Resources · n8n integration

n8n Redis Vector Store nodeConfigure Redis Vector Store in n8n.

The n8n Redis Vector Store node turns a Redis database into searchable memory for your AI workflows. Two head parameters, five operation modes, and one index name are all it takes to insert documents or rank them against a query. Built for teams already running Redis.

Try the Redis Vector Store node on n8nSee the 2 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Redis Vector Store node do?

It reads and writes a Redis vector index, meaning a store that keeps mathematical representations of your text (embeddings) so an AI can search by meaning instead of by keyword. The node description says it plainly: work with your data in a Redis vector index. It runs as a regular action node on each incoming item, and it has no sub-node port of its own.

First pattern: the semantic cache. A workflow embeds the incoming question, queries Redis with load, and only calls a model when nothing close enough comes back. The published template that pairs Redis with a Hugging Face embeddings node is built exactly this way, and it is the cheapest use of the node.

Second pattern: the knowledge base an agent can search. Set the mode to retrieve-as-tool, fill in Name and Description, then wire the node to the tool connector of an AI Agent. The description matters more than it looks: it is what the model reads to decide whether this store is worth querying. The template that chats with GitHub issues follows that flow, with an OpenAI chat model on the agent.

Third pattern: ingestion. Mode insert writes documents coming from a Default Data Loader into the index, and a new index is created for you when the name does not exist yet. Creating one by hand is only worth it when you need a custom index schema or want to reuse an existing index.

When to prefer something else: to fetch documents inside a chain rather than an agent, put a Vector Store Retriever between the Question and Answer Chain and this node. To have results summarized before they reach the model, use the Vector Store Question Answer Tool instead of connecting the store straight to the agent's tool port, the same slot an AI Agent Tool would occupy.

Known limits, in order of how often they bite. Redis needs the Redis Query Engine enabled, so Redis Open Source v8.0 and later, Redis Cloud or Redis Software. Metadata filters run as an OR query: list two fields and one match is enough. Overwrite Documents deletes the index along with the documents, while Time-to-Live expires documents only and leaves the index standing. Teams that want the wider picture can read the n8n review or follow the n8n training.

Connect

What do you need before the node runs?

  1. 01

    Get a Redis with the Query Engine

    The node searches vectors through the Redis Query Engine, so a plain Redis will not answer. Three deployments qualify: Redis Open Source in v8.0 and later, which ships the Query Engine by default, Redis Cloud as a fully managed service, and Redis Software for a self-managed setup. Pick one before touching n8n, because every operation mode depends on it.

  2. 02

    Create the Redis credential in n8n

    Open the Credentials menu and add the Redis credential: Password, Host (defaults to localhost), Port (defaults to 6379), Database Number (defaults to 0) and an SSL toggle, off by default, with an option to disable TLS verification for self-signed certificates. Fill it once; every workflow on the instance reuses it. The node has no Authentication selector.

  3. 03

    Point the node at an index

    Fill Redis Index with the name of the vector search index, or pick an existing one from the list the field offers. Nothing to prepare on the Redis side: a new index gets created when the name you type does not exist yet. Then choose the operation mode and run the node once on a single item to confirm the connection.

Parameters

Which parameters does the node expose?

The Redis Vector Store node has 2 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Operation Mode

mode

What you see in n8n

Notes & use cases

This selector decides what the node is: a writer, a reader, or a resource other AI nodes query.

Key parameters

  • Operation Mode load (Get Many): ranked documents for a query, with Prompt as the search text and Limit as the number of results.
  • Operation Mode insert (Insert Documents): writes documents into the store.
  • Operation Mode retrieve (Retrieve Documents (As Vector Store for Chain/Tool)): exposes the store to a chain.
  • Operation Mode retrieve-as-tool (Retrieve Documents (As Tool for AI Agent)): needs Name, Description and Limit.
  • Operation Mode update (Update Documents): rewrites a document through its ID.
Use cases
a night workflow inserts your help center, a day one runs in load with Include Metadata on, so answers cite a source.
02

Redis Index

redisIndex

What you see in n8n

Notes & use cases

The index is the address of your vectors. This field names it, and the same name has to come back in every mode that touches the same documents.

Key parameters

  • Redis Index: a resource locator, so either type the index name or choose an existing one from the list. Typed from a previous node, it takes an expression such as {{ $json.index }}, which reads the field index on the incoming item.
Use cases
one index per client keeps a support agent from reading another account's documents, and the workflow switches between them with a single expression instead of a duplicated branch.
Need help

Need help automating Redis Vector Store with n8n?

A person reads every message.

FAQ

Redis Vector Store and n8n, answered

01Is the Redis Vector Store node included in n8n?
Yes. It belongs to the n8n AI package, shipped with n8n, on n8n Cloud as well as on a self-hosted instance under the Community Edition and its Sustainable Use license. Nothing to install, and n8n charges nothing for the node itself. What you pay for sits elsewhere: the Redis deployment you choose, and the model provider whose embeddings you generate, since that provider bills its own API calls under its own terms. A workflow built on n8n Cloud runs identically once exported to a self-hosted instance.
02What do you need to make it work?
Two things. A Redis database with the Redis Query Engine enabled, which means Redis Open Source v8.0 and later, Redis Cloud or Redis Software, and the Redis credential created once in the Credentials menu of n8n. The node shows no Authentication selector, so there is no method to pick. You also need an embeddings sub-node, meaning a node plugged into a port that turns text into vectors, and an index name in Redis Index. If the index does not exist, it gets created.
03What are the limits of the n8n Redis Vector Store node?
The node is at version 1 and covers five operation modes, so anything outside them is out of reach. Metadata filters behave as an OR query: with several fields listed, one match is enough, which makes narrow filtering harder than it looks. Metadata itself is set when documents are loaded, through the document loader, not in this node. Overwrite Documents deletes the index too, not only the documents. Time-to-Live expires documents in seconds but never the index. Plan the index name before the first insert, since a rename means reindexing everything.
04What do you plug in, and what should you start with?
At minimum an embeddings node, since documents are stored as vectors and a query has to be embedded the same way before it can be compared. Start with Get Many on an index you filled by hand: one prompt, a Limit of 10, and you see what ranking returns without any agent in the picture. Once that works, move to Retrieve Documents (As Tool for AI Agent) and connect the node to an agent, or put a Vector Store Retriever in front of a Question and Answer Chain.
05n8n or Make for a Redis vector store?
It depends on where your data has to live. Make is hosted only, billed per operation, with no self-hosting option. n8n runs on your own server through Docker or npm, or on n8n Cloud, which matters when the documents you index are internal and should not leave your infrastructure. Cost models differ too: per operation on one side, per instance on the other once self-hosted. Make has its own visual logic and a large catalog. The deciding factor here is usually hosting and data control.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations