Terminology
Sema4.ai Product Components
Actions
Actions are tools for agents that let them search the Internet, process data, or work directly with your enterprise applications and data. We have a gallery of actions for common applications like Google Workspace, SharePoint, HubSpot, and more.
Action server
The actions server in Sema4.ai is a backend service that handles executing specific actions requested by agents. It processes defined workflows, interacts with external APIs, and performs tasks like data retrieval or automation steps. The actions server serves as the operational layer that enables agents to interact dynamically with other systems and execute predefined commands effectively within their environment.
Agent
An agent in Sema4.ai is a set of instructions and actions designed to autonomously perform specific tasks within enterprise environments. It uses LLMs, data, and documents to make decisions, all while integrating with existing business systems to automate end-to-end business processes.
Sema4.ai provides two types of agents:
- Conversational agents get tasks from users via chat.
- Worker agents work autonomously based on workload assigned by automation.
Agent Compute environment
Agent Compute environment (ACE) is a dedicated Kubernetes cluster infrastructure that runs in your AWS VPC. It provides the computing resources needed to execute tasks, handle workflows, and perform data operations securely. ACE is designed to support agents in their autonomous activities, ensuring optimal performance and integration with enterprise systems.
You can have multiple ACE instances, typically to separate production and development environments. Different environments may reside on different AWS accounts.
Agent Server
The Sema4.ai Agent Server is a dedicated runtime environment for hosting and executing AI agents. It provides the necessary infrastructure to run agents securely and efficiently, managing resources, handling agent lifecycle, and facilitating communication between agents and other Sema4.ai components. The Agent Server ensures that agents operate in a controlled, scalable environment, supporting both development and production deployments.
Control Room
The Control Room acts as a centralized management hub for overseeing the operations of AI agents. It helps users monitor agent activities, manage workflows, and ensure secure execution. It provides tools for tracking performance, scheduling tasks, and troubleshooting issues, giving users visibility and control over their deployed agents within enterprise environments.
Data Server
The Sema4.ai Data Server is a crucial component of the Dynamic Data Access solution. It provides a secure and efficient way for AI agents to access enterprise data without creating new data silos. The Data Server enables zero-copy data access, supporting over 100 data sources and allowing agents to interact with structured enterprise data efficiently. It facilitates seamless integration with existing data infrastructure, enhancing the agents' ability to make informed decisions based on real-time, historical, and even predictive data.
DocumentDB
DocumentDB is a specialized, scalable, and high-performance database designed to store, manage, and retrieve structured data extracted from documents by Document Intelligence. It provides efficient storage and query capabilities tailored for document-centric data models.
Document Intelligence
Document Intelligence in Sema4.ai is a powerful solution that enables AI agents to deeply understand and process unstructured data autonomously. It goes beyond simple extraction by allowing agents to interpret complex documents, including PDFs and emails, with high accuracy and context awareness.
SDK (VSCode Extension)
The Sema4.ai VSCode Extension integrates with Microsoft Visual Studio Code and Cursor on your local machine, providing an environment for building and managing custom actions using Sema4.ai's automation-as-code framework. It offers syntax highlighting, auto-completion, and integrated debugging tools, enabling developers to create, test, and debug custom actions within their IDE. The extension supports local action testing and direct publishing to Sema4.ai Studio, streamlining development to production deployment.
Studio
Sema4.ai Studio is a comprehensive and versatile development platform designed for building, testing, and deploying enterprise AI agents direcly on your local machine. It empowers both business users and developers to create sophisticated AI solutions tailored to specific enterprise needs.
Work Room
The Work Room in Sema4.ai is a user interface designed for business users to interact with AI agents in a focused, simplified environment. It provides tools to easily access, manage, and utilize agents for different tasks, ensuring that non-technical users can effectively engage with AI-driven processes to support their workflows without needing in-depth technical knowledge.
Product Editions
Team Edition
Team Edition provides essential features for building and deploying AI agents for Snowflake. It includes Work Room for business users to interact with agents, Studio for developers to build and test agents, and Control Room for managing agents. In Team Edition, Control Room and Work Room run in your Snowflake boundary using Snowpark Container Services.
Enterprise Edition
Enterprise Edition is a comprehensive, highly customizable version of Sema4.ai designed for large organizations with complex needs. It enables customers to build, run, and manage enterprise-scale AI agents in their own AWS Virtual Private Cloud (VPC). In Enterprise Edition, Control Room is hosted by Sema4.ai and Work Room runs in your VPC on ACE.
