How to Implement AI-Powered Search in FlutterFlow

1-- Enable pgvector (if not already done via UI)
2create extension if not exists vector;
3
4-- Documents table with embedding column
5create table documents (
6  id uuid primary key default gen_random_uuid(),
7  title text not null,
8  body text not null,
9  embedding vector(1536),
10  created_at timestamptz default now()
11);
12
13-- Index for fast similarity search
14create index on documents
15using ivfflat (embedding vector_cosine_ops)
16with (lists = 100);
17
18-- RLS: allow authenticated reads
19alter table documents enable row level security;
20create policy "Public read" on documents
21  for select using (true);

1// supabase/functions/embed-document/index.ts
2import { serve } from 'https://deno.land/std@0.168.0/http/server.ts';
3import { createClient } from 'https://esm.sh/@supabase/supabase-js@2';
4
5serve(async (req) => {
6  const { title, body } = await req.json();
7
8  // Get embedding from OpenAI
9  const embeddingRes = await fetch('https://api.openai.com/v1/embeddings', {
10    method: 'POST',
11    headers: {
12      'Authorization': `Bearer ${Deno.env.get('OPENAI_API_KEY')}`,
13      'Content-Type': 'application/json',
14    },
15    body: JSON.stringify({
16      input: `${title} ${body}`,
17      model: 'text-embedding-ada-002',
18    }),
19  });
20
21  const { data } = await embeddingRes.json();
22  const embedding = data[0].embedding;
23
24  // Store in Supabase
25  const supabase = createClient(
26    Deno.env.get('SUPABASE_URL')!,
27    Deno.env.get('SUPABASE_SERVICE_ROLE_KEY')!
28  );
29
30  const { error } = await supabase
31    .from('documents')
32    .insert({ title, body, embedding });
33
34  if (error) return new Response(JSON.stringify({ error }), { status: 500 });
35  return new Response(JSON.stringify({ success: true }), { status: 200 });
36});

1-- First, add this SQL function in Supabase SQL editor:
2create or replace function match_documents(
3  query_embedding vector(1536),
4  match_count int default 5
5)
6returns table (id uuid, title text, body text, similarity float)
7language sql stable as $$
8  select id, title, body,
9    1 - (embedding <=> query_embedding) as similarity
10  from documents
11  order by embedding <=> query_embedding
12  limit match_count;
13$$;
14
15-- Edge Function: supabase/functions/semantic-search/index.ts
16// (same imports as embed-document)
17serve(async (req) => {
18  const { query } = await req.json();
19
20  const embeddingRes = await fetch('https://api.openai.com/v1/embeddings', {
21    method: 'POST',
22    headers: {
23      'Authorization': `Bearer ${Deno.env.get('OPENAI_API_KEY')}`,
24      'Content-Type': 'application/json',
25    },
26    body: JSON.stringify({ input: query, model: 'text-embedding-ada-002' }),
27  });
28
29  const { data } = await embeddingRes.json();
30  const queryEmbedding = data[0].embedding;
31
32  const supabase = createClient(
33    Deno.env.get('SUPABASE_URL')!,
34    Deno.env.get('SUPABASE_SERVICE_ROLE_KEY')!
35  );
36
37  const { data: results, error } = await supabase
38    .rpc('match_documents', {
39      query_embedding: queryEmbedding,
40      match_count: 10,
41    });
42
43  if (error) return new Response(JSON.stringify({ error }), { status: 500 });
44  return new Response(JSON.stringify({ results }), { status: 200 });
45});

1// Custom Action: semanticSearch
2import 'dart:convert';
3import 'package:http/http.dart' as http;
4
5Future<List<dynamic>> semanticSearch(String query) async {
6  const supabaseUrl = 'https://YOUR_PROJECT.supabase.co';
7  const anonKey = 'YOUR_ANON_KEY';
8
9  final response = await http.post(
10    Uri.parse('$supabaseUrl/functions/v1/semantic-search'),
11    headers: {
12      'Content-Type': 'application/json',
13      'Authorization': 'Bearer $anonKey',
14    },
15    body: jsonEncode({'query': query}),
16  );
17
18  if (response.statusCode != 200) return [];
19  final data = jsonDecode(response.body);
20  return data['results'] as List<dynamic>;
21}