Send Prompt
Send a prompt to GPT, optionally along with an image, and receive a text message in response.
OpenAI ChatGPT Action
This integration allows you to send a prompt to GPT, optionally along with an image, and receive a text response. You can customize various parameters, such as rules, filters, system and user messages, attached image URL, model, temperature, and maximum number of tokens. Additionally, you can configure the removal of quotation marks in ChatGPT's response and customize tags and output variables.
When executing this action, you will receive information such as the generated text, the reason for completion, the number of tokens used in the prompt and the response, and the total number of tokens used. This integration facilitates the automation of tasks that require AI-generated text, optimizing processes and improving efficiency in various applications.
Customization Options
Configurable fields you can adjust in your automation
- Rules
- Filters (Only present if Rules switch is ON)
- System message
- New user message
- Attached image URL
- Model
- Temperature
- Max tokens to use
- Remove any potential quotation marks from the beginning and end of ChatGPT's response
- Customize tags
- Output Variable Renaming (Only present if Customize tags switch is ON)
Information provided
When executed, this operation delivers the following data, which can be used in the same automatic task.
Tags
-
Text
{{content}}
Predicted completion.
e.g. This is a test -
Finish reason
{{finish_reason}}
reason why the resulting text has terminated.
e.g. lenght -
Prompt tokens
{{prompt_tokens}}
Number of tokens in the Prompt. 1 token ~= 4 chars in English
-
Completion tokens
{{completion_tokens}}
Number of tokens in the Completion. 1 token ~= 4 chars in English
-
Total tokens
{{total_tokens}}
Prompt tokens + Completion tokens
-
Tool name
{{tool_name}}
Name of the function the model wants to call. Only present when json_tools is used and the model returns a tool call instead of text.
-
Tool payload
{{tool_payload}}
JSON string with the arguments for the tool call. Pass the result back using the previous_tool_result field.
Frequently Asked Questions
Why does ChatGPT respond differently to seemingly identical prompts?
Variations are almost always caused by one (or several) of these factors:
• High temperature: at higher values (≈ 0.8-1.0) the model injects more randomness. Lower it to 0.2-0.4 if you need consistency.
• Missing examples or clear rules: without few-shot examples or precise instructions, the model fills in gaps on its own. Add concrete input-output samples or define rules in the system prompt/instruction.
• Ambiguous prompt: if your text supports multiple interpretations, the model will choose one at random. Rephrase so there is only one possible reading.
• Context shifts: any previous message in the conversation influences the answer. Restart the thread or explicitly summarise the context if you need reproducibility.
• Token limits: when the conversation grows long, parts of the history may be truncated and the model "forgets" details. Shorten the prompt or summarise the context.
Follow these recommendations to obtain much more stable and predictable responses.
Common Errors
Invalid JSON response calling chatgpt: 210-Error You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.
The error indicates that you have exceeded the usage quota assigned to your account on the OpenAI API, preventing further requests. To resolve this, review the details of your plan and current usage on the OpenAI platform and consider upgrading to a plan with a higher quota if your usage requires it at https://platform.openai.com.
Let's talk
Choose day and time.
We share the screen and answer all your questions.