Skip to content

Command Palette

Search for a command to run...

AACSearch OSSearch · Discovery · GraphRAG · Security

Search OS for products,
catalogs and knowledge bases

One control plane for search and discovery: text, voice, images, vector search, recommendations, suggestions, GraphRAG, scoped access and analytics.

Find a product, document, error, user or answer…
p99 < 50 msVector + keywordScoped accessTenant isolationSearch analytics
2.5M+
Documents indexed
< 50 ms
Search latency
3
Regions live
AACSearch OS Console
AACSearch OS Console
Live request
Query intent
error 401 in SDK after token refresh
TextVoiceImageScreenshotChat
Understanding
intenttroubleshooting
entityerror_401
collectiondeveloper_docs
Route
keyword
vector
graph
rerank
38 ms·hybrid match
Results
01
SDK authentication guideDocs

How scoped tokens expire and refresh.

02
Token refresh referenceAPI

TTL, HMAC signature and origin checks.

03
Origin allow-listSecurity

Why browser requests can be rejected.

Answer with citations

The 401 is likely caused by an expired scoped token or an origin that is not included in the allow-list.

Token refresh guideOrigin allow-list
Scoped access
tenant: demo-store
TTL: 1h
origin verified

Not a search API. A Search OS.

AACSearch OS connects input, understanding, retrieval, recommendations, GraphRAG, analytics and security in one managed layer.

01

Input Layer

Text, voice, photos, screenshots and AI dialogue. Users express intent the way they prefer.

TextVoiceImageScreenshotChat
02

Understanding Layer

Transcription, OCR, embeddings, typo correction, intent detection, entity extraction and query rewriting.

EmbeddingsOCRTranscriptionIntentTypo correction
03

Retrieval Core

Full-text, vector, visual similarity, facets, graph retrieval and hybrid ranking.

Full-textVectorVisualGraphHybrid
04

Intelligence Layer

Suggestions, recommendations, no-result recovery, GraphRAG and AI answers with citations.

SuggestionsRecommendationsGraphRAGAI answers
05

Control Plane

Indexes, relevance rules, synonyms, pinned results, A/B tests, analytics, API keys, quotas and tenants.

IndexesRelevanceSynonymsA/B testsQuotas
06

Security Plane

Scoped tokens, origin allow-list, tenant isolation, SSO, audit log, IP allow-list and SLA.

Scoped tokensTenant isolationOrigin allow-listSSO
AACSearch OS — architecture
Input Layer
text · voice · image · chat
Understanding Layer
OCR · NLP · embeddings · intent
Retrieval Core
full-text · vector · visual · graph
Intelligence Layer
suggestions · recommendations · GraphRAG
Control Plane
relevance · synonyms · quotas · tenants
Security Plane
scoped tokens · isolation · audit log

Built for your use case

One API, many shapes. Bring your data — AACsearch handles the indexing, ranking, and delivery.

E-commerce product search

Faceted filters, synonym expansion, zero-results handling, and per-user scoped tokens — all in a single search call.

Help center & docs

Index your knowledge base and surface relevant articles in milliseconds. Pair with the Knowledge module for Q&A over PDFs.

Multi-tenant SaaS

One AACsearch index per tenant, scoped tokens per end-user. Your customers never see each other's data — by design.

Developer tools

Semantic search over code, configs, and logs. MCP Server lets your AI agents search and update indexes programmatically.

Internal tools

Fast search over CRM records, inventory, employee directories, or any internal dataset your team queries daily.

Content sites

Instant search for blogs, changelogs, and documentation. The Browser SDK drops into any front end in under an hour.

When you need Search OS, not just search

Fits if

  • Search runs in production and must not depend on a dedicated search team.
  • Users search from the browser and exposing permanent API keys is not acceptable.
  • You have multiple tenants, customers, regions or storefronts on one platform.
  • You want predictable monthly costs as index size and query volume grow.
  • You need visibility into what users search for, where they find nothing, and what drives conversion.

