Artificial Genius Orthix Agent version 2

The Orthix Unstructured Data Workflow Agent applies breakthrough proprietary non-hallucinatory language model technology to help enterprises unlock measurable value from their unstructured data. Organizations hold vast amounts of valuable unstructured information - and when the right insights are delivered to the right person at the right time, the result is tangible opportunities for sales growth, revenue generation, and other commercial returns.

This product is specifically designed for smaller businesses that want the power of an AI agent without the complexity of managing container infrastructure, VPCs, resource sizing, or orchestration platforms. Deployment is fully managed through Amazon Bedrock AgentCore, making it simple to get started.

Non-Hallucinatory Language Processing: Orthix is built on proprietary technology designed to deliver accurate, reliable, and relevant outputs from unstructured data. Your teams can trust the information they receive and act on it with confidence. Declarative Workflow Logic Without Scripting: Version 2 introduces sophisticated data transformation capabilities that execute directly within the Markdown of the workflow. By adding eager semantics to special Markdown links, the Orthix Markdown itself serves as a fully declarative functional language supporting transparent if-then-else logic, recursion, and looping. Business users and technical teams alike can build complex data workflows using familiar Markdown syntax - no coding required.

Full Agent Harness Included: The Orthix Agent ships as a complete, self-contained package. The included harness provides everything needed to define, execute, and manage unstructured data workflows, reducing the integration burden on your engineering team.

Built for Amazon Bedrock AgentCore: The Orthix Agent deploys directly through Amazon Bedrock AgentCore, eliminating the need to manage container orchestration, VPC networking, or resource sizing - so your team can focus on building workflows, not infrastructure.

• Unstructured data extraction and analysis - Process documents, text, and other unstructured sources to surface actionable insights
• Workflow-driven data transformation - Define multi-step processing pipelines using declarative Markdown-based logic
• Conditional and iterative processing - Build workflows with branching, recursion, and looping to handle complex data scenarios
• Enterprise-grade reliability - Non-hallucinatory language model technology designed to minimize inaccurate or fabricated outputs
• Accessible to every department - Designed so that both technical and non-technical users can build and manage workflows without writing code

Orthix Agent is delivered as a container image through AWS Marketplace, deployable as a Bedrock AgentCore Agent. This fully managed deployment model means you do not need to provision or manage container infrastructure, VPCs, or compute resources - AgentCore handles it all. This is intentional by design for smaller businesses seeking simplicity. A separate server variant is available for larger enterprises requiring standard container infrastructure deployment with full orchestration control.

Users provided with an agent URI can run any number of workflows by invoking the standard container endpoint for Bedrock AgentCore agents (/invocations on port 8080). The POST input is a JSON object with prompt and context fields, where the prompt field contains the PRD in Markdown format and the context field contains the document to be processed. The POST output is a JSON object with a response field containing the workflow output in Markdown format.

The Orthix Agent runs entirely in your AWS account with no dependencies on Artificial Genius. The container meets the enterprise security requirements for AWS Marketplace.

The Orthix non-hallucinatory language models are delivered via Bedrock Custom Model Import, a secure service that provides inference endpoints within your AWS account by hosting the Orthix models in a secure, AWS-managed escrow account that:

1. Does not allow Artificial Genius any access to the prompts, data, or responses you send to the models
2. Does not allow you any access to the proprietary weights of the Artificial Genius Orthix models or the underlying base models used in the import process.

Once the import is complete, the models are accessible via model deployment ARN(s). You will provide these ARN(s) to the teams using the agent, entering them into their PRD prompts in the Agent field.