# Coomia — Data Intelligence Platform

> Coomia is a privately deployed, Flink-native data intelligence platform. It unifies streaming + batch data pipelines, a Palantir-grade 16-module business ontology, and a five-agent AI analytics super-team in one coherent stack — replacing the typical Airbyte + dbt + Dagster + BI + catalog tool chain. Every license includes the full platform — sold by term (6-month / 1-year / lifetime), unlimited users, runs inside your own infrastructure.

---

## What is Coomia?

Coomia (internally the "Master Data Stack" / MDS) is an enterprise data-intelligence platform that collapses what most organizations currently assemble from five or six vendors into a single privately deployed product. It covers three layers end-to-end:

1. **Pipelines** — ingestion, transformation, governance, and serving of data on Apache Flink + Apache Doris.
2. **Ontology** — modeling the business domain as structured, queryable, executable objects (Palantir Foundry-style).
3. **Analytics AI** — a super-team of specialized AI agents that read the ontology and deliver answers, dashboards, root-cause explanations, and forecasts.

The platform is **always** deployed inside the customer's own infrastructure — on any cloud (AWS, Azure, GCP, Alibaba Cloud) or on-premises Linux. No tenant data ever leaves the customer's perimeter. Users are **unlimited** within an organization; every license includes every capability — pricing varies by license term only (6-month, 1-year, or lifetime), never per seat, never by feature tier.

## Positioning & differentiation

**Coomia vs Palantir Foundry.** Coomia offers the same category of "ontology-driven decision intelligence" as Foundry — object types, actions, rules, world branching, decision studio — at roughly 1/10–1/20 the commercial cost, with full source/deployment control. Coomia is purpose-designed to be self-hosted by mid-market enterprises and vertical SaaS vendors.

**Coomia vs Databricks / Snowflake.** Coomia is not a warehouse. It runs on top of Apache Doris (and can overlay existing Doris, MySQL, PostgreSQL, ClickHouse). Databricks and Snowflake provide compute + storage; Coomia provides the pipeline authoring layer, the ontology semantic layer, and the AI analytics layer. Coomia complements a warehouse; it does not replace one.

**Coomia vs Airbyte + dbt + Dagster + BI.** The typical modern data stack stitches 4-6 vendors: ingestion (Airbyte/Fivetran), transformation (dbt), orchestration (Dagster/Airflow), catalog (DataHub/OpenMetadata), BI (Looker/Superset), observability (Monte Carlo). Coomia replaces the entire chain with one platform — same features delivered through AI-authored Flink SQL, real column-level lineage (from Flink jobs + Doris queries, not mocks), a first-class ontology layer, and an agent-driven analytics surface. Result: one vendor, one deployment, one contract.

**Coomia vs metric stores / semantic layers (Cube, MetricFlow).** Coomia's ontology is a superset: it has typed objects, relations, actions, events, rules, decisions, and world branching — not just metric definitions. AI agents use it for grounding; humans use it for what-if scenario planning.

## Three engines — in depth

### 1. Flink-native Pipelines

- **AI Pipeline Builder** — natural language → production Flink SQL + Apache Doris DDL. Generates medallion architecture (Bronze/Silver/Gold) automatically. Batch + streaming unified.
- **One-Click Database Onboarding** — point Coomia at a MySQL / PostgreSQL / Doris / ClickHouse database. It scans the schema, detects relationships, drafts ObjectType candidates, and deploys CDC pipelines in a single flow.
- **Native CDC Connectors** — MySQL binlog, PostgreSQL logical replication, Doris CDC, ClickHouse connector. 20+ additional sources (MongoDB, SQL Server, Oracle, Snowflake, BigQuery, Redshift, S3, GCS, Azure Blob, Kafka, …).
- **Data Quality Engine** — 30+ built-in rule types via Great Expectations. AI auto-generates rules from column profiles; statistical profiling; anomaly detection; PII identification; real-time quality scoring.
- **Real Data Lineage** — column-level lineage sourced from actual Flink job DAGs and Doris query plans, not mocked. Visual impact analysis shows what breaks before you ship.
- **Data Contracts + Schema Drift** — contract-first code generation (schema precedes code). Automatic drift detection, breaking-change alerts, AI-generated repair proposals.
- **Asset Catalog + Profiling** — URN-unified catalog across Flink, Doris, and the ontology layer. Column-level distribution histograms, null rates, PII flags.
- **Data Service Publishing** — any SELECT query can be converted to a REST API one-click, with dynamic parameters, version management, rate limiting, caching, and access control.
- **Pipeline Template Marketplace** — pre-built templates for ERP, CRM, e-commerce. Optimization engine auto-scans for performance bottlenecks.

