# Chat

The LLM: Chat node lets you use conversation-tuned large language models like GPT-4.

Using the model selection, you can choose different models (and versions) from different providers.

Additional model parameters are available, depending on the model you choose.

# Inputs

Name	Type	Description
prompt	text	The prompt sent to the LLM as the last message
conversation	struct	The full conversation to be sent - a list of chat \| messages.
model	text	the model to use
OpenAI models
temperature	number	the `temperature` parameter

combining `prompt` and `conversation`

If both inputs are provided, the prompt will be appended to the conversation as the last message using the user role.

System prompt

We do not include an extra input for the system prompt at the moment. But you can always add a system prompt via the conversation input:

# Outputs

Name	Type	Description
text	text	The content of the models last response.
response	struct	The full vendor response - useful for inspecting e.g. used tokens.
conversation	struct	The full conversation - the input conversation appended with the last response.
error	struct	the error response from the model vendor if an error occured

# Usage examples

The most basic LLM workflow: take some user input, create a prompt and send it to an LLM. Returns the last LLM response.

A more advanced example combining conversations and memory nodes for remembering user conversations. Takes an additional userId input and will use it to lookup and store conversations in our Key/Value store.