Run Turing on your own infrastructure under Apache 2.0, no vendor lock-in for
your data, your models or your search engine. Install it, point it at any LLM
provider (OpenAI, Anthropic, Gemini, Azure OpenAI or local Ollama), and configure
per-instance behavior; API keys are stored encrypted. Scale to multiple isolated
tenants behind a single identity realm when you need to, or keep a single-tenant
deployment simple.
Operating Turing in production is first-class: Micrometer metrics, conversation and
chat analytics, cache-hit gauges, token usage accounting and cost governance
give you the visibility to run it confidently. This section covers installation, the
full configuration reference, provider and instance setup, assets, multi-tenancy,
administration, logging and observability.
Configuration ReferenceComplete reference for the Viglet Turing application.yaml configuration file.
Generative AI & LLM ConfigurationOverview of GenAI capabilities in Viglet Turing: LLM providers, RAG architecture, tool calling, MCP servers, and AI agents.
Configure LLM Instances in Viglet TuringConnect Viglet Turing to OpenAI, Anthropic Claude, Gemini, Ollama and 7 other model providers: credentials per vendor, generation parameters, and which features each model supports.
AssetsManage files and train the RAG knowledge base with Viglet Turing Assets.
Multi-TenancyRun one Viglet Turing JVM that serves many fully isolated tenants: discriminator-column data isolation, a single Keycloak realm with a tenant claim, tenant-owned infrastructure, plan quotas, and tenant lifecycle.
Administration GuideManage users, groups, roles, API tokens, global settings, and system diagnostics in Viglet Turing.
Observability: Prometheus & GrafanaThe metrics, dashboards, and traces that turn Viglet Turing from a black box into a glass box. Prometheus scrapes Spring Boot Actuator, Grafana renders the story, and you sleep through the night.
LoggingMonitor server activity, indexing pipeline, and AEM connector logs from the Viglet Turing admin console.
Chat AnalyticsTurn thousands of chat conversations into the voice of your customer. Intent, goal achievement, sentiment trajectory, funnels, conversation replay, tool-latency p95, router-decision and SSE diagnostics: engine-agnostic, cluster-safe, AI-on-AI classified.
Token UsageMonitor and analyse LLM token consumption in Viglet Turing.
Cost GovernanceSee and control your AI spend. Turing freezes a USD cost on every turn from an editable price table, rolls it up into a live dashboard by agent / model / stage, and lets each agent enforce a soft monthly budget that downgrades to a cheaper model instead of overspending.
Governed LLM GatewayPoint any OpenAI-compatible client at Viglet Turing: virtual keys, per-key budgets and rate limits, semantic cache, guardrails, model-name routing into agents and RAG, and native tools.
