¶ Intelligence Domain
The Intelligence domain enables AI models, agents, and toolchains to participate as governed, explainable actors in Fuse workflows.
Rather than treating AI as a black box, this domain turns it into a composable, auditable component that can be invoked, monitored, and refined alongside human, policy, and system steps.
Fuse treats AI as a native domain — allowing models and agents to be:
- Triggered by workflows
- Routed dynamically based on context
- Constrained by identity and policy
- Monitored, traced, and explained in real time
From autonomous agents to human-in-the-loop collaboration, the Intelligence domain ensures AI operates with transparency and governed control.
- Prompt Execution and Routing — Dynamically select and execute AI models based on input, role, or policy
- Tool Orchestration — Use AI to select and invoke tools or services within workflows
- Session Management — Track multi-turn conversations, context threads, and memory
- Explainability and Feedback Loops — Make AI-driven decisions observable and reviewable by humans
- Bias Detection and Drift Monitoring — Log AI performance and identify anomalies over time
- Scoped AI Access — Restrict tool or model usage based on user roles, domain rules, and runtime policies
- AI SDK Integration — Extend Fuse to incorporate custom models and external inference systems
| Layer |
Intelligence Role |
| Integration |
Connect to external model APIs or inference servers |
| Automation |
Use AI to trigger or evaluate logic based on events and inputs |
| Orchestration |
Invoke AI as part of workflow steps, approvals, or decisions |
| Interoperability |
Apply runtime context, scoped governance, and feedback routing |
The Intelligence domain ensures that AI tools and agents operate under secured, observable runtime conditions — not as isolated external services.
- A user uploads a document. Fuse routes the file through an AI model to classify, summarize, and flag it for review.
- An AI assistant helps a tax preparer complete a complex return, dynamically routing ambiguous cases to human reviewers.
- A multilingual AI agent selects the appropriate language model in real time to assist client support interactions.
¶ Extending the Intelligence Domain
Use these frameworks to extend and govern AI workflows:
- AI Framework — Register tools, agents, and models for orchestrated execution.
- Workflow Framework — Invoke AI steps inside modular, reusable workflow logic.
- UI Framework — Build human review panels, AI decision displays, and feedback interfaces.
- Log Framework — Monitor AI behavior, cost, performance drift, and explainability.
- Workflow Domain — Orchestrate human and AI collaboration across structured steps.
- Knowledge Domain — Feed AI models with context-rich embeddings and retrieval workflows.
- Security Domain — Enforce token-scoped access, approvals, and role-based policy gates on AI operations.
- Glossary — Learn terms like prompt routing, explainability, feedback loops, AI governance, and runtime observability.