### 2. Ontology Intelligence (16 modules)

The ontology is a Palantir Foundry-grade semantic layer. Object instances are stored as Doris VARIANT JSON with full version control. Changes automatically propagate downstream.

- **Object Type Management** — ObjectType schemas with attributes, relations, rules; version-controlled.
- **Object Explorer** — CRUD, semantic search, filtering, relationship graph, timeline.
- **Action Engine** — 5 pure data operations (CREATE/MODIFY/DELETE OBJECT or LINK), transactional.
- **Rule Engine** — YAML business rules + Rete network reasoning. TRIGGER_DECISION triggers; derived-attribute computation; event → rule → recommendation → execution closed loop.
- **Decision Studio** — interactive scenario exploration. Rule evaluation, option generation, impact assessment, AI-assisted recommendations with confidence scores.
- **World Management / What-if** — Nessie-powered branches for data. Before/After split-screen, impact propagation animation, multi-scenario comparison.
- **Event-Driven Architecture** — Event is a first-class ObjectType (kind=EVENT). Kafka-powered lifecycle transitions; full automation pipeline (Event → Rule → Recommendation → Confirmation → Execution).
- **Agent Runtime** — DecisionAgent, QueryAgent, OntologyAssistant; 4 agent capability types.
- **Vector Query + Graph Traversal** — HNSW vector embedding search for semantic similarity; multi-hop graph traversal.
- **6 Matching Functions** — vector, attribute, capability, team, risk, temporal matching.
- **Function Runtime** — user-defined Python functions in sandboxed execution; derived-attribute computation with isolation, versioning, monitoring.
- **OAG (Ontology-Augmented Generation)** — structured RAG grounded in the ontology. Prompt templates, decision-explanation generation, citation tracking.

### 3. Analytics AI Super-Team (5 agents)

- **Analytics Agent** — NL → SQL with ontology-aware semantic understanding. Aggregation, distribution, TopN; dynamic charts with confidence scoring.
- **Dashboard Agent** — AI-generated interactive dashboards. Intelligent chart-type selection. Object drill-through to the Object Explorer.
- **DQ Agent** — automated data-quality rule generation. Reports, anomaly alerts, trend monitoring.
- **RCA Agent** — root-cause analysis with anomaly-metric tracking, impact-chain visualization, causal analysis canvas, event backtracking, timeline replay.
- **Forecasting Agent** — time-series prediction with confidence-interval estimation and long-term trends.

Supporting capabilities:

- **6 Structured Renderers** — table, card, chart, metric, timeline, form. CopilotKit-powered streaming answers.
- **5 Analysis Models** — funnel, retention, path, interval, attribution (5 attribution models: first-touch, last-touch, linear, time-decay, position-based). Backed by Doris WINDOW_FUNNEL + RETENTION.
- **Pattern Discovery Engine** — vector clustering, graph patterns, time-series trend identification, cross-dimension correlation.
- **Ontology-Grounded Copilot** — page-context auto-injection, conversational action execution, entity auto-linking.

## Common enterprise capabilities

- **Project Management** — Org / Project hierarchy with Admin / Developer / Viewer roles.
- **Git-style Data Versioning** — Nessie branches. Zero-copy clones, atomic merges, Merge Request workflows with code + data diff.
- **RBAC** — world, project, and field level. SSO / SAML / OIDC. Data classification and masking.
- **Audit & Compliance** — immutable operation history, query audit logs, access tracking, compliance proof.
- **Multi-tenant Isolation** — three levels (Org → Project → Branch). Data, compute, access boundaries enforced at every layer.
- **Business Glossary** — terms, synonyms, asset association, review workflows.

