cloudflare vector search

This project is a Cloudflare Page that allows you to perform a semantic search over a set of documents.

Essentially you can search for meaning, not just keywords.

The nice thing about using Cloudflare is that it's incredibly fast and cheap. You can deploy this to a free Cloudflare Page and it will scale to millions of documents.

See below for examples on how to submit and search documents.

For large documents, we support a bulk submit endpoint that will automatically split your document into paragraphs and index each paragraph separately.

Usage

Submitting

curl -X POST -H "Content-Type: application/json" -d '{"text": "cloudflare"}' "http://127.0.0.1:8788/api/submit"

Submitting large documents

curl -X POST -H "Content-Type: application/json" -d '{"text": "big document"}' "http://127.0.0.1:8788/api/bulk-submit"

Searching

curl "http://127.0.0.1:8788/api/search?query=cloudflare"

Which returns:

[
  {
    "id": "fad58cf0-78dc-4170-9ae3-38f5c62868e3",
    "namespace": "default",
    "text": "cloudflare",
    "metadata": {},
    "indexed_at": "2023-05-08T08:02:29.058Z",
    "similarity": 0.9999999999999998
  }
]

Additional options

You can pass a namespace parameter to both the submit and search endpoints to namespace your queries.

curl -X POST -H "Content-Type: application/json" -d '{"text": "cloudflare", "namespace": "my-namespace"}' "http://127.0.0.1:8788/api/submit"

And then:

curl "http://127.0.0.1:8788/api/search?query=cloudflare&namespace=my-namespace"

A namespace could be a user ID, a document ID, or anything else you want to use to group your queries.

You can also pass in a key/value metadata object to the submit endpoint to store additional data about your query.

Development

Setup the env vars as described below and then run:

yarn
yarn dev

Deployment

Database

cloudflare-vector-search uses a Postgres database to store the vectors. We recomment using Neon.

Sign up for a free account and paste in the schema from data/schema.sql into the SQL editor.

Cloudflare

Clone this repo and then set up a Cloudflare Page with their admin.

We expect the following env vars:

• DATABASE_URL: The URL to your Postgres database
• OPENAI_API_KEY: Your OpenAI API key
• AUTH_SECRET: A secret used to authenticate requests to the API (optional)