Network and traffic architecture

See where EL-Ai-NE sits—and how traffic changes when it is present.

Explore direct user requests, indirect event-driven workflows, SD-WAN service insertion, endpoint control, and disconnected-site operation through live two-dimensional traffic animations.

Status at Baseline v1.0.0. The flows on this page illustrate the baseline design. Direct, governed requests through a site node are Available now. Network Truth (read-only) and endpoint telemetry are Included, enabled per qualification. Service chaining is qualified for each site before use. Disconnected operation, cross-site content reuse and the full endpoint fleet are Planned.

Placement models

EL-Ai-NE can participate without forcing every packet through one box.

The deployment mode is selected according to traffic class, latency, privacy, resilience, control, and evidence requirements. Direct AI use is only one path; many workflows are triggered indirectly by telemetry, events, systems, or endpoint state.

01

Endpoint-assisted

Endpoint clients initiate outbound authenticated control and telemetry sessions. Ordinary application traffic continues through existing enterprise paths.

Best for remote users and device estate control
02

Out-of-band site gateway

Users, applications, and systems call EL-Ai-NE through APIs, agents, connectors, DNS names, or a managed AI gateway—not transparent packet interception.

Best for AI mediation and workflow orchestration
03

Inline transparent

A dedicated management interface controls a bridge pair used for selected traffic. Service interfaces remain unnumbered in transparent mode.

Best for controlled inspection and local serving
04

SD-WAN service insertion

The SD-WAN edge steers selected tunneled or branch traffic through EL-Ai-NE while approved DIA traffic can retain a separate breakout path.

Best for branch and multi-site policy paths

Interactive traffic lab

Compare the same environment without and with EL-Ai-NE.

Choose a scenario, switch the operating mode, then pause or change the animation speed. The moving packets show request, context, policy, action, verification, and evidence paths.

Direct user traffic

A governed AI request with local-first routing

The user asks EL-Ai-NE directly. Identity and policy determine whether local knowledge, a local model, or an approved external route may be used.

What EL-Ai-NE is doing from an AI perspective

  • Resolves who or what is making the request.
  • Retrieves permission-filtered context.
  • Selects the smallest qualified model route.
  • Returns evidence and route metadata.

What changes operationally

  • Private context remains governed.
  • Repeated work can be served locally.
  • External AI is explicit rather than implicit.
  • Cost and evidence are recorded.

Direct and indirect participation

EL-Ai-NE does not need a person to open a chat window to create value.

Direct use begins with a person or application request. Indirect use begins with operational state: telemetry, alerts, tickets, scheduled workflows, source changes, device posture, or a policy event.

Direct path

User or application → EL-Ai-NE → result

  1. Authenticated request enters through web, voice, API, endpoint, or embedded workflow.
  2. Identity, Policy, Brain Trust, and route control determine permitted context and model placement.
  3. EL-Ai-NE answers, recommends, or proposes a typed action.
  4. Any side effect is separately authorized and verified.

Indirect path

Event or state change → EL-Ai-NE → governed workflow

  1. Telemetry, ITSM, security, network, SaaS, or endpoint state enters through a connector or agent.
  2. The operational record and Brain Trust correlate context, history, source health, and contradictions.
  3. EL-Flow creates a durable response plan.
  4. Action Fabric executes only permitted changes and independent verification closes the loop.

Proxy and distributed cache

A later request at another site can be served locally without a second full external download.

EL-Ai-NE checks the local cache first, then an approved regional or peer hotset, and only then the source origin. Every reuse decision rechecks tenant, permissions, source revision, freshness, classification, and tombstone state.

Site A learns the object

An eligible website asset, document, Wiki page, attachment, or software package is fetched once, scanned, hashed, chunked, and stored under an approved content manifest.

Site B requests the same object

The Site B edge performs a local lookup, discovers an eligible manifest, pulls only missing verified chunks from the nearest approved regional or peer source, and reconstructs the object locally.

Origin remains authoritative

ETag, Last-Modified, source revision, Cache-Control, policy epoch, and invalidation state determine when the origin must be revalidated or fetched again.

No user-to-user transfer: the second user does not call the first user or their endpoint. Node-to-node transfer uses authenticated EL-Ai-NE services, short-lived authorization, encrypted transport, content digests, and request-time access checks.

Implementation guidance

Translate the animated model into an address, interface, routing, and firewall design.

The deployment guide covers management and service interfaces, IP addressing, DNS, DHCP, NTP, certificates, firewall rules, MTU, path symmetry, SD-WAN insertion, failure behavior, and acceptance testing.

Ask EL-Ai-NEPublic website guide
Ask about capabilities, governance, privacy, Brain Trust, risk, deployment, pricing, comparisons, or public resources.