Enterprise search, cited RAG and AI agents over your own content.
AI search and grounded answers you can trace back to the source, on any LLM, in your own infrastructure. An open, self-hosted alternative to Algolia, with AI built in.
For engineering teams who need production search and RAG without shipping their data to someone else's cloud.
npx @viglet/turing-cli init my-searchTry it: type to search.
Search it, ask it, validate it, on the same content
All three run on Turing's own zero-dependency SDK, over the exact same corpus (Turing's documentation). Search finds the page; the assistant answers with a cited RAG response. You can switch personas to hear the same answer through different eyes, or validate whether your own copy lands for a given reader. Try a result's “Ask about this”.
See the same content through a persona's eyes: a fit score, what lands, and the exact phrases that miss.
The playground runs on this code
The live search and cited-chat playground above run on the product's own zero-dependency @viglet/turing-sdk, grounded in Turing's own documentation. Here's the exact code that drives them.
// The widget on the left runs this code.
import { createTuringClient, createChatController }
from "@viglet/turing-sdk";
const client = createTuringClient({
baseURL: "https://turing-demo.viglet.org/api",
});
const chat = createChatController(client, {
site: "turing-docs",
});
// streams tokens + RAG citations
chat.subscribe((s) => render(s.messages));
chat.send("Does Turing support self-hosting?");- The live search + assistant above call this exact SDK, with no bespoke demo backend
- Zero runtime dependencies: a native fetch client + observable controllers
- The same SDK you embed on your own site (React SDK & vanilla JS)
Search it. Ask it. Automate it. Run it.
Most teams stitch a search engine, a RAG pipeline and an agent framework together. Turing is all three, open source and self-hosted.
Production search over your content
Faceted, multilingual, typo-tolerant search on Solr, Elasticsearch or Lucene, with field manifests as code, currency/number localization and coverage observability.
RAG & chat with answers you can audit
Retrieval-augmented generation with a pluggable reranker (LLM, cross-encoder or Cohere) and an embedded vector store. Answers stream token-by-token with inline citations that trace back to the source passage, plus an optional groundedness check and a deterministic “not in this site” refusal.
Agents with tools, skills & MCP
Configurable AI agents that call your tools, run Anthropic-standard skills in a Docker sandbox and federate across MCP servers. Each provider's own server-side tools (web search, code execution, computer use, image generation) are opt-in per agent.
Self-hosted, any LLM, your data
Open source under Apache 2.0. Bring any LLM (OpenAI, Anthropic, Gemini, Bedrock, Ollama and more), your own search engine and storage. Multi-tenant and observable, on your infrastructure.
Outcomes, not just features
A capability list is easy to skim past. Here's what the platform actually changes for your team and your users.
Users get answers, not a list of links
Grounded, cited answers over your own content cut the time from question to answer. Every claim traces back to the exact source passage.
Fewer repetitive support tickets
A self-service assistant over your docs and knowledge base answers the questions your team answers again and again.
Ship in days, not quarters
Connectors, hybrid search, reranking, agents and SDKs come wired and tested. You configure a product instead of building a RAG stack from scratch.
Answers you can defend
Inline citations, a relevance gate and an optional groundedness check keep a governed brand from shipping confident, wrong answers.
Pluggable LLM providers & search engines
One platform, two ways in
Adopt it bottom-up as a developer, or bring it in as a governed enterprise search layer. The open-source core is the same.
Own your search stack, from npx to prod
Open source and self-hosted, with SDKs that drop into your app.
- Local stack in one command: npx @viglet/turing-cli init
- React SDK + zero-dependency vanilla-JS SDK (headless)
- Bring your own LLM and search engine
- Vendor-neutral funnel analytics to your own GA4, GTM, Segment or Matomo, zero-config
- Self-host with one docker run. No account, no telemetry
Production AI search on your own infrastructure
For regulated, content-heavy organizations that need control over their data.
- Your content and users' queries never leave your network
- Connect Adobe AEM, WordPress, databases and files
- SSO, encrypted secrets, any LLM (including fully local)
- Cited, auditable answers you can trace to the source
How Turing answers a question
Every conversation runs an agentic loop: retrieve, reason, act with tools, and respond with a grounded, cited answer.
Ingest & index
Connectors crawl AEM, WordPress, databases and files into Solr, Elasticsearch or Lucene with vector embeddings.
Retrieve
Hybrid RRF retrieval and reranking pull the passages most relevant to the user's question, across locales.
Reason & act
The agent calls tools, runs skills in a sandbox, queries MCP servers and runs a chat flow.
Respond
A grounded answer streams back with citations, rich content and structured slots your UI can render.
