Skip to content

Platform blueprint

One delivery protocol. Four workspaces. Clear accountability.

FYY models expert delivery through Providers, Expert Services, Customer Workspaces, Cases, and immutable Expert Twin Versions. Skills, MCP servers, models, and knowledge systems remain capability components instead of carrying the entire product model.

Each role sees exactly the context needed to do the work

01 Platform operator

Platform Admin

Govern tenants, risk, metering, disputes, and infrastructure without reading raw expert knowledge or client content by default.

02 Provider admin

Provider Console

Design service SKUs; configure teams, workspaces, permissions, budgets, SLA, deployments, and releases.

03 Expert

Expert Studio

Claim handoffs, inspect evidence and AI drafts, use tools, revise, and take responsibility for delivery.

04 End customer

Customer Workspace

Open cases, add context, follow progress, receive results, request a human, and manage members and retention.

How one professional delivery moves

  1. 01

    Client request

    A case starts with service scope, deadline, and workspace context.

  2. 02

    Policy execution

    The runtime loads a released Twin version, client memory, and least-privilege tools.

  3. 03

    Quality gate

    Confidence, evidence, risk rules, budget, and action permissions are evaluated.

  4. 04

    Human takeover

    One expert is atomically assigned; AI becomes an internal drafting assistant and cannot deliver directly.

  5. 05

    Versioned delivery

    The result records evidence, actions, accountable owner, and knowledge version.

Cloudflare coordinates delivery. Expertise stays in a trusted environment.

Cloudflare / control plane

  • Identity and tenant context
  • Service, Case, and Handoff metadata
  • Real-time collaboration and orchestration
  • Artifacts, metering, and audit index

fyy runtime / data plane

  • Knowledge ingestion, indexing, and ontology mapping
  • Client-isolated memory
  • Models, harnesses, and tool execution
  • Output policy and detailed local audit

The default path only requires the expert runtime to open an outbound HTTPS/WSS connection. No network driver or inbound port is required. The existing mesh remains an optional direct-connect optimization.

A working vertical slice is already underway

01 D1 multi-tenant Case and Handoff model
02 Outbound WSS Deployment Session and disconnect recovery
03 Optimistic concurrency and one accountable owner
04 Customer-visible versus Provider-internal event isolation
05 Data-plane risk policy and structured handoff packet
06 Three-service Docker Desktop topology
Open the local delivery workspace