Not needed if

  • You need an on-premises search deployment with no external dependencies.
  • Your data cannot leave your infrastructure perimeter under any conditions.
  • You need deep ranking customization at the scoring plugin level.
  • Your index has fewer than 1 000 documents and a simple database query handles the use case.

Search starts with user intent, not the index

AACsearch handles different intent types: exact entities, fuzzy queries, category navigation, natural language and no-result recovery.

Exact entities

Order ID, SKU, error code, username or document name. AACsearch finds exact matches faster than Ctrl+F.

Fuzzy queries

Users misspell, use transliteration, or abbreviate. Typo tolerance and prefix search fix to relevant results.

Category navigation

Faceted filters: brand + price + availability + size. Multi-search across collections in one round-trip.

Natural language answers

For docs and knowledge base — search-first, answer-augmented UX. Hybrid: result list + AI answer with citations.

Autocomplete & suggestions

Show relevant suggestions from the first character. Keyboard navigation, recent searches, pinned suggestions.

Zero-result handling

No-result recovery: suggestions, similar queries, automatic filter expansion. Users never leave empty-handed.

Search UX that feels like part of your product

Autocomplete, typo tolerance, facets, highlighted matches, grouped results and no-result recovery — not backend features, but the interface users see.

nkie air max — corrected to: nike air max38ms
Filters
Suggestions:nike air max 270nike air max 90nike air max 97

Showing results for nike air max — typo corrected automatically

Nike Air Max 270 React

0.97

Men's running shoe. Breathable mesh, Max Air unit in heel. Available in 8 colors.

Sneakers

Nike Air Max 90 Essential

0.91

Classic lifestyle sneaker. Leather and mesh upper, visible Air cushioning.

Sneakers

Nike Air Max 97 Silver Bullet

0.88

Iconic full-length Air sole. Rippled design inspired by bullet train.

Sneakers
152 results38ms · exact + fuzzy match

Category

Sneakers (124)
Running (89)
Casual (56)

Brand

Nike (78)
Adidas (54)
New Balance (32)

Price

Under $50 (23)
$50–$150 (87)
$150+ (42)

Autocomplete + suggestions

Suggestions from the first character. Keyboard navigation, recent searches, pinned suggestions and popular queries.

Typo tolerance

AACsearch corrects 1–2 mistakes per word, understands transliteration and does not break exact matches.

Faceted filters

Filters by brand, price, category, availability and any custom attributes — with result counts for each value.

Highlighted matches

Matching fragments are highlighted in results. Users immediately see why this document was returned.

Grouped results

Multi-search across collections: products separately, articles separately, orders separately — all in one request.

No-result recovery

Empty result → suggestions, similar queries, automatic filter expansion. Users don't leave.

What's in Search OS

Not a collection of separate APIs — a complete layer: access, index, relevance, analytics, reindex and multi-tenant security.

Scoped tokens

Issue short-lived tokens scoped to user, index and filter set. Browser gets only a narrow search right — not a permanent API key.

Origin allow-list

Pin API keys to trusted origins. Even if a token leaks outside your domain, the request is rejected.

Rate limit & quota

Per-key rate limit, monthly org quotas and pre-limit alerts. Teams see consumption before it blocks.

Multi-search

Several searches in one round-trip: autocomplete, cross-collection search, grouped results and federated search without a cascade of requests.

Zero-downtime reindex

Rebuild an index and change the schema while live traffic keeps hitting the old version. Atomically switch aliases — no downtime.

Enterprise controls

SSO, audit log, IP allow-list, dedicated cluster, SLA and DPA for teams with control and compliance requirements.

From search to discovery

AACSearch OS does not wait for the perfect query. It helps form intent, corrects mistakes, suggests similar objects and recommends the next best result.

01

Suggestions Engine

Helps the user reach the right query.

  • Autocomplete
  • Query completions
  • Recent searches
  • Pinned suggestions
  • Typo suggestions
  • Facet suggestions
  • No-result suggestions
02

Recommendation Engine

Helps the user choose the next object.

  • Similar items
  • Related documents
  • Frequently searched together
  • Trending queries
  • Session-based recommendations
  • Personalized recommendations
  • Tenant-specific recommendations

Suggestions help formulate the query. Recommendations help choose the next object.

