Resources · n8n integration

n8n Oracle Database Vector Store nodeConfigure Oracle Database Vector Store in n8n.

Oracle Database already holds the data your AI keeps guessing about. The n8n Oracle Database Vector Store node turns a vector table into something a chain or an agent can query, through 2 parameters and 4 operation modes, for teams that would rather not move their data to a second database.

Try the Oracle Database Vector Store node on n8nSee the 2 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Oracle Database Vector Store node actually do?

It reads and writes a vector table inside Oracle Database, so a workflow can store documents as vectors, mathematical representations of text, and later fetch the ones closest to a question. The node covers the whole round trip: it inserts documents, ranks them for a search prompt, or exposes the table to an AI agent as a tool. Your Oracle Database instance has to support Oracle AI Vector Search for any of that to work.

The first scenario is the loading side. You point Table Name at a table, pick insert under Operation Mode, and every item arriving in the workflow becomes a row of vectors. The table does not have to exist yet: the node creates it. That is how a folder of product sheets or a support knowledge base ends up queryable, one run at a time.

The second scenario is plain search inside a workflow, with no agent involved. In load mode you fill Prompt with the question, set Limit to the number of results you want, and the node returns ranked documents you can push straight into Slack or a row in Google Sheets. Nothing here needs a chat model.

The third scenario hands the table to an AI Agent. With retrieve-as-tool, the node shows up on the agent's tool connector, and the agent decides on its own when to search it. This is where Description earns its keep: the wording you type there is what the model reads to decide whether the tool is relevant. A vague description means a tool that never gets called, or gets called for everything.

When should you reach for something else? If the question is answered by a single lookup on a known key, a regular database query is faster and cheaper than a similarity search. If you only need a prompt and an answer with no stored corpus, a chat model node connected to OpenAI or Anthropic is enough on its own. The vector store earns its place when the corpus is larger than a context window and changes over time.

The known limits are worth stating up front. This node is at version 1 and exposes 2 head parameters, so most of the fine tuning lives in Options rather than on the panel itself. Embedding quality is not its job: the embeddings sub-node you plug in decides how good the ranking is, and the provider behind that sub-node bills its own API calls. If your team is still picking a pattern, the n8n training walks through the same choices on real workflows.

Parameters

Which parameters do you set on the node?

The Oracle Database 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 searcher, or a resource other AI nodes lean on. Set it first, the rest of the panel changes with it.

Key parameters

  • Operation Mode: insert writes documents, load returns ranked documents, retrieve opens the store to a chain or tool, retrieve-as-tool opens it to an agent.
  • Prompt and Limit: the query, often {{ $json.question }}, and how many top results come back.
  • Name, Description, Include Metadata, Rerank Results, ID, Options: the tool identity an agent reads before calling, plus the extra switches.
Use cases
one workflow inserts nightly, a second answers questions, a third leaves the table to an agent.
02

Table Name

tableName

What you see in n8n

Notes & use cases

Here you name the Oracle table the vectors live in. It is the only thing tying an insert workflow and a search workflow together, so a typo returns an empty answer rather than an error.

Key parameters

  • Table Name: the table to store the vectors in; if it does not exist, the node creates it. Use a plain value like company_knowledge_base, or {{ $json.tableName }} when one workflow serves several corpora.
Use cases
keep one table per corpus, product docs on one side, support tickets on the other. Since a missing table is created on the fly, check the spelling before the first insert run.
Need help

Need help automating Oracle Database Vector Store with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the Oracle Database Vector Store node included in n8n?
Yes. It belongs to the n8n AI package, shipped with n8n, so there is nothing to install and no extra cost on the n8n side, on n8n Cloud as well as on a self-hosted instance. n8n self-hosts through Docker or npm, or runs on n8n Cloud, the hosted offer n8n operates, and a workflow behaves the same either way. What you do pay for sits elsewhere: the provider behind the embeddings sub-node bills its own API calls according to its own terms, and n8n adds nothing on top of that.
02What do you need to set up before the node works?
Nothing on the node itself. It has no credential and no Authentication selector, which is why there is no connection step on this page. The account side is handled by the model sub-node you plug in: that sub-node carries the provider credential. Two things still have to be true. Your Oracle Database instance must support Oracle AI Vector Search, since that is what the vector operations run on. And any field marked as required, Prompt in a search, Name and Description for an agent tool, blocks the node from running when left empty.
03What are the limits of the n8n Oracle Database Vector Store node?
The node is at version 1 and exposes 2 head parameters, so the panel is deliberately short and most tuning happens through Options. Distance Strategy is only available in Get Many and Retrieve Documents modes, not when inserting, and Metadata Filter is available in Get Many and in both Retrieve Documents modes. Filters set through the interface are combined as an AND query: every field has to match. For richer filtering, the node passes filters through to Oracle AI Vector Search, which supports arrays, nested filters, and operators such as $gte, $nin and $and. A workflow built on an older version may show fewer options.
04What do you plug into it, and where should a beginner start?
Start in load mode with a table you filled yourself through insert, and no agent at all. You get ranked documents in the output panel, which makes it obvious whether your embeddings and your prompt are doing their job. Once that is stable, move up: retrieve exposes the store to a chain or a tool, and retrieve-as-tool hands it to an agent that calls it on its own. Distance Strategy offers Cosine, Inner Product, Euclidean, Manhattan, Euclidean Squared and Hamming, so leave it alone until you have a reason to change it.
05n8n or Make for a vector store workflow?
It depends on where the data has to stay. n8n self-hosts through Docker or npm, so your documents and your Oracle connection never leave your own environment, and it also runs on n8n Cloud when you would rather not operate anything. Make is a hosted automation platform with no self-hosting option, billed per operation, which is predictable when volumes are low and less so when a workflow embeds thousands of documents. The visual logic differs too: n8n wires AI sub-nodes onto ports, which is exactly how this node is meant to be assembled.
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