## Pricing

One product — every license includes all three engines, unlimited users, and private deployment. Pricing varies by license term only:

| License term | USD | CNY | Best for |
| --- | --- | --- | --- |
| 6-Month | $11,999 | ¥79,999 | Lowest commitment; evaluation-to-production projects. Upgrade anytime by paying the difference. |
| 1-Year | $19,999/yr | ¥139,999/yr | Ongoing operation; saves ~17% vs two 6-month terms. Most popular. |
| Lifetime | $42,999 one-time | ¥299,999 one-time | Perpetual use; 3 years of updates + priority support included, then optional maintenance at 20% of the 1-Year price per year. |

Support by license term (all remote, no on-site):

| License term | Response SLA (workdays) | Channels | Extras |
| --- | --- | --- | --- |
| 6-Month | 24 hours | Email + Discord + GitHub + private Slack / WeChat channel | — |
| 1-Year | 24 hours | Email + Discord + GitHub + private Slack / WeChat channel | Quarterly remote health-check call + priority bug fixes |
| Lifetime | 8 hours (first 3 years) | Email + Discord + GitHub + private Slack / WeChat channel | Quarterly remote health-check call + priority bug fixes |

## Professional services (all remote)

Remote-only, no on-site engagements. Available as add-ons to any license:

| Service | Day rate (USD) | Day rate (CNY) |
| --- | --- | --- |
| Remote deployment assistance | $1,500 | ¥10,000 |
| Remote custom integration / migration | $2,000 | ¥14,000 |
| Remote training session | $2,000 | ¥14,000 |
| Consulting package (10-day minimum) | 15% off day rate | 日费 85 折 |

## Deployment

- **Runtime**: Docker Compose or Kubernetes (Helm charts provided).
- **OS**: Linux (Ubuntu 20.04+, CentOS 7+).
- **Minimum**: 4 CPU cores, 16 GB RAM, 100 GB storage. Scales horizontally per engine.
- **Cloud**: AWS, Azure, GCP, Alibaba Cloud, or on-prem hardware. No network egress of tenant data required.
- **Bring your own warehouse**: overlays existing Doris / MySQL / PostgreSQL / ClickHouse; augments rather than replaces.
- **Licensing**: license key bound to deployment instance. Transfer assistance available.

## Who should use Coomia?

- Data platform teams at mid-market to enterprise organizations tired of maintaining 5–6 open-source tools.
- Vertical SaaS vendors who want to embed a Palantir-grade ontology without Palantir-grade procurement.
- Data-regulated industries (finance, healthcare, manufacturing, government) that cannot send data to SaaS BI or SaaS pipeline vendors.
- Organizations building LLM-grounded decision tools that need a real ontology (not a vector store) backing their AI agents.

## Who is Coomia NOT for?

- Teams with <50 GB of data and no compliance constraints — a single data warehouse + BI tool is probably sufficient.
- Workloads that are 100% ad-hoc SQL with no governance, semantic, or AI-agent needs.

## Source-of-truth links

- Home: https://coomia.com/
- Features: https://coomia.com/features
- Pricing: https://coomia.com/pricing
- Support / FAQ: https://coomia.com/support
- Trial: https://coomia.com/trial
- Contact: https://coomia.com/contact
- Blog: https://coomia.com/blog
- Concise version of this file: https://coomia.com/llms.txt

## Localized sites

The website is fully localized in six languages (this file is English, the canonical language):

- English: https://coomia.com/ (default)
- 中文 (Chinese): https://coomia.com/zh/
- 日本語 (Japanese): https://coomia.com/ja/
- Deutsch (German): https://coomia.com/de/
- Bahasa Indonesia (Indonesian): https://coomia.com/id/
- Tiếng Việt (Vietnamese): https://coomia.com/vi/

## Canonical contact

- Website: https://coomia.com
- Support: support@coomia.com