Turing for Adobe Experience Manager
AEM search, reimagined with AI — without the ceiling
The AEM-native pitch is “Cloud-Service-native.” That's also its limit. Viglet Turing ES gives AEM the same cited, streaming AI answers — then goes deeper: it searches beyond AEM, adds agents, tools, skills and MCP, runs on any LLM and any search engine, and keeps every page and query in your own infrastructure. Open source, Apache 2.0.
Live on real AEM content
Search and ask Adobe's WKND site
This is the Adobe WKND reference site — its pages and assets indexed live via the Dumont AEM connector. Search it with facets and typo tolerance, or ask a question and get a grounded, cited answer. Every result and citation links back to the public WKND site. Same SDK, same corpus — two lenses.
Home to the best right-handers in Bali — big-wave surfing at Nusa Dua for advanced surfers.
A craft-beer tour through Oregon's Willamette Valley, from taprooms to hop farms.
A guided multi-day trek across granite domes, alpine meadows and giant sequoias.
Backcountry ski touring across the Chamonix valley beneath Western Europe's highest peak.
A weekend of cellar-door tastings and vineyard tours through California wine country.
Chasing cold-water swells inside the Arctic Circle — wetsuits, aurora and empty line-ups.
Coast-to-outback across WA: quokkas, karri forests and the Coral Coast.
The bowls, ledges and DIY spots that shaped Southern California skate culture.
A collective of outdoors, music, crafts, adventure-sports and travel enthusiasts sharing experiences.
Browse curated surf, ski, cycling and camping adventures around the world.
How it plugs in
Sits alongside AEM — no re-platforming
AEM stays your content source of truth. Turing connects, indexes into the engine you choose, and drops a cited-answer UI wherever your users are.
- 1
Connect your AEM tier
Point the Turing AEM connector at your author or publish instance. It crawls pages, components and DAM assets — extracting text from PDFs and Office files — into a Semantic Navigation site. Works with AEM 6.5+ and AEM as a Cloud Service; no code changes to your templates.
- 2
Index into the engine you choose
Content lands in Apache Solr, Elasticsearch or an embedded Lucene index with vector embeddings — not locked to AEM's Oak repository. Fields are described as code in a versioned manifest, and hybrid RRF fuses keyword and semantic relevance for every query.
- 3
Drop the UI anywhere
Mount cited AI search and chat into an AEM page component, an Adobe Edge Delivery Services (EDS) block, or your own SPA — with the zero-dependency vanilla-JS SDK or the React SDK. Answers stream token-by-token over SSE with inline citations back to the source page.
- 4
Turn search into an agent
Go past answers: configure an agent that calls your tools, captures form slots, runs Anthropic-standard skills in a Docker sandbox, federates over MCP and uses each LLM's native tools — the depth an AEM-only add-on can't reach.
<!-- Vanilla, zero-dependency: AEM or EDS block -->
<div id="turing-search"></div>
<script type="module">
import {
createTuringClient,
createSearchController,
createChatController,
} from "https://esm.sh/@viglet/turing-sdk";
// Live demo: Adobe's WKND site, indexed from AEM.
// Swap baseURL for your own instance — data stays put.
const client = createTuringClient({
baseURL: "https://turing-demo.viglet.org/api",
});
// Faceted, typo-tolerant search over WKND content…
const search = createSearchController(client, {
site: "wknd-publish",
locale: "en_US",
});
// …and a grounded, cited chat answer streamed over SSE.
const chat = createChatController(client, {
site: "wknd-publish",
});
search.subscribe((s) => render("#turing-search", s));
search.query("surf camps in Bali");
</script>Everything an AEM AI search does — and the parts it can't
Turing matches a Cloud-Service-native add-on on the core AI-answer story. The depth gap opens on everything around it.
| Viglet Turing ES | AEM-native AI add-on | Adobe native search | |
|---|---|---|---|
| AI-generated answers with citations | |||
| Vector / semantic search | |||
| Real-time SSE streaming | |||
| Faceted, typo-tolerant search | partial | partial | |
| Indexes AEM 6.5 + AEM as a Cloud Service | connector | native | |
| PDF & DAM asset text extraction | partial | ||
| Searches beyond AEM (CMS, DBs, files) | |||
| AI agents, tools & skills | |||
| MCP server & client | |||
| Native provider tools (web search, code, computer use) | |||
| LLM providers | OpenAI · Anthropic · Gemini · Azure · Ollama | OpenAI · Anthropic | — |
| Bring your own search engine | Solr · ES · Lucene | no (Oak) | no (Oak) |
| Pluggable reranker (LLM / cross-encoder / Cohere) | |||
| Groundedness check & honest refusal | partial | ||
| Open source (Apache 2.0) | |||
| Where content & queries live | your infra | Adobe cloud | Adobe cloud |
| Pricing model | free / self-host | license + consulting | bundled w/ AEM |
| Time to first result | minutes | 2–4 week engagement | bundled |
High-level comparison as of 2026 — verify current vendor capabilities. “AEM-native AI add-on” = a Cloud-Service-native AI search product; “Adobe native search” = built-in Query Builder / Omnisearch.
