---
orphan: false
---

# Genesis

## Building on the preceding platform: where models come from and how they are provided externally

Agent conversations, knowledge base embeddings, and workflow model nodes in [AI Studio](ai-studio.md) all need to call models, while [MatrixOne](matrixone.md) handles only data storage and computation, not inference. So where do these models come from, how is their usage measured, and how are they made available to your own applications? The answer is **Genesis**, MOI's platform for managing model supply.

Its internal code name is **TaaS** (Token as a Service). Bills and logs use the name Genesis; both names refer to the same product.

## Positioning

**Genesis is MOI's standardized model API service for external users.** It brings models from multiple vendors and with multiple capabilities together behind a unified entry point, a unified API specification (OpenAI compatible), and a unified Credit billing system. With a single API Key, callers can access, measure, audit, and operate model calls.

Genesis has an **independent entry point** in the product's top navigation and uses a separate sign-in flow and console. It has no workspace switcher and does not distinguish between data and agent modes.

## Core concept: the API Key controls both access and cost

In Genesis, an API Key determines more than simply whether a caller has access. It controls **which models can be called, how much can be consumed, how quickly calls can be made, and how long access lasts**. These four groups of controls are central to understanding Genesis:

| Control | What it controls |
| --- | --- |
| Available models | Whether the Key can call all models or only specified models |
| Quota limits | Total and daily quota limits, measured in Credits |
| Rate limits | Concurrency, RPM (requests per minute), and TPM (tokens per minute) |
| Expiration | Never expires, or becomes invalid automatically at the specified time |

The primary Key issuance path is a **service account** in the account system (UC), which issues keys in the `sk-gsa-v1-...` format. The complete key is displayed only once in the creation dialog. For details, see [Guide · Manage API Keys](../guides/genesis/api-keys.md).

## Console capabilities

The console is organized around the flow "select a model → debug → obtain a Key → call → operate":

| Area | Purpose |
| --- | --- |
| Models | Model Marketplace: filter by type, vendor, context length, and health; view capabilities, pricing, and examples |
| Debug | Playground: try different types of models online, tune parameters, and write prompts |
| Use | Integration guide: Base URL, authentication, request parameters and response examples, and FAQ |
| Keys | Create, edit, rotate, enable, disable, and delete API Keys, and configure the four groups of controls above |
| Dashboard | Operational overview of QPS, availability, health, today's requests and cost, and more |
| Usage | Analyze usage and cost by time, model, and Key |
| Logs | Per-call logs for troubleshooting and cost attribution |
| Notifications | Alerts for anomalies in usage, failures, latency, security, and more |

Models are organized by capability, including chat, OCR, text-to-image / video, TTS, ASR, embedding, and reranking. The actual set depends on the current deployment.

## The difference between "selling models" and AI Studio's "selling capabilities" (separate ledgers)

This distinction is central to understanding Genesis billing and is also the most common source of confusion between Genesis and AI Studio. Which ledger records a model inference depends **only on its entry point**:

| | AI services (inside AI Studio) | Genesis |
| --- | --- | --- |
| What is sold | AI capabilities packaged by the platform | External model services |
| Whether the model is exposed | Hidden; the platform can switch the underlying model | **Exposed; the caller chooses the model** |
| Who initiates the call | A user works with an agent, workflow, or knowledge base in Studio | A caller uses an API Key outside the platform or in their own application |
| Which ledger records it | AI services (not included in usage shown in the Genesis console) | Genesis Credits |

In one sentence: **intelligence used through the interface → AI services; models called with your own API Key → Genesis.** BYOM (bring your own model) is a separate billable item. For the complete boundaries among these three, see [How the three platforms work together · Separate ledgers](platforms-together.md#separate-ledgers) and [Credits and billable items](../guides/billing/credit-and-pricing.md).

## Returning to the three platforms

The three gateways now form a complete path: [MatrixOne](matrixone.md) stores data correctly, governs it, and makes it queryable; [AI Studio](ai-studio.md) turns that foundation into AI-ready data and puts it to work with agents; Genesis turns model supply into an operable, measurable service that can be provided externally. All three share one account system and one commitment. For details about how they fit together, see [How the three platforms work together](platforms-together.md).