Vectors

Storefiredcore.vectors

Index and search an embedded vector store. The operation knob reshapes the knobs, inputs and outputs. Search takes one or more query embeddings, keeps each item’s best score across them and returns the top k. It compares every stored vector by cosine similarity, which suits thousands of items.

Fired. It runs when an event or a value lands on trigger, and pulls its other inputs at that moment.

Inputs

Port	Type	Notes
trigger	event	Fires the node
vectors	vectors
embedding	embedding	Only when operation is search or index
text	text	Optional
tag	any	Optional, grows: one socket per wire

Outputs

Port	Type	Notes
results	memory-set	Only when operation is search
ids	json	Only when operation is search
ref_id	int	Only when operation is index
affected	int	Only when operation is delete or clear
trigger	event

Knobs

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

Knob	Kind	Default	Notes
operation	choice	search	search, index, delete, clear
namespace	text	default	{tag} and {{secret.NAME}} work here; placeholder facts / messages / summaries
top_k top k	number	5	1 to 100; only when operation is search
min_score min score	number	0.45	0.0 to 1.0, step 0.01; only when operation is search
metadata	text block	empty	{tag} and {{secret.NAME}} work here; only when operation is index
ref	text	empty	{tag} and {{secret.NAME}} work here; only when operation is delete; placeholder external ref to delete

Placeholder for metadata

ref = {key}
role = user

Example

One node for four jobs on a vector store. index stores an embedding with its text, search returns the top_k closest hits above min_score, delete removes the entries with a ref, and clear empties the namespace. embedding is a required input in every operation, so clear and delete need a wire there too, though they do not read it. Search also takes a list of embeddings and merges the hits, for a question asked two ways.

More in Store

• Database: An embedded SQLite database.
• DB: Read and write a SQLite store.
• KV: Read and write a key-value store.
• KV Store: A key-value store that keeps its values between runs.
• Vector Store: A vector store kept in a file on this machine.