AI reply

Type: llm-streaming.

#What it does

1. Builds the AI's instructions from the prompt template, your system_prompt, and the shape of the reply.
2. Adds this turn's input (each name in input_variables with its value) to the step's memory (history_key).
3. Streams the reply. The pipeline speaks speak as the words arrive, ends the call after speaking if the reply sets hangup, and uses interruption_type if the reply includes one.
4. Saves the exchange to memory. The reply's fields are not saved to the call's state.

#Settings

Everything goes in the step's parameters. In the editor, click the step to open its settings.

system_promptstring

In the editor: Instructions. Tell the agent who it is, what to do and how to talk. Default: "".

modelstring

In the editor: Model. Which AI writes the reply. The provider comes with it. Default: "openai/gpt-oss-120b".

llm_return_type{ name: { type, description } }

In the editor: What it decides (More tab). Things the AI fills in every turn. The spoken reply is what the caller hears; the rest goes to your webhook.

interruption_typestring

In the editor: If the caller interrupts. What happens when the caller talks over the agent. Default: "full".

input_variables{ name: { type, description } }

In the editor: What it reads (More tab). Information from earlier steps the AI can use.

history_keystring

In the editor: Shared memory (More tab). Steps with the same name here remember the same conversation. Default: "conversation_history".

summariserboolean

In the editor: Summarise long calls (More tab). Keeps long conversations quick by summarising older turns.

stream_timeoutnumber (seconds)

In the editor: Give up after (More tab). When two AI replies race, drop this one if it takes longer than this.

interruption_typestring

What happens if the caller talks while this is being said. See Interruptions. Default "full".

interruption_metadataobject

Numbers for the timed and length-based interruption types: duration (seconds), length (words), ignore-words.

interruption_type_keystring

Read the interruption type from this value in the state instead, so an earlier step can decide it.

interruption_metadata_keystring

The same, for the interruption numbers.

prompt_templatestring

Built-in instructions to start from. See Prompt templates. Default "base_llm".

summariser_configobject

Tune summariser. See Memory.

#Examples

1{
2  "type": "llm-streaming",
3  "parameters": {
4    "system_prompt": "You are the receptionist at Sunrise Dental. Be warm and brief.",
5    "service": "groq",
6    "model": "openai/gpt-oss-120b",
7    "input_variables": {
8      "user_input": {
9        "type": "str",
10        "description": "What the caller said."
11      }
12    },
13    "llm_return_type": {
14      "speak": {
15        "type": "str",
16        "description": "What to say."
17      },
18      "hangup": {
19        "type": "bool",
20        "description": "True to end the call."
21      }
22    },
23    "history_key": "conversation_history"
24  },
25  "next": "ask_for_input"
26}

#Next steps

• Steps: all the other steps.
• Patterns: see it in a full workflow.