GraphRAG for questions where connections matter

Regular RAG finds similar fragments. GraphRAG uses connections between entities: documents, products, users, errors, incidents, services and owners.

Entity extraction

AACSearch OS extracts entities: users, organizations, errors, docs, incidents, SKUs, policies and services.

Relationship graph

Builds connections: owns, depends on, caused by, similar to, mentioned in, purchased with.

Answers with citations

Answers are built on top of sources and the relationship graph so users can verify context.

Example queries
  • What billing changes affected Enterprise customers?
  • Which incidents are linked to checkout?
  • Which documents explain error 401 in the SDK?
  • Who owns the service linked to the last incident?
Knowledge graph — billing service
Service: billing
→ incidents (3)
→ owners (2)
→ docs (12)
→ API methods (8)
→ customers affected (147)

Relevance must be manageable

Speed matters, but product search wins on result quality. AACsearch gives you a control plane for ranking, synonyms, rules and experiments.

Ranking management

Field weights, exact-match boost, prefix search and partial match. You define which fields and match types influence ranking most.

Synonyms & stop words

Connect "laptop" and "notebook", "sneakers" and "trainers", internal abbreviations and user terms.

Business rules

Pin results, boost priority categories and manage promotional listings without reindexing.

Personalization

Scoped relevance by roles, groups, or tenants. Each tenant sees their own results.

A/B testing

Compare two sets of relevance rules on live traffic. Metrics: CTR, zero-result rate, conversion.

Search analytics shows what your product is missing

Every query is a signal. AACsearch shows what users search for, where they find nothing, which filters they use and which queries lead to conversion.

Search Analytics

Last 7 days

Live

Total searches

128.4K

+18.2%

Zero results

3.2K

-6.4%

CTR rank #1

24.6%

+7.1%

Conversions

8.7%

+3.3%

Search volume

MonTueWedThuFriSatSun

Top queries

nike air max 270

4,812 38%

wireless headphones

3,540 29%

macbook pro m3

2,910 22%

standing desk

2,280 31%

coffee maker

1,760 18%

Top queries

Most frequent queries show which objects, products and documents really matter to users.

Zero-result queries

Queries without results help find gaps in catalog, documentation, synonyms or filters.

CTR by position

What percentage of clicks does position 1 get? Position 2? Where does engagement drop?

Filter usage

Which facets are popular? Which filters go unused? Simplify UX based on data.

Post-search conversion

How many search sessions led to a target action — purchase, view, form submission.

Per-key consumption

Per-key and per-tenant usage analytics. Who's exceeding limits, where are anomalies.

How AACsearch works

Three paths, one platform. Your backend, your users, and your team — all connected to one search layer.

Ingest path

Your backend

Admin API

Bearer ss_connector_*

Ingest queue

DB buffer

Search index

Search path

Browser / app

Scoped token

HMAC-signed

Search API

Results

Manage

Dashboard

Analytics

top queries, CTR

Keys & quotas

Zero-downtime reindex via atomic alias swap

Every query scoped to your organization — no cross-tenant reads

Integration without a separate search project

Create a collection, push documents, issue a scoped token and run your first search from browser or server.

import { AdminClient } from '@aacsearch/client';

const admin = new AdminClient({
  baseUrl: 'https://app.aacsearch.com',
  apiKey:  'ss_admin_••••••••••••', // server-side only
});

await admin.indexes.create({
  slug: 'products',
  fields: [
    { name: 'title', type: 'string' },
    { name: 'price', type: 'float'  },
    { name: 'brand', type: 'string', facet: true },
  ],
});

View full documentation

GraphRAG for questions where relationships matter

Standard RAG retrieves similar chunks. GraphRAG uses relationships between entities to answer questions about causes, dependencies, owners and processes — with cited sources.

Keyword + vector hybrid search

AACSearch OS combines exact matches with semantic similarity to rank results by both keyword relevance and vector proximity.

AI answers with cited sources

Answer comes with source links so users can verify context and open the original document.

