VATES is an embeddable B2B AI infrastructure, built on the Celtic architecture with three roles: bard for cross-language reception, druid for knowledge management, and vates answering in the questioner's native language.

Problem to be solved: Traditional RAG systems compress meaning into vector coordinates as soon as they are indexed. Meaning is frozen before a question arises, requiring re-embedding and re-indexing with every update, and no one can explain why a piece of text is retrieved. VATES maintains knowledge in a structured text format and only interprets it at the time of observation.

Main features:

1. No vector retrieval, no embedding pipeline, no re-ranking. No need for Pinecone, Weaviate or pgvector, no need for GPU.

2. Instant updates — modifying a line in ES takes effect immediately, no re-indexing required.

3. A single knowledge base responds in the reader's language, without storing any translation data.

4. White box — the content is plain text readable by you, not numerical vectors.

5. Remote MCP server with OAuth 2.1 for AI agents. Idempotent by design.

Full Vietnamese support: The introduction page, admin dashboard, and all operation guides are available in Vietnamese — not just the interface, but also detailed documentation for each screen (https://vates.standout.jp/vi/legal/manual.html). VATES supports 23 languages with 9 writing systems, and Vietnamese has been included from the start. The responding language is not limited: a knowledge base written in any language can still respond in Vietnamese.

Target audience: Developers wishing to embed conversational AI into their products via API — much like integrating Stripe or Twilio. And small to medium enterprises that need an immediately usable customer support widget.

How to get started: Web platform. Embed the widget with a line of JavaScript, or use it as a standalone application via URL, or call the API. The entire dashboard operates on mobile. Pay based on usage, with a minimum load of 5 USD, no monthly fees, no subscriptions.

Difference compared to existing products: Other chatbot platforms are all based on vectors. VATES is based on our proprietary ES-IFM theory — meaning is not pre-fixed but emerges at the moment of observation, in the observing intelligence. This is not an optimization of RAG, but another principle.