OpenAI Extended Endpoints

Map of OpenAI-compatible pass-through endpoints beyond chat, including embeddings, images, audio, files, batches, assistants, threads, and videos.

These routes use the OpenAI-compatible prefix but are not the recommended first integration path. Use them when your client already expects these OpenAI API shapes or when a specific workflow requires them.

Status

Pass-through. Request and response details generally follow the upstream OpenAI-compatible or LiteLLM-compatible behavior for the selected model/provider. Validate each route with your target model before production use.

Base URL and auth

https://gateway.api.anyint.ai/openai/v1
Authorization: Bearer <ANYINT_API_KEY>
Content-Type: application/json

Routes

Route	Typical use
`POST /embeddings`	Generate vector embeddings
`POST /images/generations`	Generate images; Kling model IDs can route through Transtreams adaptation
`POST /audio/speech`	Text-to-speech
`POST /audio/transcriptions`	Speech-to-text transcription
`POST /audio/translations`	Audio translation
`POST /moderations`	Moderation checks
`/files`, `/files/{file_id}`, `/files/{file_id}/content`	OpenAI-compatible file operations
`/assistants`, `/assistants/{assistant_id}`	Assistants API pass-through when enabled upstream
`/threads`, `/threads/{thread_id}`, `/threads/{thread_id}/messages`, `/threads/{thread_id}/runs`	Threads and runs pass-through when enabled upstream
`/batches`, `/batches/{batch_id}`, `/batches/{batch_id}/cancel`	Batch job pass-through when enabled upstream
`/videos`, `/videos/{video_id}`, `/videos/{video_id}/content`	OpenAI-style video generation and lookup

Embeddings example

Embeddings request fields

Field	Type	Required	Meaning	Example
`model`	string	Yes	Embedding model ID available to your account.	`text-embedding-v4`
`input`	string or array	Yes	Text or list of texts to embed.	`AnyInt routes one application...`
`encoding_format`	string	No	Vector encoding format when supported.	`float`
`dimensions`	integer	No	Requested vector dimensions when supported by the model.	`1024`
`user`	string	No	Customer-side end-user identifier for tracing or abuse monitoring.	`user_123`

curl https://gateway.api.anyint.ai/openai/v1/embeddings \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-v4",
    "input": "AnyInt routes one application through multiple model families."
  }'

Embeddings response fields

Field	Meaning
`data[].embedding`	Embedding vector
`data[].index`	Input index for batched requests
`model`	Model used for embedding
`usage.prompt_tokens`	Input tokens when returned
`usage.total_tokens`	Total tokens when returned

Image generation example

Image generation request fields

Field	Type	Required	Meaning	Example
`model`	string	Yes	Image-capable model ID. For full provider-native controls, use Media APIs.	`kling-v3-omni`
`prompt`	string	Yes	Text prompt describing the desired image.	`A cinematic product poster...`
`size`	string	No	Output size when supported by the selected path.	`1024x1024`
`n`	integer	No	Number of images requested when supported.	`1`
`response_format`	string	No	Response format when supported.	`url`
`user`	string	No	Customer-side end-user identifier.	`user_123`

curl https://gateway.api.anyint.ai/openai/v1/images/generations \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-v3-omni",
    "prompt": "A cinematic product poster for an AI music launch."
  }'

For provider-native image examples, see Image Generation and Media APIs.

Audio request fields

Route	Important fields	Notes
`POST /audio/speech`	`model`, `input`, `voice`, `response_format`, `speed`	Text-to-speech. Supported voices and formats depend on the selected model.
`POST /audio/transcriptions`	`file`, `model`, `language`, `prompt`, `response_format`, `temperature`	Use multipart upload when required by your client.
`POST /audio/translations`	`file`, `model`, `prompt`, `response_format`, `temperature`	Translates audio to English or the provider-supported target behavior.

File, assistant, thread, and batch fields

These endpoints are pass-through compatibility routes. Use them when your client already expects the OpenAI-compatible shape and verify support with your selected model path.

Area	Common fields	Customer guidance
Files	`file`, `purpose`, `file_id`	Use only for clients that require OpenAI-compatible file objects.
Assistants	`model`, `name`, `instructions`, `tools`, `file_ids`	Prefer Chat Completions or Responses for new simple apps unless your client requires Assistants.
Threads	`messages`, `metadata`, `thread_id`	Useful for OpenAI-compatible clients with built-in thread handling.
Runs	`assistant_id`, `model`, `instructions`, `tools`, `stream`	Verify tool and streaming behavior per model.
Batches	`input_file_id`, `endpoint`, `completion_window`, `metadata`	Use for clients that already generate compatible batch files.

Video route behavior

OpenAI-style video requests can be accepted at /openai/v1/videos. Kling model IDs such as kling-v3, kling-v3-std, kling-v3-pro, and kling-v3-omni are adapted to Transtreams routes where applicable.

If you want full control over Kling request fields, use Kling video directly instead of the OpenAI-style video route.

Common mistakes

Treating every OpenAI upstream feature as enabled for every AnyInt model
Sending provider-native media payloads to OpenAI-compatible endpoints without checking the route page
Assuming Assistants, Threads, Batches, or Files are the right abstraction for new apps when Chat Completions or Responses is enough

OpenAI Extended Endpoints

On this page