Project as an Operational Boundary
A project is the primary boundary for organizing AI assets and operations in Bud AI Foundry. Most day-to-day activities—deployment lifecycle, routing, analytics, and safety controls—are scoped to a project.Project Scope Diagram
Core Concepts
1. Project Metadata
Each project has a name, icon, description, and tags. Metadata helps teams discover and classify workloads across environments.2. Permission Scope
Project permissions determine who can:- View project assets
- Manage project settings and members
- Create/edit/delete deployments and routes (depending on role scope)
3. Tab-Based Functional Areas
Project detail pages are organized into focused tabs for operational clarity:- Deployments: endpoint lifecycle and runtime settings
- Agents: prompt- and agent-centric assets
- Routes: traffic distribution strategy
- Analytics: performance and capacity trends
- Clusters: infrastructure placement context
- Observability: runtime traces, logs, and request signals
- Guardrails: policy and safety controls
4. Safe Lifecycle Management
Project deletion is intentionally guarded. If active deployments are present, deletion can be blocked until dependencies are removed or paused.5. SLO-Aware Scaling
Deployment settings can be tuned with SLO-aligned autoscaling signals (for example latency or queue-related metrics), allowing teams to balance performance and cost.How Projects Relate to Other Modules
| Module | Relationship to Projects |
|---|---|
| Models | Models are selected and deployed into project-scoped endpoints |
| Deployments | Deployments are managed and governed within projects |
| Routes | Routing policies are configured from project context |
| Observability | Monitoring is filtered by project for faster diagnostics |
| User Management | Access is granted to project resources via RBAC |
Best Practices
Use one project per product, tenant, or major workload boundary.
Keep project tags consistent for reporting and governance.
Grant manage permissions to a minimal set of operators.
Review analytics and observability before scaling or route changes.
Enable guardrails early in lifecycle, not only after incidents.