Skip to content

Bridge connectivity

Status: Implemented and verified opt-in baseline; permanent default rollout deferred

The bridge-connectivity specification defines how a remote Crystal Lattice frontend discovers, pairs with, and sends HTTP-shaped requests to a registered local server through the distributed bridge service.

It covers the complete current subsystem, including the compatibility WebSocket path and the opt-in HTTP and hybrid frontend paths:

Ownership

The contract is cross-package because no single package owns the end-to-end behavior:

Responsibility Current owner
WebSocket channels, pairing, server registry, sessions, HTTP API, and transfer coordination bridge-elixir
Frontend transport contract and WebSocket/HTTP/hybrid implementations packages/frontend-sdk
Registered server connection and local HTTP forwarding application/python/agent_terminal_app
Saved bridge connection, mode selection, foreground recovery, and repair flow mobile/crystal-lattice-control-rn
Cluster and secure-edge configuration bridge-elixir deployment assets

The formal specification owns the portable target and requirement status. Package documentation describes the current implementation and operating commands without redefining this contract.

Scope

The bridge-connectivity subsystem begins when a local server opens its bridge control WebSocket and ends where that server forwards a proxied request into its local HTTP application. It includes:

  • registration, discovery, health, conflict handling, and cluster visibility;
  • pairing codes, signed pairing credentials, resume, and ephemeral legacy sessions;
  • legacy WebSocket proxying and text chunking;
  • HTTP v1 control/proxy routes and v2 server control events;
  • raw request and response streaming, limits, cancellation, and cleanup;
  • frontend transport selection, persistence, and rollback;
  • single-node, clustered, load-balanced, and TLS-edge behavior.

Explicit exclusions

The specification does not define:

  • the semantics or payload schemas of proxied application routes such as /sessions, /messages, or /events;
  • event contents, polling cadence, or frontend refresh orchestration;
  • agent, provider, plugin, tool, or model behavior;
  • cloud-orchestrator connectivity;
  • direct HTTP or Bonjour discovery beyond distinguishing them from the bridge;
  • mobile presentation outside bridge connection selection and lifecycle;
  • desktop behavior that is not wired to the new bridge modes;
  • historical or abandoned bridge designs.

Sources and evidence

Current claims are bounded to the directly connected bridge modules, SDK transports, Python server adapter, mobile connection host, deployment definitions, and their focused tests. Inspection alone is not accepted as verification. The requirements table distinguishes implemented behavior, automated evidence, pending evidence, and human rollout gates.

Primary delivery tracking:

  • tasks/20260610-mobile-message-refresh-performance/050-plain-http-bridge-large-payload-path.md
  • commit 61140c0e (Add opt-in HTTP bridge transport path)
  • commit 70346cec (Harden HTTP bridge integration coverage)