OpenAI

Audience: Low-code Engineers

Skill Prerequisites: Actions, Tokens, Connectors

The OpenAI add-on adds actions that send text or audio to an AI model and save the result in a token. They chat with a model, summarize long text, turn a document into knowledge base articles, transcribe MP3 files and count AI model tokens. Despite the add-on's name, it also has chat actions for Amazon Bedrock (Aws AI Chat) and Google Vertex AI (VertexAI Chat). The OpenAI actions also work with Microsoft Azure OpenAI, except Speech to Text (Whisper).

note

This add-on is the PlantAnApp.OpenAi package. It's installed separately and needs the AI feature package in your license. If it isn't licensed, the actions fail with a not-licensed error. The AI connector types need the same license. If you don't see these actions, the add-on isn't installed.

Choosing an action

Action	What it does	Use it to
Chat	Sends a message to an OpenAI or Azure OpenAI chat model, with optional history, JSON replies and workflows as tools.	Answer users' questions, pull structured JSON out of free text, or let the model look something up with a workflow.
Aws AI Chat	Sends a message to a Claude, Llama or Titan model on Amazon Bedrock.	Use a model hosted in your own AWS account and region.
VertexAI Chat	Sends a message to a Gemini or older PaLM model on Google Vertex AI.	Use a Gemini model in your own Google Cloud project.
Summarize Content	Summarizes text with OpenAI or Azure OpenAI, cutting or splitting text that's too long for the model.	Summarize a support ticket, email thread or meeting transcript.
Create Articles	Splits a long text into topics and writes a knowledge base article for each, with OpenAI or Azure OpenAI.	Turn a product manual or training transcript into help articles.
Speech to Text (Whisper)	Transcribes an MP3 file with OpenAI's Whisper model, as text, JSON or SRT or VTT subtitles.	Transcribe a recorded call, or create subtitles for a video.
Count Tokens	Counts the AI model tokens a text uses for an OpenAI model. It runs on your server and sends nothing to OpenAI.	Check whether a text fits a model, or estimate the cost, before sending it.

Choosing a chat action

	Chat	Aws AI Chat	VertexAI Chat
Provider	OpenAI or Microsoft Azure OpenAI	Amazon Bedrock	Google Vertex AI
Temperature and maximum length	Yes	No, fixed per model	Yes
JSON or JSON Schema reply	Yes	No	No
Workflows as tools	Yes	Claude and Llama only	No
Streaming to a Conversation View	Optional	Optional	Required for Gemini, not possible for PaLM
Ignore Errors and On Error	Yes	No	Yes
PAA usage records	Yes	No	No

All three send text only, not images or files. To keep a conversation going, save Store Entire Conversation JSON in a token and pass it back as the history next time. With VertexAI Chat this only works for Gemini models.

Connectors

Every action except Count Tokens signs in through a connector. Create it first in the Connectors area or with Add Connector, and check it with Test Connector. See Connectors.

• OpenAI or Microsoft Azure OpenAI for Chat, Summarize Content and Create Articles. With Azure, you deploy the model in Azure first.
• OpenAI only for Speech to Text (Whisper). The Azure option doesn't work in 1.28.
• Amazon Bedrock AI for Aws AI Chat. The model must be enabled for Bedrock in the connector's region.
• Google VertexAI for VertexAI Chat.

Things to know before you send data

• Data leaves your server. Everything you send, including instructions, history, content, audio and workflow results, is processed by the provider under your account's terms. Don't send personal or confidential data unless your agreement with the provider allows it.
• Cost. The provider bills you for every token sent and received. Long histories, workflow calls, and long content split into many requests add up. Use Count Tokens to check the size of OpenAI requests first. It doesn't work for Bedrock or Vertex AI models.
• Replies can be wrong. Check important answers before you act on them, and don't use these actions for decisions that must always be correct, such as approvals or calculations.
• Long runs. Create Articles and Speech to Text (Whisper) can take a long time on long documents or recordings. Consider running them in a workflow instead of a form submit.
• Errors. Aws AI Chat and Speech to Text (Whisper) have no On Error or Ignore Errors settings. To handle their errors, put them inside Execute Actions and use its On Error actions.

For storing connector details, see the Connectors actions. To read values from a JSON reply, see the Parsing actions.

Revised 10/02/2026