Skip to content

ShreeGopi/OneSource

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 

Repository files navigation

OneSource

Structured Attention Intelligence

OneSource is a custom intelligence system for studying internet attention patterns.

It starts with ecommerce creatives because those examples have clear, repeatable signals: hooks, emotions, visuals, CTAs, platforms, niches, and creative structures. OneSource normalizes those signals, builds relationships between them, and turns a creative dataset into an exploration workspace.

The goal is not to generate more content.

The goal is to understand the repeatable structures behind attention, trust, and action.

Clean signal > more AI

What It Is

OneSource is structured persuasion infrastructure.

It is designed to answer questions like:

  • which hooks repeatedly show up with specific emotions
  • how visuals, CTAs, and niches reinforce a creative pattern
  • which signals travel across platforms and which stay platform-specific
  • what co-signals appear around a selected attention trigger
  • how a creative library can become reusable intelligence instead of static examples

The current proving ground is ecommerce creative analysis, but the deeper direction is broader:

a structured intelligence layer for internet attention.

What It Is Not

OneSource is not:

  • an AI wrapper
  • a content generator
  • a viral prediction tool
  • a recommendation engine
  • a generic analytics dashboard

AI may later help with tagging, clustering, summarization, and organization.

AI is not the moat.

The moat is the structure:

  • clean signal vocabulary
  • normalized creative data
  • reusable intelligence primitives
  • accumulated signal relationships
  • governed ingestion
  • searchable attention patterns

Architecture

OneSource uses a layered pipeline:

Creative records
-> signal normalization
-> aggregation
-> intelligence pipeline
-> exploration state
-> UI interpretation

The intelligence layer is intentionally separated from the presentation layer. The UI renders the results, but the core pattern logic lives in pure TypeScript modules.

Key architecture areas:

  • src/lib/config - controlled vocabularies, thresholds, and signal weights
  • src/lib/signals - extraction, normalization, and relationship helpers
  • src/lib/intelligence - summaries, patterns, reinforced structures, taxonomy, platform analysis, and exploration orchestration
  • src/components/intelligence - signal tags, weighted fields, pattern cards, and exploration panels

The central idea is that a creative should not just be stored as content. It should become structured signal data that can be compared, traversed, and reused.

Intelligence System

The current intelligence system includes:

  • normalized signal extraction
  • emotion + hook pattern detection
  • relationship building across hooks, visuals, CTAs, and niches
  • reinforced structure analysis
  • platform-specific pattern comparison
  • taxonomy-style clustering
  • strongest co-signal detection
  • weighted signal rendering
  • universal signal exploration
  • workspace lens behavior across tabs

The Workspace Lens is important: when a user clicks a signal like curiosity or before-after, the whole Gallery recalculates around the matching creative set. Overview, Patterns, Structures, Platforms, Library, and Exploration all read from the same active context.

That turns signal exploration from a side panel into the working state of the intelligence environment.

Signal Governance

OneSource treats signal quality as product infrastructure.

The project uses controlled vocabularies and validation rules so the dataset does not drift into messy tags, duplicated labels, or inconsistent naming.

Examples of governed signals:

  • emotions
  • hook types
  • visual styles
  • CTAs
  • platforms
  • niches

This matters because weak signal quality breaks the intelligence layer. If before-after, before_after, and Before After all exist as separate values, the system starts splitting relationships that should be unified.

Clean data is not cleanup work here. It is part of the core product.

Product Shape

Current routes:

  • / - product entry screen
  • /gallery - attention intelligence workspace
  • /admin - protected creative ingestion and signal governance

The Gallery is not just a list of creatives. It is an analysis environment with:

  • Overview
  • Patterns
  • Structures
  • Platforms
  • Library
  • persistent exploration
  • breadcrumb traversal
  • workspace lens state

The Admin area is intentionally narrow. It exists to keep ingestion clean, not to become a full CMS.

Production Hardening

OneSource is moving from local MVP into:

Production Hardening + Real Data Validation

Current production direction:

  • Vercel free/Hobby deployment
  • Supabase Free backend
  • public Homepage and Gallery
  • protected Admin ingestion
  • Supabase Auth + Row Level Security
  • draft | published creative status
  • test | live source type
  • dataset labels for trust

The next learning layer is not more features. It is seeing how the system behaves with real data.

The biggest production risk is dataset quality drift, not infrastructure.

Verification

The app includes lightweight Vitest coverage around the pure intelligence logic.

Useful commands:

cd onesource
npm run test
npm run lint
npm run build

The production build may need network access because the app uses Google-hosted Next fonts.

Technology

Current stack:

  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • Supabase
  • Vitest
  • Vercel target deployment

The system is intentionally simple at the infrastructure level. The complexity belongs in the signal model and intelligence pipeline, not in unnecessary backend machinery.

Documentation

The documentation is organized around the project evolution:

  • step docs explain how the technical architecture developed
  • Pause Phase docs explain how the system was stabilized
  • production hardening docs and SQL describe the move toward live usage

Pause Phase was not a pause in progress. It was the point where OneSource stopped expanding features and made the system coherent.

Intentionally Out Of Scope

OneSource is not currently adding:

  • graph databases
  • AI reasoning agents
  • recommendation systems
  • embeddings or vector infrastructure
  • ontology expansion
  • prediction systems
  • feature-heavy dashboard UX

Those ideas may become useful later.

They are not the right next move yet.

Long-Term Direction

Most tools help people produce more content.

OneSource is trying to understand the structure behind content that earns attention.

Long-term, the direction is closer to:

Bloomberg Terminal for Internet Attention

or:

Creative Genome for ecommerce persuasion.

The value is not more AI output.

The value is accumulated, normalized, searchable attention intelligence.

About

A structured intelligence layer for internet attention.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages