fix(edge): meter and size-cap knowledge embeddings in embed-chunks and search-knowledge

server/supabase/functions/search-knowledge/index.ts

@@ -1,100 +1,46 @@
-// deno-lint-ignore no-import-prefix
-import { createClient } from 'https://esm.sh/@supabase/supabase-js@2.39.7'
-import { corsHeaders, handleCorsPreflightRequest } from '../_shared/cors.ts'
-import { requireUser, authErrorResponse, type AuthError } from '../_shared/auth.ts'
+// server/supabase/functions/search-knowledge/index.ts
+// Composition root: wires the real adapters into the search-knowledge use case.
+
+import { createClient } from '@supabase/supabase-js'
+import { requireUser } from '../_shared/auth.ts'
+import { createServiceRoleClient } from '../_shared/quota.ts'
 import { readProviderKey } from '../_shared/provider-key.ts'
+import { createOpenAIEmbeddingProvider } from '../_shared/openai-embeddings.ts'
+import { createSupabaseEmbeddingUsageStore } from '../_shared/embedding-quota.ts'
+import { type ChunkSearcher, createSearchKnowledgeHandler, SIMILARITY_THRESHOLD } from './handler.ts'
 
-const EMBEDDING_DIMENSIONS = 1536
-const PROVIDER_TIMEOUT_MS = 45_000
+let serviceClient: ReturnType<typeof createServiceRoleClient> | null = null
+function service(): ReturnType<typeof createServiceRoleClient> {
+  serviceClient ??= createServiceRoleClient()
+  return serviceClient
+}
 
-function json(status: number, body: Record<string, unknown>): Response {
-  return new Response(JSON.stringify(body), {
-    status,
-    headers: { ...corsHeaders, 'Content-Type': 'application/json' },
+/** match_knowledge_chunks runs as the caller so RLS limits results to their own chunks. */
+const searchChunks: ChunkSearcher = async (req, embedding, count) => {
+  const authHeader = req.headers.get('Authorization') ?? ''
+  const supabaseUrl = Deno.env.get('SUPABASE_URL') ?? ''
+  const anonKey = Deno.env.get('SUPABASE_ANON_KEY') ?? ''
+  if (!supabaseUrl || !anonKey) return { ok: false, reason: 'unavailable' }
+
+  const userClient = createClient(supabaseUrl, anonKey, {
+    global: { headers: { Authorization: authHeader } },
+    auth: { persistSession: false, autoRefreshToken: false },
   })
+  const result = await userClient.rpc('match_knowledge_chunks', {
+    query_embedding: embedding,
+    match_count: count,
+    similarity_threshold: SIMILARITY_THRESHOLD,
+  })
+  if (result.error) return { ok: false, reason: 'failed' }
+  return { ok: true, results: Array.isArray(result.data) ? result.data : [] }
 }
 
-function isEmbedding(value: unknown): value is number[] {
-  return Array.isArray(value)
-    && value.length === EMBEDDING_DIMENSIONS
-    && value.every((entry) => typeof entry === 'number' && Number.isFinite(entry))
-}
-
-Deno.serve(async (req: Request) => {
-  const preflight = handleCorsPreflightRequest(req)
-  if (preflight) return preflight
-  if (req.method !== 'POST') return json(405, { error: 'method_not_allowed' })
-
-  try {
-    await requireUser(req)
-    const body = await req.json().catch(() => null) as {
-      query?: unknown
-      count?: unknown
-    } | null
-    const query = typeof body?.query === 'string' ? body.query.trim() : ''
-    const count = body?.count === undefined ? 5 : body.count
-    if (!query || query.length > 4_000) return json(400, { error: 'invalid_query' })
-    if (typeof count !== 'number' || !Number.isInteger(count) || count < 1 || count > 20) {
-      return json(400, { error: 'invalid_count' })
-    }
-
-    const openaiKey = readProviderKey('OPENAI_API_KEY')
-    if (!openaiKey) return json(503, { error: 'embedding_provider_unavailable' })
-
-    let embeddingResponse: Response
-    try {
-      embeddingResponse = await fetch('https://api.openai.com/v1/embeddings', {
-        method: 'POST',
-        headers: {
-          Authorization: `Bearer ${openaiKey}`,
-          'Content-Type': 'application/json',
-        },
-        body: JSON.stringify({
-          model: 'text-embedding-3-small',
-          input: query,
-          dimensions: EMBEDDING_DIMENSIONS,
-        }),
-        signal: AbortSignal.timeout(PROVIDER_TIMEOUT_MS),
-      })
-    } catch {
-      return json(502, { error: 'embedding_upstream_failed' })
-    }
-    if (!embeddingResponse.ok) return json(502, { error: 'embedding_upstream_failed' })
-
-    const embeddingPayload = await embeddingResponse.json().catch(() => null) as {
-      data?: Array<{ embedding?: unknown }>
-    } | null
-    const queryEmbedding = embeddingPayload?.data?.[0]?.embedding
-    if (!isEmbedding(queryEmbedding)) return json(502, { error: 'embedding_response_invalid' })
-
-    const authHeader = req.headers.get('Authorization') ?? ''
-    const supabaseUrl = Deno.env.get('SUPABASE_URL') ?? ''
-    const anonKey = Deno.env.get('SUPABASE_ANON_KEY') ?? ''
-    if (!supabaseUrl || !anonKey) return json(503, { error: 'knowledge_storage_unavailable' })
-
-    const userClient = createClient(supabaseUrl, anonKey, {
-      global: { headers: { Authorization: authHeader } },
-      auth: { persistSession: false, autoRefreshToken: false },
-    })
-    const result = await userClient.rpc('match_knowledge_chunks', {
-      query_embedding: queryEmbedding,
-      match_count: count,
-      similarity_threshold: 0.5,
-    })
-    if (result.error) return json(500, { error: 'knowledge_search_failed' })
-
-    return json(200, { results: result.data ?? [] })
-  } catch (error) {
-    if (
-      error
-      && typeof error === 'object'
-      && 'status' in error
-      && (error.status === 401 || error.status === 403)
-      && 'message' in error
-      && typeof error.message === 'string'
-    ) {
-      return authErrorResponse(error as AuthError, corsHeaders)
-    }
-    return json(500, { error: 'internal_error' })
-  }
-})
+Deno.serve(createSearchKnowledgeHandler({
+  authenticate: requireUser,
+  embeddingProvider: () => {
+    const key = readProviderKey('OPENAI_API_KEY')
+    return key ? createOpenAIEmbeddingProvider(key) : null
+  },
+  usageStore: () => createSupabaseEmbeddingUsageStore(service()),
+  searchChunks,
+}))