Load

Your data

Connectors or approved imports bring in emails, cookies, events, CRM rows, tickets, billing, documents, and outcomes.

Store

The trail

Scout keeps ordered attribution paths in your estate (PostgreSQL/pgvector for proof). Fortress moves heavy load into Rust and LanceDB.

Match

Past outcomes

The Rust engine compares the current path with similar journeys that converted, retained, escalated, delayed, or lost.

Return

A task brief

Checked JSON with top examples, confidence, caveats, and ranked next steps for the goal you asked about.

Explain

Your model or team

Your approved model can explain the brief. It does not invent the plan. Elite covers the leadership walkthrough when needed.

Context Engine map →

Architecture / Fortress

The enterprise relationship engine and path-weight relationship store inside the customer estate.

Enterprise/Fortress takes the free open-source Scout data-plane foundation into private enterprise operation: private connectors, approved imports, local credential control, governance control surfaces, audit export, identity integration, sovereign relationship storage, LanceDB path-weight storage, and the proprietary Rust relationship, weighting, traversal, and outcome path matching engine. Fortress sends schema-validated execution packet to the customer's approved model boundary, such as an internal model, approved gateway, or deployment-specific hosted-provider adapter.

View FortressPlatform Architecture

Core layers

How Fortress is structured

01

Private connector/import runtime

Enterprise connectors and approved imports read exact authorised records, metadata, and permitted source structure without exporting customer records to KynticAI product operations.

02

Proprietary Rust engine and LanceDB

Relationship facts are linked, weighted, embedded, traversed, and searched in LanceDB with tenant, source, role, confidence, audit, and masking boundaries before JSON consumption.

03

Customer LLM handoff

Schema-validated execution packet goes to the customer's approved model endpoint, including internal LLMs, approved gateways, or deployment-specific hosted-provider adapters.

04

Operational control

Installer assets, lifecycle status, config delivery, operational routing control surfaces, and support boundaries give enterprise teams a controllable runtime.

Operating flow

The request path through the product

Inject

Systems stay where they are

SQL, CRM, ERP, SharePoint, email engagement, cookies, web events, support, billing, usage, and outcome systems remain under customer control.

Store

Relationships gain attribution

Scout stores each item in PostgreSQL/pgvector for inspectable proof. Fortress uses the Rust/LanceDB runtime when the relationship memory needs high-load enterprise relationship traversal.

Explain

The right model gets JSON, not data sprawl

Fortress feeds the customer's approved model endpoint; the Elite path can show how local/private open-source model routing would fit where approved.

Example signal path

Fortress analyses sensitive supplier relationships inside the customer estate

Illustrative sample data only. Private connectors, permissions, and governance must be confirmed per pilot.

01 / Source

Example fields

ERP.open_po_value = material

QMS.defect_rate_pct = 4.8

WMS.late_shipments_60d = 9

02 / Evidence

What KynticAI creates

supplierRisk = high

commercialExposure = material

sourceTrail = ERP + QMS + WMS

03 / Action

What the business does

brief procurement

escalate quality review

protect open orders

Operating model

How the product stays useful at enterprise scale

Identity

Access follows enterprise roles

Fortress patterns align evidence access with customer identity, role boundaries, and local administration.

Audit

The source trail stays visible

Governance control surfaces and export paths focus on what was read, why it was used, and which policy applied.

Sovereignty

Customer data plane owns operational data

KynticAI can manage licences, support, updates, and aggregate posture while the customer data plane owns operational data.

Integration points

Where it connects to the wider stack

Scout

Free open source

Scout proves the Context Engine data-plane mechanics with source registration, selectors, snapshots, relationship facts, PostgreSQL/pgpath-weight storage, JSON output, and APIs before enterprise deployment.

Enterprise

Rust/LanceDB engine

Enterprise adds the proprietary Rust relationship, weighting, traversal, and LanceDB outcome path matching engine plus private runtime controls and sovereign storage patterns.

Operate

KynticAI operations layer

KynticAI manages commercial metadata, licences, downloads, support, aggregate posture, and updates without raw customer records.

Product Reveals

Fortress turns private relationship memory into JSON for your LLM.

These examples focus on the private runtime, Rust/LanceDB scale path, and customer model handoff.

KynticAI Reveal
Private Runtime

Fortress scenario - connectors, credential boundary, local evidence store

Enterprise relationship analysis without data sprawl

“Fortress runs close to customer systems, reads approved source structure, and keeps raw operational data under customer control.”
KynticAI Reveal
Rust/LanceDB

Fortress scenario - millions of relationship sets, fast comparison, traversal

Fast private comparison at enterprise scale

“The proprietary Rust engine and LanceDB store compare one relationship path with many similar won/lost, converted/not-converted, or saved/lost examples.”
KynticAI Reveal
Customer LLM

Fortress scenario - JSON to an approved model boundary

Keep the customer's model choice

“Fortress sends the relationship output to the customer's approved model boundary, such as an internal model, approved gateway, or deployment-specific hosted-provider adapter.”