Introduction
FLNet is a platform for executing reproducible, governed scientific tools in a controlled runtime environment.
At a high level, FLNet provides:
- Standardized tool runtime interface (tools implement domain logic, the platform implements orchestration)
- Isolated execution (production runs are strictly containerized)
- End-to-end traceability (configuration, inputs, outputs, and telemetry are captured as provenance)
- Live observability (status transitions, logs, and metrics streamed via WebSocket)
- Security controls (OAuth2/OIDC authentication, runtime credential injection, policy enforcement)
Mental model
Think of FLNet as two cooperating layers:
- Tool layer (inside the container): scientific logic, deterministic transformations, model training/inference, evaluations
- Platform layer (control plane): authentication, validation, orchestration, lifecycle control, provenance storage, auditing
What you should read next
- Architecture at a glance - components and data/control flows
- Security & trust model - where trust is placed and how risks are controlled
- FAQ - common operational questions (dev vs prod, inputs/outputs, failures)