ENGINE 01 / CONTENT INTELLIGENCE BACKBONE

TARGETED
DECOMPOSITION
ENGINE.

The TDE takes any document, deck, video, or podcast and produces atomic intelligence units — small, independently addressable elements tagged on nine analytic dimensions, cached for instant reuse across every application on the platform.

How It Works Request White Paper
9D
Metadata
3
Privacy Classes
6
Resource Types
Watch the decomposition
DECOMPOSITION PIPELINE

FROM RAW CONTENT
TO ATOMS.

Every piece of content that enters the platform follows the same five-stage pipeline. Decomposition is front-loaded. Retrieval is fast and cheap. Cost flips from linear-with-queries to logarithmic.

INGESTION
PDF · DOCX · PPTX · HTML · Audio · Video
OUTPUT
Tagged atoms · Vector index · Cache layer
01
Ingest
Source document accepted. Format detected. Raw text, structure, and embedded media extracted into a staging buffer.
02
Segment
Content is split into candidate atomic units at natural semantic boundaries — paragraphs, slides, chapters, timestamps.
03
Tag (9D)
Each atom is classified across all nine analytic dimensions. Multi-label where appropriate. Confidence scores attached per axis.
04
Index
Tagged atoms are embedded into the vector index and registered in the intelligence cache with TTL and isolation class.
05
Serve
Applications query the cache by any combination of dimensions. Sub-document retrieval returns the precise atoms — never the whole PDF.
9D ATOM SCHEMA

NINE DIMENSIONS.
ONE ATOM.

Every atom emitted by the TDE carries metadata on all nine analytic axes. This is what makes sub-document retrieval work — queries can target any combination of dimensions to find the exact atoms that answer the question.

D1
Persona
10 values — CTO, CFO, End User, Champion, etc. Who this atom is relevant to.
D2
Buying Stage
6 values — Unaware, Problem-Aware, Solution-Aware, Evaluating, Deciding, Post-Sale.
D3
Emotional Driver
7 values — Fear, Aspiration, Frustration, Curiosity, Urgency, Trust, Pride.
D4
Evidence Type
10 values — Stat, Testimonial, Case Study, Analogy, Demo, Benchmark, etc.
D5
Credibility
1–5 scale — Strength of source attribution and verifiability of the claim.
D6
Recency
5 tiers — Breaking, Current, Recent, Aging, Archival. TTL-governed decay.
D7
Economic Driver
ROI · Cost-Out · Speed · Quality · Growth · Risk-Reduction · None.
D8
Status Quo Pressure
Sunk Cost · Change Fatigue · Risk Aversion · Political Cost · Procedural Gravity · Counter-Inertia · None.
D9
Industry
NAICS top sector + SIC division — both required per atom. Cross-geography classification.
CAPABILITIES

WHAT THE TDE
ACTUALLY DOES.

Not a wrapper on a vector database. Not an embeddings-only play. The TDE is a purpose-built decomposition pipeline that produces intelligence, not just retrieval hits.

Sub-Document Retrieval
Not "here's the 40-page PDF." Instead: "here are the three precise atoms that answer the question." Queries hit the atom level, not the document level.
Intelligence Cache
Decomposition is front-loaded. Retrieval is fast and cheap. Cost flips from linear-with-queries to logarithmic. The cache persists across applications.
Three-Class Isolation
Shared · Tenant-isolated · User-isolated. Regulated-vertical privacy with shared-pool economics where appropriate. Isolation inherited by TrueGraph.
Decomposition Profile Registry
Named, versioned, reusable decomposition configurations. A "Sales Enablement" profile decomposes differently than a "Compliance Audit" profile. Both are cached.
Multi-Format Ingestion
PDF, DOCX, PPTX, HTML, audio, and video. Each format has a specialized extraction pipeline — slide decks are decomposed differently than podcasts.
TTL-Governed Decay
Atoms have time-to-live. The recency dimension and TTL system ensure the cache self-cleans — stale intelligence ages out, not overwritten, but properly decayed.