Skip to content
Documentation
Contact
← Back to documentation
Technical Reference

OpenAI

Use OpenAI models for generation, transcription, embeddings and structured workflow steps.

CategoryAI
Authenticationcustom
Version0.11.0
Published scopes0 published scopes
Last updatedSeptember 21, 2026

Overview

Connect OpenAI with Flows360

Use OpenAI models for generation, transcription, embeddings and structured workflow steps.

Use the OpenAI connector to include its declared actions and events in Flows360 workflows without duplicating the integration logic inside each consuming system. This pinned connector version currently exposes 19 actions and no declared triggers.

Available actions

  • Ask ChatGPT
  • Ask Assistant
  • Vision Prompt
  • Extract Structured Data from Text
  • Classify Text (Moderation)
  • Analyze Text Sentiment
  • Create Embedding
  • Search Embeddings
  • Plus 11 additional declared actions.

Available triggers

  • No triggers are declared in this pinned package.

The capability list above is generated from the exact reviewed connector package so the public description stays aligned with the version Flows360 operates.

Technical guide

Technical reference

Package identity

  • Connector: OpenAI
  • Package: @activepieces/piece-openai@0.11.0
  • Exact artifact digest: sha256:3617dc454f66d368969a75fb56c058946983defa8c19a5a6f29b3c8609e78389
  • Runtime: Activepieces-compatible deterministic worker
  • Source: the exact reviewed Activepieces source

Authentication

SECRET TEXT. Connection values are tenant-specific and remain in the consuming Studio instance; CDK stores definitions and field metadata, not customer credential values.

Action contracts

  • Ask ChatGPT (ask_chatgpt), Ask ChatGPT anything you want! Classification: read.
  • Ask Assistant (ask_assistant), Ask a GPT assistant anything you want! Classification: read.
  • Vision Prompt (vision_prompt), Ask GPT a question about an image Classification: read.
  • Extract Structured Data from Text (extract-structured-data), Returns structured data from provided unstructured text. Classification: read.
  • Classify Text (Moderation) (classify_text), Classify whether the supplied text violates OpenAI safety policies (harassment, hate, self-harm, sexual, violence, etc.). Classification: read.
  • Analyze Text Sentiment (analyze_sentiment), Analyzes text for sentiment (positive, negative, or neutral). Classification: read.
  • Create Embedding (create_embedding), Generate a vector embedding for the supplied text. Useful for semantic search, clustering and RAG pipelines. Classification: read.
  • Search Embeddings (search_embeddings), Matches a query string to a list of document strings for best results. Classification: read.
  • Generate Image (generate_image), Generate an image using text-to-image models Classification: read.
  • Edit Image (edit_image), Edit an existing image using a text prompt with gpt-image-2 Classification: read.
  • Text-to-Speech (text_to_speech), Generate an audio recording from text Classification: read.
  • Transcribe Audio (transcribe), Transcribe audio to text using whisper-1 model Classification: read.
  • Translate Audio (translate), Translate audio to text using whisper-1 model Classification: read.
  • Upload File (upload_file), Upload a file to OpenAI for use with Assistants, Vector Stores, Batch jobs, Fine-tuning, or Vision. Classification: read.
  • List Files (list_files), Return the list of files uploaded to the connected OpenAI account, optionally filtered by purpose. Classification: read.
  • Find File (find_file), Check whether a file with the given name already exists in the connected OpenAI account. Classification: read.
  • Delete File (delete_file), Delete a file previously uploaded to the connected OpenAI account. Classification: read.
  • List Models (list_models), Return the list of models available to the connected OpenAI account, optionally filtered by an ID substring. Classification: read.
  • Custom API Call (custom_api_call), Make a custom API call to a specific endpoint Classification: read.

Trigger contracts

  • No trigger delivery contracts are declared.

Verification

Deterministic execution evidence is recorded for the pinned artifact, including 1 exercised action entrypoint and 0 exercised trigger entrypoints.

