Data Model and GraphQL

How Consuelo OS reads structured workspace data.

Data model and GraphQL

Consuelo OS has an ontology and an app schema.

The ontology is how agents understand the company: people, companies, calls, tasks, notes, files, skills, reports, approvals, and artifacts. The schema is how Consuelo stores records today. Some ontology objects already exist as app records. Others begin as artifacts or integration-backed records before becoming first-class objects.

Ontology vs schema

Concept	Meaning	Example
OS ontology	Product map agents use to reason about work	Campaign brief, approval, skill, artifact.
App schema	Records stored in the app today	Person, company, task, note, opportunity.
Integration record	Data owned by a connected system	Ad campaign or call state.
Artifact category	Output tracked by OS first	Report, landing page draft, campaign brief.

This split keeps docs honest. Product docs should name the concept. Implementation docs should say where it lives today.

Structured data path

Skills should use structured GraphQL/API facades for workspace records.

skill
  -> OS data facade
    -> GraphQL/API
      -> workspace records

GraphQL is useful because it preserves workspace scope, user context, object metadata, and backend guardrails.

Current structured objects

The current app already has a strong workspace object model. OS docs can safely talk about people, companies, tasks, notes, opportunities, attachments, dashboards, workflows, workflow runs, calendar objects, messaging objects, connected accounts, and workspace members as structured concepts.

Call, dialer, queue, voice, and cadence behavior exists in Consuelo-specific backend modules and should be surfaced through app-shaped facades when OS skills need it.

Objects that can start as artifacts

Concept	First home	Why
Campaign brief	Artifact	Useful before campaign lifecycle management.
Landing page draft	Artifact	Draft content and provenance matter first.
Ad variants	Artifact or integration draft	External platform state comes later.
Weekly report	Artifact	Reports need storage, provenance, and sharing.
Skill output	Artifact	Durable outputs should be tracked.

Structured first, vector second

Structured records are the source of truth. Vector context is supporting memory, search, and recall.

Use structured queries for exact records, relationships, status, money, permissions, approvals, artifact metadata, and execution logs. Use vector context to locate relevant text, summarize prior decisions, or retrieve supporting context.

First useful data query

The first real revenue brief should fetch a small workspace snapshot through approved data facades: workspace identity, recent calls, recent tasks, recent notes, touched people and companies, useful opportunities, and recent artifacts.

The exact field names should follow the current generated schema. Docs should describe the stable product contract and leave implementation-specific query text to reference docs once the facade is final.

Related

• Context and memory
• Files and artifacts
• Observability