We have more depth
Where an AEM-only add-on reaches its ceiling, Turing keeps going.
One index across every repository
AEM is rarely your only source. Turing federates AEM pages and DAM assets with WordPress, databases, file shares and any REST source into a single index — so one answer can cite a policy PDF, a product record and an AEM page together. An AEM-native tool stops at the AEM boundary.
Agents, not just answers
Beyond RAG, configure agents that call your tools, capture form slots, run skills in a sandbox and hand off to a human. An AEM search box becomes an AI concierge that can actually do things — book, look up, escalate — not only summarize a page.
Any LLM, any engine — your choice
Not two providers, but OpenAI, Anthropic, Gemini, Azure OpenAI and local Ollama — running on Solr, Elasticsearch or Lucene. Switch per agent, store keys encrypted, and pin nothing to a single vendor's cloud.
Your content stays in your infrastructure
A Cloud-Service-native add-on runs — and reasons over your content — inside Adobe's cloud. Turing is self-hosted under Apache 2.0: pages, embeddings and query logs never leave your network, and you can run fully local models to keep even the LLM in-house.
MCP + native provider tools
Expose AEM search as an MCP server for any AI client, federate out to other MCP servers, and switch on each provider's server-side web search, code execution and computer use — opt-in per agent, wired into the same cited-answer UI.
Answers a governed brand can trust
A pluggable reranker, a configurable relevance gate, an optional groundedness check and a deterministic “not in this site” refusal keep a regulated AEM property from shipping confident hallucinations — auditable end to end.
Measure it in GA4 / GTM — zero config on EDS
The search and cited-chat UI emit a vendor-neutral analytics event for every funnel step — query, result click, chat, conversion, abandonment. The GA4 / Google Tag Manager sink auto-detects the gtag / dataLayer your AEM pages or Edge Delivery Services blocks already load, so the funnel shows up in the analytics your marketing team already lives in — with a Segment / Matomo / Plausible escape hatch, and nothing phoned home to a vendor.
AEM teams ask us
The objections worth raising before you add AI search to AEM — answered straight.
- What is AEM search?
- AEM search is how visitors and employees find content managed in Adobe Experience Manager — pages, components and DAM assets. Adobe's built-in Query Builder / Omnisearch handles basic keyword lookups; for typo-tolerant, faceted and AI-powered search (semantic results and cited answers) you add a dedicated search layer. Viglet Turing ES indexes your AEM content and serves exactly that, without moving off AEM.
- How do I add AI search to AEM?
- Point the Turing AEM connector at your author or publish instance — it crawls pages and DAM assets (including PDFs) into a search index on Solr, Elasticsearch or embedded Lucene. Then drop the search + cited-chat UI into an AEM component, an Edge Delivery Services (EDS) block or your own SPA with the zero-dependency JavaScript SDK. It works with AEM 6.5 and AEM as a Cloud Service, needs no template changes, and first results come back in minutes — no multi-week engagement.
- Do I have to move off AEM?
- No. Turing sits alongside AEM. Its connector indexes AEM 6.5+ and AEM as a Cloud Service — pages, components and DAM assets — without touching your templates or authoring workflow. AEM stays your content source of truth; Turing adds the AI search and agent layer on top.
- Isn't it limited to AEM content, like the AEM-native tools?
- The opposite. An AEM-native add-on stops at the AEM repository. Turing federates AEM with WordPress, databases, file shares and any REST source into one index, so a single cited answer can span your whole content estate — not just what lives in Oak.
- Where do my content and my users' queries go?
- Into your own infrastructure. Turing is self-hosted under Apache 2.0 — indexed AEM content, embeddings and query logs stay on your hardware. A Cloud-Service-native add-on reasons over your content inside Adobe's cloud; Turing doesn't. Run local Ollama models and even the LLM stays in-house.
- Which LLMs can it use — just OpenAI and Anthropic?
- More than that. OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, local Ollama and any OpenAI-compatible endpoint plug in behind one interface, switchable per agent, with encrypted keys. Reranking and embeddings are pluggable too.
- Do I need a multi-week consulting engagement to install it?
- No. It's open source with a zero-dependency CLI: scaffold a project, run a local stack and point the connector at AEM in one session. First searches come back in minutes, not a 2–4 week paid implementation.
- Can I drop it into AEM pages or Edge Delivery Services?
- Yes. The vanilla-JS SDK is zero-dependency and framework-agnostic — it runs in an AEM component, a plain <script> tag or an Adobe EDS block (example blocks ship in the repo). There's also a React SDK with headless hooks and components for search, facets, autocomplete and cited chat.
- Does it handle PDFs and DAM assets?
- Yes. The connector extracts text from PDFs and Office documents in the DAM and indexes them alongside pages, so answers can cite an asset the same way they cite a page.
Add AI search to AEM — on your terms
Self-host under Apache 2.0, connect your AEM tier in a session, bring your own LLM and search engine, and keep every page and query in your infrastructure.