Capability summary

Authenticationcustom
Supported objects19
Supported actions19
TriggersNone published

Authentication

Authentication method: custom

Authentication profiles

connection · custom
  • API Key Required · protected credential

Supported objects

ObjectReadWriteNotes
Ask ChatGPT Yes No Ask ChatGPT anything you want!
Ask Assistant Yes No Ask a GPT assistant anything you want!
Vision Prompt Yes No Ask GPT a question about an image
Extract Structured Data from Text Yes No Returns structured data from provided unstructured text.
Classify Text (Moderation) Yes No Classify whether the supplied text violates OpenAI safety policies (harassment, hate, self-harm, sexual, violence, etc.).
Analyze Text Sentiment Yes No Analyzes text for sentiment (positive, negative, or neutral).
Create Embedding Yes No Generate a vector embedding for the supplied text. Useful for semantic search, clustering and RAG pipelines.
Search Embeddings Yes No Matches a query string to a list of document strings for best results.
Generate Image Yes No Generate an image using text-to-image models
Edit Image Yes No Edit an existing image using a text prompt with gpt-image-2
Text-to-Speech Yes No Generate an audio recording from text
Transcribe Audio Yes No Transcribe audio to text using whisper-1 model
Translate Audio Yes No Translate audio to text using whisper-1 model
Upload File Yes No Upload a file to OpenAI for use with Assistants, Vector Stores, Batch jobs, Fine-tuning, or Vision.
List Files Yes No Return the list of files uploaded to the connected OpenAI account, optionally filtered by purpose.
Find File Yes No Check whether a file with the given name already exists in the connected OpenAI account.
Delete File Yes No Delete a file previously uploaded to the connected OpenAI account.
List Models Yes No Return the list of models available to the connected OpenAI account, optionally filtered by an ID substring.
Custom API Call Yes No Make a custom API call to a specific endpoint

Supported actions

Ask ChatGPT · Read

Ask ChatGPT anything you want!

ask_chatgpt · Risk: low · Retry: safe

Inputs

  • model · string · Required, The model which will generate the completion. Some models are suitable for natural language tasks, others specialize in code.
  • prompt · string · Required
  • temperature · number · Optional, Controls randomness: Lowering results in less random completions. As the temperature approaches zero, the model will become deterministic and repetitive.
  • maxTokens · number · Required, The maximum number of tokens to generate. Requests can use up to 2,048 or 4,096 tokens shared between prompt and completion depending on the model. Don't set the value to maximum and leave some tokens for the input. (One token is roughly 4 characters for normal English text)
  • topP · number · Optional, An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.
  • frequencyPenalty · number · Optional, Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim.
  • presencePenalty · number · Optional, Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the mode's likelihood to talk about new topics.
  • memoryKey · string · Optional, A memory key that will keep the chat history shared across runs and flows. Keep it empty to leave ChatGPT without memory of previous messages.
  • roles · object · Optional, Array of roles to specify more accurate response
Ask Assistant · Read

Ask a GPT assistant anything you want!

ask_assistant · Risk: low · Retry: safe

Inputs

  • assistant · string · Required, The assistant which will generate the completion.
  • prompt · string · Required
  • memoryKey · string · Optional, A memory key that will keep the chat history shared across runs and flows. Keep it empty to leave your assistant without memory of previous messages.
Vision Prompt · Read

Ask GPT a question about an image

vision_prompt · Risk: low · Retry: safe

Inputs

  • image · object · Required, The image URL or file you want GPT's vision to read.
  • prompt · string · Required, What do you want ChatGPT to tell you about the image?
  • detail · string · Optional, Control how the model processes the image and generates textual understanding.
  • temperature · number · Optional, Controls randomness: Lowering results in less random completions. As the temperature approaches zero, the model will become deterministic and repetitive.
  • maxTokens · number · Optional, The maximum number of tokens to generate. Requests can use up to 2,048 or 4,096 tokens shared between prompt and completion, don't set the value to maximum and leave some tokens for the input. The exact limit varies by model. (One token is roughly 4 characters for normal English text)
  • topP · number · Optional, An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.
  • frequencyPenalty · number · Optional, Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim.
  • presencePenalty · number · Optional, Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the mode's likelihood to talk about new topics.
  • roles · object · Optional, Array of roles to specify more accurate response
Extract Structured Data from Text · Read