General Terminology
Agent architecture
Sema4.ai agents can be of two architecture types:
- Assistant: Single agent architecture that uses actions and knowledge to complete work outlined in your runbook
- Plan and execute: Extends the assistant architecture and uses multiple agents to create a multi-step plan and executes each step sequentially.
Embedded vector
An embedded vector is a numerical representation of data, often used in machine learning to capture the characteristics of text, images, or other inputs in a structured form. In Sema4.ai, embedded vectors are used to transform unstructured information into a mathematical format that agents can process to understand context, relationships, and meaning, enabling effective analysis and automation of tasks.
IAM – Identity and access management
Identity and access management (IAM) is a framework used to control user access to systems and resources. In Sema4.ai, IAM ensures that only authorized users can interact with specific agents, actions, or data. It provides security by managing user identities, defining roles, and enforcing permissions within the platform.
LLM – Large language model
A Large Language Model (LLM) is a type of artificial intelligence model trained on vast amounts of text data to understand and generate human-like language. LLMs are capable of recognizing patterns, generating coherent text, and understanding context, making them useful for tasks like natural language processing, translation, and summarization. In Sema4.ai, LLMs are utilized to power language capabilities and decision-making processes of AI agents.
LOB – Line of business
Line of Business (LOB) refers to a specific organizational unit or business segment focused on a particular product, service, or set of related activities. LOBs are typically responsible for their own profit and loss and operate somewhat independently, often with dedicated processes and resources tailored to their market or operational focus. In enterprise environments, AI agents can be aligned to support various LOBs, optimizing specific workflows and decision-making processes within each segment.
OIDC – OpenID Connect
OIDC stands for OpenID Connect, an identity layer built on top of the OAuth 2.0 protocol. It enables secure user authentication by allowing third-party applications to verify a user's identity based on the authentication performed by an identity provider. In Sema4.ai, OIDC is used to integrate with external identity providers, facilitating seamless and secure user authentication across platforms.
OAuth2
OAuth2 is an open standard authorization protocol that allows third-party services to access user data without exposing login credentials. It enables users to grant limited access to their resources on one site (the resource provider) to another site (the client), typically by using access tokens. In Sema4.ai, OAuth2 is used to securely connect agents to external services or platforms while maintaining user privacy and security.
Organization
An organization refers to your company or one of its subdivisions. It houses critical elements such as the Agent Gallery, LLM connections, SSO settings, and more. Each organization is assigned one Organization Storage, which provides storage services from your VPC.
Organization Storage
Organization Storage in Sema4.ai refers to the centralized data repository where an organization's resources, such as agents, runbooks, and knowledge bases, are securely stored. It allows multiple teams and users within the organization to access, share, and manage these resources, ensuring consistency and collaboration across different projects. Organization Storage provides controlled access to ensure security and maintain data integrity across the platform.
OTEL collector
An OTEL collector (OpenTelemetry Collector) is a service that collects, processes, and exports telemetry data like metrics, logs, and traces from software systems. It is part of the OpenTelemetry framework and serves as a central component to gather observability data, enabling effective monitoring and troubleshooting across distributed systems. The OTEL collector supports various formats and endpoints to ensure seamless integration with different observability backends.
RPA – Robotic process automation
Robotic process automation (RPA), a technology that uses software robots to automate repetitive, rule-based tasks traditionally done by humans. RPA tools mimic user actions such as data entry, form completion, and transaction processing across various software systems. It is commonly used to improve efficiency and accuracy in business operations by taking over routine tasks, freeing up human workers to focus on more complex and creative work.
Runbook
Define how an agent works using simple, natural language, enabling non-technical users to quickly create agents and update them as their business processes change - without needing to write code.
SSO – Single sign-on
Single Sign-On (SSO) is a user authentication process that individuals can use to log into multiple applications or systems using one set of credentials. In Sema4.ai, SSO simplifies access management by enabling users to authenticate once and gain access to Studio, Control Room, and other connected services without needing to log into each service separately.
VPC – Virtual private cloud
A Virtual Private Cloud (VPC) is an isolated network environment within a public cloud infrastructure. It enables secure communication between resources such as servers and databases by providing configurable network settings, including IP address ranges, subnets, and gateways. VPCs offer control over network traffic flow and are commonly used to ensure secure connectivity for cloud-based services and applications.
Workspace
A workspace serves as a central hub set up under an ACE where you can test, iterate on, monitor, and deploy agents. You divide agents based on your company structure or the purpose of agents you intend to use it for. In a workspace, you also set up necessary configurations, security policies, and deployment settings to ensure smooth agent operations. Each workspace is tailored to specific tasks, helping you manage agents throughout their lifecycle.