# inference-api

## Pages

- [Inference API](https://docs.ezghcloud.com/inference-api/index.md) — Hosted model inference behind an OpenAI-compatible API. The OpenAI surface is under a project base, https://inference-api.ezghcloud.com/v1/organizations/{orgId}/projects/{projectId}/openai/v1: the stock OpenAI SDKs work with only base_url and api_key set. Models are the platform's Model Catalog; a request names one as <id> (its latest version) or <id>:<version>, and every response carries the resolved <id>:<version>.
- [Create embeddings (OpenAI-compatible)](https://docs.ezghcloud.com/inference-api/CreateEmbeddings/index.md) — OpenAI's embeddings API. input is a string or an array of strings (token-ID arrays are refused with token_input_unsupported); dimensions truncates the vectors (Matryoshka), within the model's minDimensions and dimensions; encoding_format is float (the default) or base64. user is accepted and ignored. Errors use OpenAI's body, with the platform's stable code in error.code.
- [Get the project's inference limits](https://docs.ezghcloud.com/inference-api/GetLimits/index.md) — The rate, concurrency and monthly limits in force for the project. The per-request limits are each model's properties in the catalog.
- [Get a model (OpenAI-compatible)](https://docs.ezghcloud.com/inference-api/GetModel/index.md)
- [List the models the project can call (OpenAI-compatible)](https://docs.ezghcloud.com/inference-api/ListModels/index.md) — Every model in the catalog offered by Inference with a version that isn't retired.
- [Embedding](https://docs.ezghcloud.com/inference-api/schemas/Embedding/index.md)
- [EmbeddingProperties](https://docs.ezghcloud.com/inference-api/schemas/EmbeddingProperties/index.md)
- [EmbeddingsRequest](https://docs.ezghcloud.com/inference-api/schemas/EmbeddingsRequest/index.md)
- [EmbeddingsResponse](https://docs.ezghcloud.com/inference-api/schemas/EmbeddingsResponse/index.md)
- [Error](https://docs.ezghcloud.com/inference-api/schemas/Error/index.md) — Every non-2xx answer. Clients branch on code, never on message.
- [Limits](https://docs.ezghcloud.com/inference-api/schemas/Limits/index.md)
- [OpenAIError](https://docs.ezghcloud.com/inference-api/schemas/OpenAIError/index.md) — Every non-2xx answer of an OpenAI-compatible endpoint: the OpenAI API's error body, with the platform's code. OpenAI clients read type; branch on type and code, never on message.
- [OpenAIModel](https://docs.ezghcloud.com/inference-api/schemas/OpenAIModel/index.md)
- [OpenAIModelList](https://docs.ezghcloud.com/inference-api/schemas/OpenAIModelList/index.md)
- [QuotaExceeded](https://docs.ezghcloud.com/inference-api/schemas/QuotaExceeded/index.md) — quota_exceeded: the quota that refused the request. Never with Retry-After.