/headless-vector-search

Supabase Toolkit to perform vector similarity search on your knowledge base embeddings.

Primary LanguageTypeScriptMIT LicenseMIT

Headless Vector Search

Provides a vector/similarity search for any documentation site. It's headless, so that you can integrate it into your existing website.

How it works:

  • This repo initializes a new docs schema inside your database.
  • The accompanying GitHub Action ingests your markdown docs into your database as embeddings.
  • This repo provides an Edge Function that handles user queries, converting them into ChatGPT-like responses.

Tech stack:

  • Supabase: Database & Edge Functions.
  • OpenAI: Embeddings and completions.
  • GitHub Actions: for ingesting your markdown docs.

Set-up

Start by creating a new Supabase Project: database.new.

  1. Clone this repo
  2. Link the repo to your remote project: supabase link --project-ref XXX
  3. Apply the database migrations: supabase db push
  4. Set your OpenAI key as a secret: supabase secrets set OPENAI_KEY=sk-xxx
  5. Deploy the Edge Functions: supabase functions deploy --no-verify-jwt
  6. Expose docs schema via API in Supabase Dashboard settings > API Settings > Exposed schemas
  7. Setup supabase-vector-embeddings GitHub action in your Knowledge Base repo. You will see the embeddings populated in your database after the GitHub Action has run.

Usage

  1. Find the URL for the vector-search Edge Function in the Functions section of the Dashboard.
  2. Inside your appliction, you can send the user queries to this endpoint to receive a streamed response from OpenAI.
See cURL example 
 curl -i --location --request GET 'https://your-project-ref.functions.supabase.co/vector-search?query=What%27s+Supabase%3F'
See EventSource example 
const onSubmit = (e: Event) => {
  e.preventDefault()
  answer.value = ""
  isLoading.value = true

  const query = new URLSearchParams({ query: inputRef.current!.value })
  const projectUrl = `https://your-project-ref.functions.supabase.co`
  const queryURL = `${projectURL}/${query}`
  const eventSource = new EventSource(queryURL)

  eventSource.addEventListener("error", (err) => {
    isLoading.value = false
    console.error(err)
  })
  
  eventSource.addEventListener("message", (e: MessageEvent) => {
    isLoading.value = false

    if (e.data === "[DONE]") {
      eventSource.close()
      return
    }

    const completionResponse: CreateCompletionResponse = JSON.parse(e.data)
    const text = completionResponse.choices[0].text

    answer.value += text
  });

  isLoading.value = true
}

Showcase

License

MIT