RAG & Chat
Ground answers in your own content. Turing's retrieval-augmented generation pipeline pulls the most relevant passages, sharpens them with a pluggable reranker (LLM, cross-encoder or Cohere), and streams a cited answer token by token. Every citation traces back to the exact source passage: provider-agnostic provenance built on Turing's own retrieval metadata, not a vendor's citation API.
Answers stay honest by construction: a configurable relevance gate drops weak matches, an optional groundedness check appends a low-confidence caveat to anything it can't support, and a deterministic refusal fires when a question simply isn't covered by your data. Embeddings can use the built-in Lucene vector store or your engine's own vectors, and conversation memory keeps long chats coherent.
What is RAG?
Retrieval-Augmented Generation explained: how LLMs, Embedding Models, and Embedding Stores work together to ground AI responses in real content.
Answer Engine
The /answer lens in Viglet Turing: a streaming, grounded, cited answer over a Semantic Navigation site's knowledge base, with freshness re-checks, interaction signals and shareable permalinks.
Vectorless (Structured-Data) RAG
Answer questions over a structured catalog (products, prices, specs, a model registry) grounded and cited, using only an LLM and your search index. No embeddings, no vector store.
Embedding Stores
Vector database backends for RAG and semantic search in Viglet Turing: ChromaDB, PgVector and Milvus.
Embedding Models
Configure embedding models for vector generation, supported providers, local transformers, model selection and the REST API.
Reranking
The reranking stage in Viglet Turing: a selectable, fail-open second pass (LLM, cross-encoder, Cohere, Voyage, Bedrock, Vertex AI) that sharpens retrieved context before generation.
Chat
The Viglet Turing Chat interface is where AI Agents meet customers. Direct LLM, Semantic Navigation grounded chat, and agent-per-tab, each tuned for a different conversion moment.
Chat Memory
Persist conversation history per agent, surface the most relevant older turns with BM25 retrieval, and compress long histories into a workspace-stored summary.
Intent
Configure intent categories and action prompts for the Viglet Turing chat interface.
Personas
Give your AI Agents a brand voice. Personas turn a generic LLM into the salesperson, the support specialist, the executive advisor your company always wanted to have.
The Persona Book
The complete, didactic, strategic guide to Personas in Viglet Turing, from why a generic AI voice costs you money to Persona Match, Dialogue, and Synthetic User Research, with worked examples for every feature.
Transcription
Set up speech-to-text in Viglet Turing with OpenAI or a self-hosted OpenAI-compatible server, split recordings past the upload limit, and transcribe multi-hour audio as async jobs.