Skip to content

Raw to Knowledge User Guide

Raw to Knowledge is a governed knowledge pipeline. It ingests raw information artifacts such as meeting transcripts, support conversations, notes, Markdown pages, HTML pages, and PDF-derived text, extracts question-and-answer candidates, classifies them by information type, routes them through role-based human validation, and publishes approved answers into a versioned registry — where they are available for readouts, semantic retrieval, graph analysis, and team dashboards.

This guide is organized by role and task. Find your path below, then follow the links to the content most relevant to your work.


Which path is yours?

Persona Primary need in Raw to Knowledge Recommended starting point
Consultant / Account Lead Prepare for customer meetings using curated, approved answer readouts; retrieve answers on demand Core ConceptsReadouts & Retrieval
Validator / SME Review extracted and generated answer candidates; approve, revise, or escalate Validation LifecycleValidation Workflow
Technical Writer / Content Strategist Export registry answers, apply Horn-type guidance, manage knowledge lifecycle Horn Information TypesRegistry Export
Support Engineer Retrieve troubleshooting answers, inspect contradictions, surface post-meeting knowledge gaps Flow A vs Flow BSearch & Filter Answers
Knowledge Architect / Governance Lead Govern the full answer lifecycle, manage quality, monitor dashboard metrics Governance OverviewDashboard
Platform Admin / Integration Developer Configure services, connectors, Celery workers, API integrations, and monitoring System Surfaces & Data StoresAdministration

The Raw to Knowledge knowledge pipeline

The diagram below shows the highest-level view of the pipeline. Every source artifact that enters Raw to Knowledge travels this path before its knowledge becomes reusable.

flowchart LR
    T([Transcript\nVTT / SRT / TXT / JSON])
    I[Intake\nSegmentation · Privacy · Flow A/B]
    C[Classification\nHorn Type · Assertion Grammar]
    V[Validation\n9-State Machine · Role Authority]
    R[(Registry\nApproved Answers\nVersion Chain)]
    RD[Readouts]
    RT[Retrieval]
    G[Graph\nNeo4j · Contradiction]
    D[Dashboard\nMetrics · Quality]

    T --> I --> C --> V --> R
    R --> RD
    R --> RT
    R --> G
    R --> D


Note: The full interactive API reference is served by the running FastAPI service at /docs (OpenAPI/Swagger UI) and /redoc. System architecture diagrams are in the Architecture overview.