GraphRAG pipeline

AACSearch OS builds a knowledge graph over documents, products, users and events. GraphRAG traverses entity relationships to answer questions where context matters.

Entity graph and relationships

AACSearch OS extracts entities — products, users, errors, incidents, docs — and connects them with typed relationships: owns, depends on, caused by, similar to, purchased with.

Security built into the search layer

Search from the browser must not expose permanent keys. AACsearch uses scoped tokens, origin allow-list and tenant filters so every request is bound to the user's context.

Scoped token flow

Zero API key exposure

Browser / App

No API key exposed

Scoped Token

HMAC · TTL: 1h

Search API

Origin verified

Tenant Filter

tenant: acme

Your Index

Results returned

API key never sent to browser
Each token scoped to one tenant
Expired tokens auto-rejected
Origin allowlist enforced
Leaked token still harmless

Token payload

{
  "org": "acme",
  "ttl": 3600,
  "origin": "shop.acme.com",
  "filter": "tenant:acme",
  "scope": ["search"]
}
Request allowed

Short-lived scoped tokens

HMAC-signed tokens bound to an index, origin, and TTL. A leaked token only accesses one index for one hour.

Tenant isolation

Every search query is scoped to your organization ID. Cross-tenant reads are architecturally impossible.

Origin restriction

Pin every API key to specific origins. A browser-console leaked key won't work from another domain.

SSO + Audit log

SAML/OIDC for enterprise. Every administrative action is logged with timestamp and actor.

Compliance & DPA

SOC 2, encryption at rest and in transit, dedicated cluster regions. Signed DPA for GDPR/CCPA.

99.95% SLA

Contractually guaranteed uptime. Incident response, backups, restore procedures.

Pricing

Search OS plans

Not just a search API — a full operating layer. Pick the plan that fits your product stage.

Free

For prototypes, test indexes and first integrations.

$0/mo
Start free
  • 10K search-units / month
  • 1 index
  • 3 API keys
  • Community support

Pro

For production products that need fast search, scoped access and analytics.

$29/mo
Start free
  • 1M search-units / month
  • 20 indexes
  • Unlimited keys & scoped tokens
  • Custom origin allow-list
  • Email support, 1 business day

Enterprise

For teams with SSO, audit log, dedicated region, SLA and contractual requirements.

Custom
Contact sales
  • Unmetered search-units
  • SSO (SAML / OIDC)
  • Audit log + IP allow-list
  • Dedicated cluster region
  • 99.95% SLA, signed DPA

Pricing FAQ — unit economics explained

1 search request OR 1 indexed document. A document update = 1 unit. Reindex is not billed — only the final write to the new index.
Soft cap: you receive a dashboard warning and email notification. Hard cap: search is blocked until the next billing cycle. No surprise bills.
Algolia charges for every record every month. At 1M searches + 10K documents, Algolia costs ~$5,500/mo — AACsearch costs $100/mo. At 10M searches, the difference reaches 50x.
10,000 search-units/month, 1 index, 3 API keys. Enough for a prototype, MVP, or small side project.
Zero-downtime reindex out of the box. Export your index → import into AACsearch → switch the alias. All online, no search interruption.

SDK, API and contracts without magic

TypeScript SDK, REST API, OpenAPI, webhooks and React components work on top of the same stable contracts.

TypeScript SDK

Typed SDK for browser and server-side scenarios. Type autocomplete, predictable errors and ready-made helpers.

REST API

HTTP API with stable contracts for indexing, search, tokens, keys and analytics.

Webhooks

Indexing events, sync errors, quota breaches — subscribe to real-time changes.

Next.js / React

SearchProvider, UseSearch hook, ready-made components for React and Next.js App Router.

Migration

Zero-downtime reindex out of the box. Migrate from Algolia, Meilisearch, or Elasticsearch — physical index dump.

Playground

Interactive console in the dashboard. Try queries, see results, copy ready-to-use code.

Launch Search OS in one evening

Create an index, add documents and connect search from your app. Free tier covers a prototype — data persists when you upgrade.