AI Package

RAG-powered conversational AI via @skam/ai

Overview

The @skam/ai package provides a simple TypeScript wrapper around the Supabase Edge Function rag-chat. It abstracts the complexity of vector search and Gemini AI invocation into a single function call.

API Reference

packages/ai/src/index.ts

import { SupabaseClient } from '@supabase/supabase-js'

export async function askRagChat(
  supabase: SupabaseClient,
  question: string
): Promise<{
  answer?: string
  error?: string
  cached?: boolean
}>

Parameters

supabase

SupabaseClient

An authenticated Supabase client instance

question

string

The user's prompt or question to ask the RAG AI

Usage Example

Server Component / Server Action

import { askRagChat } from '@skam/ai'
import { createClient } from '@/lib/supabase/server'

const supabase = await createClient()
const { answer, error, cached } = await askRagChat(
  supabase,
  'What are the shipping options?'
)

if (error) {
  console.error(error)
} else {
  console.log(answer) // AI-generated response
  console.log(cached) // true if served from cache
}

How It Works

Validate

Question string is checked for non-empty input.

Invoke

Calls supabase functions invoke("rag-chat").

Semantic similarity search over embedded chunks.

Generate

Context is fed to Google Gemini for grounded answer.

Return

Returns { answer, cached } or { error }.