Returns structured data from provided unstructured text.

extract_structured_data · Risk: low · Retry: safe

Inputs

  • model · string · Required
  • text · string · Required
  • params · array · Required
Classify Text (Moderation) · Read

Classify whether the supplied text violates OpenAI safety policies (harassment, hate, self-harm, sexual, violence, etc.).

classify_text · Risk: low · Retry: safe

Inputs

  • model · string · Required, Moderation model to use. omni-moderation-latest also supports images.
  • input · string · Required, The text to classify.
Analyze Text Sentiment · Read

Analyzes text for sentiment (positive, negative, or neutral).

analyze_sentiment · Risk: low · Retry: safe

Inputs

  • model · string · Required, The chat model used to judge sentiment.
  • text · string · Required, The text to analyze.
Create Embedding · Read

Generate a vector embedding for the supplied text. Useful for semantic search, clustering and RAG pipelines.

create_embedding · Risk: low · Retry: safe

Inputs

  • model · string · Required, The embedding model to use.
  • input · string · Required, The text to embed. To embed multiple strings in one call, provide a newline-separated list or use the Custom API Call action.
  • dimensions · number · Optional, Number of dimensions for the embedding (only supported by text-embedding-3-* models).
  • user · string · Optional, A stable identifier for the end user u2014 helps OpenAI detect abuse.
Search Embeddings · Read

Matches a query string to a list of document strings for best results.

search_embeddings · Risk: low · Retry: safe

Inputs

  • model · string · Required, The embedding model to use for both query and documents.
  • query · string · Required, The text to match against the documents.
  • documents · array · Required, The list of document strings to search.
  • topK · number · Optional, Return only the top K best matches. Leave empty to return all documents ranked by score.
Generate Image · Read

Generate an image using text-to-image models

generate_image · Risk: low · Retry: safe

Inputs

  • model · string · Required, The model which will generate the image.
  • prompt · string · Required
  • resolution · string · Optional, The resolution to generate the image in.
  • quality · string · Optional, Image quality level.
Edit Image · Read

Edit an existing image using a text prompt with gpt-image-2

edit_image · Risk: low · Retry: safe

Inputs

  • image · object · Required, The image to edit (PNG, JPEG, or WebP).
  • prompt · string · Required, A description of the desired edits or the new image.
  • mask · object · Optional, Optional mask image. Transparent areas indicate where to apply edits. Must match the input image dimensions.
  • size · string · Optional, The size of the generated image.
  • quality · string · Optional, Image quality level.
Text-to-Speech · Read

Generate an audio recording from text

text_to_speech · Risk: low · Retry: safe

Inputs

  • text · string · Required, The text you want to hear.
  • model · string · Required, The model which will generate the audio.
  • speed · number · Optional, The speed of the audio. Minimum is 0.25 and maximum is 4.00.
  • voice · string · Required, The voice to generate the audio in.
  • format · string · Required, The format you want the audio file in.
  • fileName · string · Optional, The name of the output audio file (without extension).
Transcribe Audio · Read

Transcribe audio to text using whisper-1 model

transcribe · Risk: low · Retry: safe

Inputs

  • audio · object · Required, Audio file to transcribe
  • language · string · Optional, Language of the audio file the default is en (English).
Translate Audio · Read

Translate audio to text using whisper-1 model

translate · Risk: low · Retry: safe

Inputs

  • audio · object · Required, Audio file to translate
Upload File · Read

Upload a file to OpenAI for use with Assistants, Vector Stores, Batch jobs, Fine-tuning, or Vision.

