OpenAI actions: generation, transcription, files.
Analyzes one or several images with a vision-capable model (Pixtral-12b) in the OpenAI integration. Inputs: in, type, multipleImages, streaming. Outputs: error, result (the analysis).
Deletes a vector store from OpenAI. Inputs: in, vectorStoreId (the store to remove). Outputs: out, on error, error, success.
Deletes a file from OpenAI. Inputs: in, fileId (the file to remove). Outputs: out, on error, error, fileId, deleted (the deletion status).
Lists the files available on OpenAI. Inputs: in, purpose (to filter the files). Outputs: out, on error, error, files (the list of files).
Generates embeddings from plain text, as listed in the OpenAI integration. Inputs: in, model, text. Outputs: out, on error, error, result (the embeddings).
Performs a web search with OpenAI's web search tool. Inputs: in, model, prompt. Outputs: out, on error, error, result, annotations (the annotations attached to the result).
file search (OpenAI) performs a file search in a given OpenAI vector store from a prompt. Inputs: in, model, prompt, vectorStoreId. Outputs: out, on error, error, result, annotations. It works with a vector store prepared beforehand in OpenAI.
create vector store (OpenAI) creates a vector store in OpenAI, optionally from a list of files. Inputs: in, name, files. Outputs: out, on error, error, vectorStore. The store can then be used by the file search (OpenAI) block.
upload file (OpenAI) uploads a file to OpenAI. Inputs: in, type, purpose. Outputs: out, on error, error, fileId, fileName. The returned fileId can then be reused by other OpenAI blocks, for example to add the file to a vector store.
generate speech (OpenAI) generates speech with the OpenAI text to speech API from an input text. Inputs: in, voice, model, streaming, input, speed. Outputs: out, on error, error, result. Voice, model and speed shape the audio produced.
speech to text (OpenAI) generates a transcription from an audio file using the OpenAI Whisper model. Inputs: in, type. Outputs: out, on error, error, result (the transcription). It is the reverse of the generate speech block.
list vector stores (OpenAI) lists all the vector stores available in OpenAI. Input: in. Outputs: out, on error, error, vectorStores. Useful to find the identifier of a store before using it with other OpenAI blocks.
generate image (OpenAI) generates an image using Dall-E 2, Dall-E 3 or GPT Image 1. Inputs: in, size, model, prompt. Outputs: out, on error, error, url (the address of the generated image).
Edit Image (OpenAI) edits an image from its URL using the OpenAI API. Settings: size, prompt, URL, quality. Calls can take 30 seconds to over a minute, and the returned URL is large, so the WriteToLog block is recommended to view it.
