Google Doc to Pinecone vector store

What it is

This n8n workflow copies one Google Doc into a Pinecone index. You press Execute workflow, and the Google Docs node fetches the document. A data loader packages its content, an OpenAI embeddings node turns the text into vectors, and the Pinecone Vector Store node writes them into your index under a namespace. It has five nodes and no schedule. Afterwards a chatbot or an agent that reads the same index can find the document by meaning.

Who it suits

Developers and teams building a question-answering bot over their own documents, who need the first step: getting a document into a vector database where a model can look it up by meaning. Skip it if you want documents kept in sync on their own. It loads one document per click. It never searches.

Where to use it

Any n8n that can reach Google, OpenAI and Pinecone will do. It only fills the index, so a second workflow, a chatbot or an app of your own has to query it, and that part is not in this file.

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Its author puts setup at about 13 minutes, which assumes the three accounts exist.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add a Google Docs credential and select it in Get a document. Replace YOUR_DOC_ID with the ID of your document, the long string between /d/ and /edit in its address.
3. Add a Pinecone credential and select it in Pinecone Vector Store. Replace YOUR_INDEX_NAME with your index, and change the namespace Contacts if you want another.
4. Add an OpenAI credential and select it in Embeddings OpenAI.
5. Save and press Execute workflow.

How to use it

Press Execute workflow. Each node shows how many items it handled, and the vector count of your index in Pinecone’s console should rise. Each press adds the document’s vectors again. Insert mode does not replace earlier ones, so two clicks store the document twice. To refresh a document, clear its namespace first, or turn on the Pinecone node’s Clear Namespace option.

Example

Put the ID of a returns-policy document in Get a document and press Execute workflow. The index then holds the policy’s text as vectors under Contacts. A retrieval step can answer a question such as how long a customer has to send an item back by meaning, with no exact wording to match.

For newcomers and experts

If you are new. An embedding is a list of numbers that stands for the meaning of a piece of text. A vector database such as Pinecone keeps those lists and finds the stored texts closest in meaning to a question. This workflow only fills the database, so nothing here answers questions yet.

For experts. The embeddings node has no options set, so it uses the node’s default model, and the index’s dimension has to match that model’s output. The data loader is on its defaults and no text splitter is connected. Open it to see what it takes as text and how it cuts a long document before you load a big one. The namespace in the file is Contacts, which suits a contacts list better than a policy, so rename it.

Good to know

The cost is the embedding calls and your Pinecone plan. Duplicates and a mismatched index are the two things that go wrong.

• Billing: OpenAI charges for the embeddings, and Pinecone bills by your plan
• Duplicates: every run inserts again, so clear the namespace before a reload
• Index: it must exist already, and a wrong dimension makes Pinecone refuse the insert
• Access: the Google account you authorize has to be able to open the document
• Privacy: the document’s text goes to OpenAI and then to Pinecone
• Credentials: the JSON holds none, and nothing from its author’s account