Skip to content
Vessios Docs
English
Esc
navigateopen⌘Jpreview
On this page

Introduction

Vessios is a wiki that writes itself — knowledge that compounds with every ingest and query.

Vessios is a knowledge management platform built on the Karpathy LLM Wiki pattern: instead of treating every question as a fresh, stateless retrieval problem, Vessios maintains a living wiki that grows and improves with use.

Standard RAG vs. Vessios

Standard RAG Vessios
Knowledge handling Re-derived from raw documents on every query (stateless) Accumulated and compounded in a wiki (stateful)
Role of the LLM Synthesizes an answer from scratch each time Reads and extends an existing, curated wiki

Three core operations

Vessios is built around three operations that work together to keep the wiki accurate and useful over time.

Why Lint matters

The biggest risk in any system that accumulates knowledge automatically is compounding hallucination: an error introduced during one ingest can be picked up and reinforced by later ingests and queries. Vessios treats Lint as a first-class, architecturally required component — run at least weekly, and immediately after any large ingest — specifically to contain this risk.

What’s in these docs

  • Quickstart — create a tenant, ingest your first source, and ask your first question.
  • Source ingestion — supported file types, URL ingest, and the safety checks every ingest goes through.
  • Search & query — how the four-layer retrieval pipeline answers questions.
  • Automatic maintenance — contradiction detection, link audits, and confidence decay.
  • Integrations — API keys, webhooks, and connectors.
  • Security — tenant isolation, PII handling, and other safeguards.
  • Plans & billing — quotas, pricing, and what each plan includes.
  • FAQ — quick answers to common questions.
  • API reference — the full REST API, generated from the live OpenAPI spec.

Was this page helpful?