Resource
The Generative AI Stack
a comprehensive suite of services designed to facilitate the creation, deployment, and management of AI-driven applications.
a new generative AI stack has emerged.
a multi-vendor approach.
This multi-vendor, hybrid approach ensures that users have the flexibility to leverage the best tools and services available, while maintaining a cohesive and manageable system.
marketplace.
The Ecosystem of AI Solutions
The Marketplace is a curated collection of AI assets. Creators publish plugins, workflows, assistants, LLMs and agents to a marketplace catalog. Users discover and use assets published by the community.
apps, chats & copilots.
User-Facing Innovations
At the top of the stack, we have Apps, Chats, and Copilots, which represent the user-facing components. These are the applications and interfaces that directly interact with end-users, providing intelligent assistance, automating tasks, and enabling natural language interactions.
agents.
Goal Oriented Achievers
Agents are software systems that use artificial intelligence to perform routine tasks on behalf of users. They are given a goal, and use a sequence of actions to achieve an outcome.
assistants.
Domain Specific Chat Bots
Assistants are a type of chat bot focused on a topic area such as HR or finance. They are given access to the files, plugins, etc necessary to help you perform tasks. Unlike Agents, they are not autonomous, or goal oriented.
generative workflows.
The Automation Backbone
Generative Workflows are the processes that define how different AI services interact and collaborate to achieve specific goals. They are the automated sequences that ensure tasks are executed efficiently and effectively, often without human intervention.
plugins & plugin runtime.
Customization and Extension
The Plugins and Plugin Runtime layer allows for customization and extension of the AI stack's capabilities. By enabling the integration of third-party or custom-built plugins, users can tailor the stack to their specific needs, enhancing the stack's functionality and adaptability.
LLMs & RAGs.
The AI Engines
LLMs (Large Language Models) and RAGs (Retrieval-Augmented Generation models) form the AI engine layer of the stack. These are the sophisticated algorithms responsible for understanding and generating human-like text, making sense of large datasets, and providing the intelligence behind the generative capabilities.
FMops, train/tune, host.
The Operational Foundation
Finally, at the base of the stack, we have FMops (Foundation Model Operations), Train/Tune, and Host services. These are the operational components that handle the training, tuning, and hosting of AI models. They ensure that the AI systems are running optimally, are well-maintained, and can scale according to demand.
agent mesh.
The Integration and Control Hub
The Agent Mesh is the linchpin of the Generative AI Stack, providing a control plane. It simplifies the complexity of working with multiple platforms and vendors by offering a single point of control, making it easier to orchestrate workflows, manage resources, and maintain consistency across the stack.