Python Additional Invoke OpenAI

Additional: Invoke and OpenAI

OpenAI-compatible endpoints

Endpoint	Назначение
GET /v1/models	список моделей
POST /v1/chat/completions	chat completions
POST /v1/chat/completions with stream=True	streaming chat completions
POST /v1/responses	Responses-like text output
POST /v1/embeddings	embeddings

Direct invoke endpoints

Используйте runtime.invoke(endpoint, payload) или bridge.invoke(endpoint, payload).

Endpoint	Unit	Required input	Return shape
chat.completions, chat	LLM	messages or prompt	{content, raw?}
responses	LLM	input	response-like text content
embeddings, embedding	embedding	input	{embeddings}
rerank, reranker	reranker	query, documents	{results} sorted by score
translate, translator	translator	text/input, optional languages	translation output
tts	tts	text/input	audio-like result
image.classify, image-classification	image-classification	image/input/url	{labels, raw}
image.detect, object-detection	object-detection	image/input/url	{boxes, raw}
image.segment, image-segmentation	image-segmentation	image/input/url	{segments, raw}
depth, depth-estimation	depth-estimation	image/input/url	{depth, raw}
image-to-text, vlm	vlm	image/input/url	{text, raw}
asr, speech-to-text	asr	audio/input/url	{text, chunks?, raw}
zero-shot-image	zero-shot-image	image, labels	{labels, raw}
language-id	language-id	audio/input/url	{labels, raw}
audio.classify, audio-classification	audio-classification	audio/input/url	{labels, raw}
ocr	ocr	image/input/url	{text, raw}
document-layout	document-layout	image/input/url	{boxes, raw}
table-detection	table-detection	image/input/url	{boxes, raw}
document-qa	document-qa	image, question/query	{answers, raw}
text.classify, text-classification	text-classification	text/input	{labels, raw}
ner, token-classification	ner	text/input	{entities, raw}
zero-shot-text, zero-shot-classification	zero-shot-text	text, labels	{labels, raw}
summarization, summarize	summarization	text/input	{summary, raw}
text2text, text2text-generation	text2text	text/input	{text, raw}
code, code.embed	code	text/input	{features, raw}
vector.add/search/delete/clear/stats	vectorstorage	unit + operation payload	storage result
rag.add/search/delete/clear/reindex/stats	RAG	unit + operation payload	RAG result
onnx.predict	onnx/custom	inputs	custom ONNX output

Convenience wrappers

Wrapper	Описание
runtime.chat(...)	shortcut для chat.completions
runtime.embed(...)	shortcut для embeddings
runtime.client(...)	OpenAI client, требует optional openai package
unit.invoke(...)	single-unit runtime invoke