upload_file · Risk: low · Retry: safe

Inputs

  • file · object · Required, The file to upload.
  • purpose · string · Required, The intended use of the file. Each purpose has different allowed file types u2014 see https://platform.openai.com/docs/api-reference/files/create.
  • fileName · string · Optional, Optional. By default the original filename is used. Provide a value to override (must include the extension).
List Files · Read

Return the list of files uploaded to the connected OpenAI account, optionally filtered by purpose.

list_files · Risk: low · Retry: safe

Inputs

  • purpose · string · Optional, If supplied, only files with this purpose are returned.
  • limit · number · Optional, Maximum number of files to return (1-10000). Defaults to 10000.
Find File · Read

Check whether a file with the given name already exists in the connected OpenAI account.

find_file · Risk: low · Retry: safe

Inputs

  • fileName · string · Required, The exact filename to look for (case-insensitive).
  • purpose · string · Optional, If supplied, only files with this purpose are searched.
Delete File · Read

Delete a file previously uploaded to the connected OpenAI account.

delete_file · Risk: low · Retry: safe

Inputs

  • fileId · string · Required, The OpenAI file ID returned by Upload File (starts with `file-`).
List Models · Read

Return the list of models available to the connected OpenAI account, optionally filtered by an ID substring.

list_models · Risk: low · Retry: safe

Inputs

  • contains · string · Optional, Optional case-insensitive substring used to filter model IDs (e.g. `gpt-4`, `embedding`, `whisper`).
Custom API Call · Read

Make a custom API call to a specific endpoint

custom_api_call · Risk: low · Retry: safe

Inputs

  • url · object · Required
  • method · string · Required
  • headers · object · Optional, Authorization headers are injected automatically from your connection.
  • queryParams · object · Optional, Appended to the URL as ?key=value.
  • body_type · string · Optional
  • body · object · Optional
  • response_is_binary · boolean · Optional, Enable for files like PDFs, images, etc.
  • failsafe · boolean · Optional, On a failed request, output the error instead of failing the step.
  • timeout · number · Optional, Seconds to wait for a response. Empty: up to the flow limit (10 min).
  • followRedirects · boolean · Optional, Follow 3xx redirects instead of returning them as the response.

Setup

Setup

  1. Add the OpenAI connector to the workflow in the target Flows360 Studio instance.
  2. Create or select the tenant-specific connection required by the connector. Connection secrets remain in the Studio connection boundary and are not copied into CDK documentation.
  3. Provide the connection fields declared by the pinned package:
  • No provider credential fields are declared by the pinned package.
  1. Select the required action or trigger and complete its declared input fields. Required and optional inputs are defined by the exact package schema.
  2. Test the workflow in the appropriate environment before enabling production scheduling or event delivery.

Limitations

Limitations and operational notes

  • This content describes @activepieces/piece-openai@0.11.0; provider behaviour can change independently and should be revalidated when the provider or connector version changes.
  • Only the 19 actions and 0 triggers declared by this pinned package are represented here.
  • Tenant credentials and connection values are not stored in the public connector record.
  • Provider-side permissions, account entitlements, quotas and rate limits remain subject to the connected provider account and are not inferred when the package does not declare them.
  • Deterministic execution evidence is recorded for the pinned artifact, including 1 exercised action entrypoint and 0 exercised trigger entrypoints.

Troubleshooting

Why canu2019t the connector authenticate?

Check the tenant connection in Studio and confirm every required OpenAI connection field is present. Re-authorise OAuth-based connections if the provider token or consent has expired. Do not place credential values in CDK content or logs.

Why is an action or trigger unavailable?

Confirm that the workflow is using @activepieces/piece-openai@0.11.0 and compare the requested capability with the declared action and trigger list for this version. A capability that is not declared by the pinned package should not be presented as supported.

What should I check after a provider-side change?

Revalidate authentication, required fields, action/trigger behaviour and provider documentation before publishing refreshed connector content or moving a new package version into production.

github.com