Tool

AIcalledcore.ai.tool

A tool the LLM can call. It declares a name, a description and its parameters. When the model calls it, call carries the arguments into the body you wired, and the answer that comes back on result goes back to the model.

Called by the LLM. When the model calls the tool, call carries the arguments into the body you wired. The value that body sends to result is the tool’s answer, handed back to the LLM, so the Tool never runs as a step of its own.

Inputs

Port	Type	Notes
result	text	The tool’s answer, handed back to the LLM

Outputs

Port	Type	Notes
call	tool-call	Right-click the node to add a wired Tool Args
trigger	event

Knobs

Right-click a knob and pick Convert to input to drive it from a wire.

Knob	Kind	Default	Notes
name name	text	my_tool
description	text block	Describe what this tool does for the model.
schema_mode	choice	auto	auto, manual
schema	text block	{"type":"object","properties":{},"required":[]}	only when schema_mode is manual

Example

Tells the model it can call something. Give it a name and a description the model reads, wire call into the LLM tools, and right-click the node to add its Tool Args. When the model calls the tool, Tool Args fires with the arguments; wire the answer at the end of that branch back into result and it goes to the model. No answer within two minutes, and the model gets an error for that call.

In auto mode the parameters come from the Tool Args fields; manual takes a JSON schema. The model finds a tool by its name, so two tools with one name keep the graph from turning On.

More in AI

• Embed: Turn text into an embedding vector with the embed model you pick, the way the LLM picks its model.
• LLM: A chat / multimodal model.
• Rerank: Reorder candidates by how well each one answers the query, and keep the best.
• STT: Speech to text.
• Tool Args: Where a tool’s body starts.
• TTS: Text to speech.