OpenAI LLM

Sends a prompt to an OpenAI-compatible large language model via Salt’s LLM service and returns the generated text. Supports a system prompt for instruction, a user prompt with dynamic placeholders, temperature, and max token controls. Automatically lists available OpenAI models from the service and falls back to a built-in set if the service is unavailable.

Usage

Use this node to generate text, analyze content, or perform instruction-following tasks with OpenAI models. Provide a clear prompt, optionally combine it with context inputs (input_1–input_4) using {{placeholders}} in the prompt, select a model, and tune temperature and max_tokens to control creativity and length. Typically used in workflows for drafting, summarization, Q&A, content transformation, and reasoning steps.

Inputs

Field	Required	Type	Description	Example
model	True	SELECT	Choose an OpenAI model to run. The list is fetched from the LLM service; if unavailable, a fallback list (e.g., gpt-4o, gpt-4o-mini, gpt-4, gpt-3.5-turbo, o-series) is shown.	gpt-4o
system_prompt	True	STRING	Instructions that guide how the AI should respond (style, format, tone). Leave concise rules for best results.	You are a helpful, concise assistant. Use bullet points when appropriate.
prompt	True	DYNAMIC_STRING	User message to the model. You can include placeholders like {{input_1}}–{{input_4}} to inject context dynamically.	Summarize the following document in 5 bullet points: {{input_1}}
temperature	True	FLOAT	Controls creativity and randomness. Lower for deterministic outputs, higher for more varied responses.	0.5
max_tokens	True	INT	Maximum tokens in the response. Higher allows longer outputs but uses more compute.	1024
input_1	False	STRING	Optional contextual input that can be referenced in the prompt as {{input_1}}.	Full text of a report to summarize
input_2	False	STRING	Optional contextual input that can be referenced in the prompt as {{input_2}}.	User profile JSON for personalization
input_3	False	STRING	Optional contextual input that can be referenced in the prompt as {{input_3}}.	Glossary of terms
input_4	False	STRING	Optional contextual input that can be referenced in the prompt as {{input_4}}.	Style guide or formatting rules

Outputs

Field	Type	Description	Example
Output	STRING	The generated text response from the selected OpenAI model.	Here are five concise bullet points summarizing your document: ...

Important Notes

• This node targets the OpenAI provider via the Salt LLM service and automatically retrieves the current model list; if retrieval fails, it uses a predefined fallback set.
• Dynamic placeholders {{input_1}}–{{input_4}} in the prompt are replaced with the corresponding input values before sending to the model.
• Default timeout is 90 seconds. Very long or complex requests may require prompt simplification or reduced max_tokens to avoid timeouts.
• If you select a model not present in the available options, the node will raise an error. Refresh model options or select a listed model.
• Temperature range is 0–1. Max tokens typically supports up to 4096 for responses, but actual limits depend on the selected model.

Troubleshooting

• Model not found: Ensure you picked a model from the provided selector. If the list looks outdated, try refreshing the workflow or reloading the model list.
• Empty or truncated output: Increase max_tokens or reduce prompt/context length. Some models may have stricter token limits.
• Inconsistent answers: Lower temperature for more deterministic outputs (e.g., 0.1–0.3).
• Placeholders not replaced: Verify your prompt uses {{input_1}}–{{input_4}} exactly and that the respective inputs are connected/provided.
• Timeouts or service errors: Simplify the prompt, lower max_tokens, or try again. If persistent, check LLM service availability or credentials.