Back to the catalog

HelixPay — Company Brain

Bundle OKF 0.1 · 6 conceitos · vood/company-brain-demo

Open source Repository Open in the app JSON README (API)

About

# HelixPay — Company Brain

An Open Knowledge Format bundle compiled from the HelixPay data snapshot.
Each concept links to its sources; contested and superseded facts are flagged inline.

## Contents

- [People](./people/index.md) — 13
- [Teams](./teams/index.md) — 3
- [Products](./products/index.md) — 5
- [Customers](./customers/index.md) — 2
- [Projects](./projects/index.md) — 4
- [Metrics](./concepts/index.md)

## Provenance

Compiled 2026-07-08T12:00:00.000Z from 44 source documents · 27 entities · 31 claims · 15 relationships · 10 metrics.

Details

Kind
OKF bundles
Topic
Cloud & DevOps
Publisher
vood
Origin
okf_github
Category
dados
Version
0.1
Last push
2026-07-13T10:47:17Z
Repository state
ativo
Language
HTML
Added
2026-09-08 09:05:43
Updated
2026-09-08 09:05:43
Origin id
vood/company-brain-demo:knowledge/index.md

README

# HelixPay Company Knowledgebase

This repository contains a source-backed knowledgebase for a fictional B2B payments
company, **HelixPay**.

It has two parts:

- `data/` — the original messy company context dump.
- `knowledge/` — a curated, agent-readable knowledgebase compiled from that dump.

An additional self-contained [Data Analytics MCP UI demo](examples/data-analytics-demo/README.md)
shows how Codex can turn synthetic product-growth data into an interactive,
source-backed report with KPI cards, charts, and a sortable evidence table.

There is no app server, API, package install, or runtime required. The deliverable is
the knowledgebase itself.

## What Is Included

The checked-in knowledgebase covers:

- 44 source documents
- 27 resolved entities
- 31 sourced claims
- 15 relationship edges
- 10 scoped metrics

The structured source of truth is:

```text
knowledge/compiled/ontology.json
```

The markdown knowledgebase is organized by concept type:

```text
knowledge/people/
knowledge/teams/
knowledge/products/
knowledge/projects/
knowledge/customers/
knowledge/concepts/
```

Start with:

```text
knowledge/index.md
```

## What The Knowledgebase Demonstrates

The knowledgebase is intentionally built around the hard parts of company context:

- **Entity resolution** — Maria Santos vs. Maria Silva, Daniel Tan vs. Tan Wei Ming,
  and HelixPay POS vs. POS Self-Service.
- **Contradictions and stale information** — public June messaging for Project
  Confluence vs. the internal/board Q3 timeline; Q3 vs. Q4 timing for company-brain
  platform value.
- **Scoped metrics** — Q1 revenue, Q2 pacing gap, NPS by segment, Brazil SMB ARR at
  risk, affected merchants, and churn impact.
- **Causal chains** — Tap + Loyalty reconciliation defect, Açaí Express risk,
  Cosmos Hotels churn, and multi-property reporting as an enterprise blocker.
- **Org and ownership context** — reporting lines, ownership, blockers, and
  dependencies.

Each concept file includes citations back to the original source documents.

## Layout

```text
data/
  Raw company corpus: docs, chats, emails, interviews, dashboards, PDFs, and images.

knowledge/
  Curated knowledgebase.

knowledge/compiled/ontology.json
  Machine-readable ontology containing documents, entities, claims, edges, and metrics.

knowledge/index.md
  Human/agent entrypoint for browsing the markdown knowledgebase.

knowledge/log.md
  Compilation log and superseded fact notes.

examples/data-analytics-demo/
  Synthetic Data Analytics MCP report, executed notebook, and reproducible artifact builder.
```

## Source Corpus

The raw `data/` directory contains:

```text
overview.md
org-chart.md
all-hands-2026-04-15.md
weekly-review-2026-04-21.md
board-update-2026-04-22.md
q1-2026-results.pdf
board-deck-q1-2026.pdf
interviews/
dashboards/
images/
chat/
email/
code/
```

The corpus is deliberately realistic: multiple formats, inconsistent naming, mixed
languages, stale documents beside fresher ones, informal Slack alongside board
materials, and contradictions that should be preserved rather than flattened.

## Notes

- All people, companies, and events are fictional.
- The knowledgebase is checked in and usable as files.
- No API key is required.
- No build step is required.

More