA, GDPR), or production workloads

o Region Availability: Available in all AWS regions supported by AWS Marketplace container products o Data Retention: Application-level data retention during the lifetime of the container and user account. Persistence across container termination or redeployment is not guaranteed. o No High Availability: Single container deployment; no built-in clustering or failover o Metering Disabled: AWS Marketplace metering hooks are present but disabled in free demo

o Enterprise SSO/SAML authentication o Role-Based Access Control (RBAC) o VPC Private Link deployment o SOC 2 compliance features o Dedicated support SLA

Enterprise features are available in Verbis Graph Professional and Enterprise editions.

o Two-Layer Auth: SERVICE_AUTH_TOKEN for service access + workspace operations o Token Security: Always used strong, randomly generated tokens;

o TLS Termination: Use ALB or API Gateway with HTTPS for production deployments o Security Groups: Restrict ports access to trusted IPs or internal VPC only

o LLM queries are sent to configured providers

Issue Solution

Image pull failures Verify ECR pull permissions and AWS Marketplace subscription is active Container OOM killed Increase memory allocation to 4 GiB; large documents require more memory for indexing

Q1 2026: o Native MCP Server for AI Agent integration (Claude, GPT) o LangChain and LlamaIndex framework integrations o Professional tier launch on AWS Marketplace

Q2 2026: o Enterprise SSO/SAML authentication o Role-Based Access Control (RBAC) o SOC 2 Type II certification (in progress) o Amazon Bedrock Agents native integration

Q3-Q4 2026: o VPC PrivateLink deployment option o Multi-tenant SaaS deployment

For the free demo version, support is provided on a best-effort basis: o Response Time: Within 1 business day during EU business hours o Channels: Email and documentation only o Scope: Deployment assistance and basic troubleshooting

Email: support@verbis-chat.com  Documentation: https://docs.verbisgraph.com  Website: https://verbisgraph.com 

This product is provided 'as is' under the Apache License 2.0. ProdigyAI Solutions makes no warranties regarding fitness for a particular purpose. AWS infrastructure costs are the customer's responsibility. This demo is intended for evaluation purposes only and should not be used with sensitive, regulated, or production data.

2025 ProdigyAI Solutions. All rights reserved. Verbis Graph is a trademark of ProdigyAI Solutions.