The open-source way to do AI search
A hosted SaaS locks your data in someone else's cloud. A DIY stack is months of glue code. Turing gives you both halves, self-hosted.
| Viglet Turing | Hosted search SaaS | DIY (LangChain + vector DB) | |
|---|---|---|---|
| Open source (Apache 2.0) | partial | ||
| Self-host: data stays in your infra | |||
| Faceted enterprise search | build it | ||
| RAG + pluggable reranker | partial | build it | |
| Cited answers you can audit | partial | build it | |
| AI agents, tools & skills built in | partial | build it | |
| Native provider tools (web search, code, computer use) | build it | ||
| Any LLM provider | partial | ||
| Bring your own search engine | partial | ||
| Time to first result | minutes | minutes | weeks |
High-level comparison as of 2026; verify current vendor capabilities. “DIY” = assembling LangChain/LlamaIndex with a vector database yourself.
$ docker run -p 2700:2700 ghcr.io/openviglet/turing-ce
# pulling ghcr.io/openviglet/turing-ce:latest ...
✓ image pulled no build step
✓ turing-app running
# console + API ready on http://localhost:2700Self-host the whole platform in one command
No account, no sales call, no data leaving your infrastructure. There is nothing to clone or build: the container image is already published to a public registry. Pull it, run it, then open the console and index your first site. Want the full stack (Solr, MariaDB, monitoring)? A readydocker composefile ships in the repo too.
- Prebuilt container image: nothing to clone, nothing to build
- Your content and embeddings never leave your network. No SaaS tier, no telemetry
- Point it at any LLM provider and your own Solr, Elasticsearch or embedded Lucene
docker run -p 2700:2700 ghcr.io/openviglet/turing-ceTurn on the tools your model already has
OpenAI, Anthropic and Gemini ship tools that run in their own infrastructure. Turing wires them into the agent loop (opt-in per agent, no glue code, no extra crawler) and routes their citations into the same source-chip UI as your own search.
Web search & grounding
The model searches the live web (and Google grounding on Gemini) and grounds its answer. Citations flow into the same source-chip UI as your own RAG.
Code execution
A managed Python sandbox runs alongside your own code interpreter for charts, data wrangling and computed answers, rendered inline.
Computer use
A multi-turn screenshot/action loop behind a pluggable driver seam. The model browses and operates a UI to complete a task, with the same contract across all three vendors.
URL / web fetch
The model pulls a URL or PDF named in the prompt directly in the provider's infrastructure, with no Turing crawler in the path.
Image generation
Images are generated inline in the answer as self-contained data URIs, with no separate artifact endpoint to host.
File search
Retrieval over a provider-hosted vector store, available as a built-in tool next to Turing's own search.
Strictly opt-in via a two-level capability gate, so your existing agents stay unchanged. Provider availability as of 2026.3.
Ship agents from your terminal
@viglet/turing-cli is a zero-dependency developer CLI: scaffold a project, run a full local stack, deploy to any environment, and evaluate your agents in CI.
- Scaffold an agent project and run a full local stack with Docker Compose
- Deploy agents, flows, tools and skills to any environment
- Tail live chat events and run YAML eval suites in CI
$ npm i -g @viglet/turing-cli
$ turing init my-copilot # scaffold an agent project
$ turing dev # local stack via Docker
$ turing deploy --env=prod # push agent + flows + tools + skills
$ turing eval # run YAML eval suites
$ turing logs --conversation abc # tail chat eventsStart from a working example
Example sites with search templates. Import a manifest and explore.
Mythical Creatures
A fantasy bestiary search site with dragons, griffins and phoenixes. Faceted search plus a React SPA template.
Faceted · ReactSpace Missions
Historic and fictional space missions with semantic search over mission data, crews and launch details.
Semantic · DataVinyl Records
A vinyl records catalog: browse albums by artist, genre, year and label with a retro template.
Catalog · RetroSDKs, tools & documentation
Everything you need to build, integrate and ship with Turing.
React SDK Storybook
Interactive component docs: headless hooks and UI components with live examples for enterprise search.
Storybook@viglet/turing-react-sdk
React 19 hooks & UI components. Headless architecture with full TypeScript support.
npm@viglet/turing-sdk
Vanilla JS, zero-dependency SDK: a framework-agnostic client and observable controllers for search, chat and autocomplete. Works in Adobe EDS blocks, a plain <script>, or any bundler.
Vanilla JS@viglet/turing-cli
Zero-dependency developer CLI: scaffold a project, run a local stack, deploy agents/flows/tools/skills, and run YAML eval suites.
CLI@viglet/turing-flow-dsl
TypeScript DSL for chat flows. Write flows as typed source and transpile them to the editor's JSON wire format; enum typos (tone, guardrail, trigger) fail at compile time.
DSLDocumentation
Architecture guides, configuration reference, connector setup, API docs and deployment instructions.
DocsOwn your AI search stack
Open source, self-hosted, and works with any LLM. Stand up the full platform in one command and ship cited AI answers over your own content, without your data leaving your infrastructure.