The AI‑ready data warehouse that lives inside your Oracle Database
Cosmos extracts from Oracle Fusion ERP and EPM, builds a governed finance star schema, and turns every report your users save into a reusable data API — with no middle tier: no ETL server, no integration platform, no extra infrastructure to license, patch or secure. Everything runs as PL/SQL and Oracle APEX, where your data already lives.
1,462
curated Fusion BICC datastores, catalogued and ready to extract
65,928
curated columns with keys, types and incremental watermarks pre-mapped
24
star-schema reporting views — 11 dimensions, 13 facts across GL, AP and AR
0
middle-tier servers — the entire platform is Oracle Database + APEX
How it Works
Source to insight, entirely in-database
A data shuttle is Cosmos’ unit of movement: source interface in, target interface out. Shuttles compose into load plans with serial and parallel steps, dependencies, scheduling and full execution logging.
Generating a warehouse is a single guided step: choose a functional area — Financials, Procurement, Supply Chain, Projects or HCM — and Cosmos creates the BI Cloud Connector job on your Fusion pod, enables exactly the columns the curated catalogue marks as needed, extracts, and builds the load pipeline. No hand-mapping of view objects.
SPOTLIGHT· SELF-SERVICE REPORTING
Build a report once. Call it from anywhere.
In most BI stacks, a report a user builds is trapped in the tool that built it. Cosmos separates every report into two portable artifacts: a Saved Filter (what to select — ledgers, periods, cost centres, hierarchy nodes, exclusions) and a Saved Report (how to lay it out — dimensions and measures). Both can be personal, shared, or set as defaults.
The Saved Report API then makes each combination a governed, callable dataset. Reference a filter and report, and Cosmos returns exactly the data the report requests — ready to use in any other tool, and ready to publish as a REST service. The figures in your board pack, your Power BI model, your finance extract and your AI prompt are the same figures, from the same definition.
ONE DEFINITION · EVERY CONSUMER
SAVED FILTER – THE WHAT
Ledgers, periods, cost centres, hierarchy nodes, exclusions
SAVED REPORT – THE HOW
The dimensions, measures and layout to present
ONE GOVERNED DATASET
The same figures in APEX, your BI tools, downstream systems — and your AI prompts
Subject areas today:
GL balances (incl. 24-month pivot), GL transactions, Accounts Payable, Accounts Receivable and Staffing — each with KPI totals for Actual, Budget, Forecast and Plan scenarios, plus variances.
FROM REPORT TO INSIGHT
ANY SAVED REPORT
The governed dataset your users already trust
YOUR AI PROVIDER
Ask anything free-form, or run curated pre-set prompts
DEEPER INSIGHT
Narrative analysis, variance explanations, anomalies worth a closer look
AI-ready, not AI-required
Cosmos detects your database’s capability level automatically. It runs fully on Oracle 19c estates today, and the in-database vector features simply light up when the platform is upgraded — no re-implementation, no migration project.
SPOTLIGHT· AI-READY DATA WAREHOUSE
AI-ready data. Insight on your terms.
Cosmos builds its search capability on the database’s own vector engine, so documents and data are chunked, embedded and searched inside your database — and when you want generative insight, you decide which AI provider sees your data, and exactly what it is asked.
Bring your own AI.Pass the data from any report to the AI provider of your choice — with free-form questions or governed, pre-set prompts — and turn the numbers into narrative insight, variance commentary and recommendations.
In-database embeddings.A sentence-transformer model loads directly into Oracle; documents are chunked and vectorised by the database’s own vector engine — no external embedding service, no per-call AI costs.
Semantic search, built in.A ready-made search UI ranks content by vector distance — find “that invoice about scaffolding hire” without knowing the exact words.
Document understanding.Integration with OCI Document Understanding extracts structure from scanned documents landed in your data lake catalogue.
A warehouse LLMs can trust.Curated metadata, conformed dimensions and a verified star schema are exactly the foundation AI assistants and agents need — governed, described, reconciled data rather than raw extracts.
Platform capabilities
Everything a finance data platform needs, nothing to bolt on
One-step Fusion warehouse
Pick a functional area and Cosmos does the rest — creates the BICC job on your pod, extracts, generates warehouse tables and incremental loads from a curated catalogue covering Financials, SCM, Procurement, Projects and HCM.
1,462 datastores · 65,928 curated columns · keys & watermarks pre-mapped.
Orchestration built in
Load plans sequence shuttle groups with serial or parallel steps, dependency waits, retries and email alerts. Scheduling rides on the database’s own scheduler — one less moving part to operate.
Full execution log per run, task and shuttle.
Multi-cloud landing zone
Extracts land in the object storage you already run — OCI, Azure Blob or AWS — and are queried in place as external tables before loading. Your files stay in your tenancy, under your retention rules.
Cloud access brokered through one hardened, audited package.
Finance-grade star schema
Conformed dimensions and facts for GL balances with budget variance, journals, AP spend, AR income and debtors — hierarchy-aware, ledger-scoped, and reconciled back to journals by an automated verification suite.
COA segment roles resolved from your chart, never hard-coded
Security by design
A deny-by-default model maps users to organisations, ledgers, balancing segments and cost centres. Grant one hierarchy node and access flows to every cost centre beneath it — and updates when the hierarchy does.
PBKDF2 password verifiers · lockout · login audit · install-time verification
Commentary where the numbers live
Attach commentary directly to GL balances, transactions, AP invoices, AR transactions and purchase orders. Reports flag commented figures automatically, so the story behind a variance travels with the number.
The month-end narrative, kept with the data it explains
At a glance
Technical Summary
| Platform | Oracle Database (19c → 26ai) — cloud or on-premises. Vector and AI features enable automatically on capable databases. |
|---|---|
| Front end | Three Oracle APEX applications: Cosmos Console (ELT, shuttles, load plans), Cosmos Reporting, Cosmos Security. |
| Sources | Oracle Fusion ERP (BI Cloud Connector, BI Publisher, REST APIs), Oracle EPM (data & metadata extracts), object storage files, database tables, views and PL/SQL transformations. |
| Storage | OCI Object Storage, Azure Blob Storage, AWS Object Storage; external tables for query-in-place. |
| Reporting | Star-schema layer (11 dimensions, 13 facts) over GL, AP, AR and staffing; saved filters and reports with sharing and defaults; row-level commentary on figures; KPI scenarios for Actual, Budget, Forecast and Plan. |
| Integration | Saved Report API returning exactly the data each report defines — consumable from any SQL-capable tool, ready for REST publication, and able to feed report data to downstream systems and AI services. |
| AI | Report data passed to your choice of AI provider with free-form or pre-set prompting for deeper insight; in-database embeddings and native vector search; semantic search UI; OCI Document Understanding integration; data lake object catalogue. |
| Operations | One-time DBA provisioning, then routine upgrades that need no DBA involvement; every change fully audited and reversible. |
See your Fusion data in a new light
Cosmos is built and supported by Beyond, specialists in Oracle ERP, EPM and data platforms — and proven in production on complex, security-sensitive multi-organisation estates.
Oracle, Oracle Fusion, Oracle APEX and related marks are trademarks of Oracle Corporation. Power BI, Excel and Tableau are trademarks of their respective owners; Cosmos integrates with them through standard SQL and REST interfaces. Feature availability may vary by release — details describe Cosmos 2.x.