Evnao documentation
Product setup, API access, and recovery paths.
Core concept
Knowledge is stored in PostgreSQL, split into tenant-scoped chunks, indexed for full-text search, embedded, and retrieved through a hybrid lexical/vector path.
Publish articles or upload PDF, DOCX, XLSX, TXT, Markdown, CSV, JSON, and HTML files. PDFs can be up to 100 MB and are read on your device before you import the extracted text. Other formats can be up to 10 MB. Each import supports up to 1,000,000 characters of readable text; scanned PDFs need text recognition (OCR) first.
Import public same-origin pages with SSRF, DNS, redirect, size, and timeout controls. Private and authenticated pages are not crawled.
Published chunks are ranked from full-text and vector candidates. Retrieval selects evidence; it does not make model output infallible.
Pro workspace workflow
A workspace owner can choose connected store data, business social content, public website pages, supported files, and existing Evnao knowledge. Evnao estimates the number of AI responses first, then runs the confirmed build as a durable background job.
Store permissions are read-only and scoped. Evnao records when business data is accessed, never asks for database credentials, and lets the owner revoke the connection.
Generated questions and answers are stored separately as drafts. Conflicting facts are marked for review. Nothing becomes live knowledge until the owner publishes it.
Source retrieval, local parsing, and unchanged cached analysis do not consume AI responses. Each real model request consumes one response; failed calls are released and idempotent retries are not charged twice.
GET /api/v1/knowledge lists articles with cursor pagination and an optional state filter, and requires knowledge:read. POST /api/v1/knowledge creates a DRAFT or PUBLISHED article and requires knowledge:write plus an Idempotency-Key. The default state is draft. File, website, and multi-source builds remain dashboard workflows; review source status and generated drafts there.