---
title: "New documentation"
description: "Guides rebuilt around Avra's models, Avra Core, data ingestion, and deployment, with concept pages, figures, complete examples, a glossary, and an FAQ."
type: changelog
date: 2026-11-09
changelog:
  category: Documentation
---

The guides now follow Avra's products. A new [Models](/models) page compares Avra's two foundation models, and each has its own section: [Avra RFM](/rfm), and [Avra GFM](/gfm), which now holds the [embeddings](/gfm/embeddings) guides. [Avra Core](/core), [Data ingestion](/data-ingestion), and [Deployment](/deployment) follow, then [Use cases](/use-cases), the [Dashboard](/dashboard), [Security & compliance](/security), and a new Resources section. Pages that moved keep working at their old addresses.

## New pages

Two concept pages explain how Avra RFM reads a relational database and what you can predict with it: [Relational data](/rfm/relational-data) and [What you can predict](/rfm/what-you-can-predict). [A worked example](/rfm/worked-example) follows one prediction from the data to the list a team works every week. Data ingestion has pages on [preparing your data](/data-ingestion/preparing-data), [time and history](/data-ingestion/time), and [data contracts and checks](/data-ingestion/data-contracts). Deployment covers [installing the data layer](/deployment/data-layer) in your Kubernetes cluster and [running Avra RFM on AWS SageMaker](/deployment/aws-sagemaker).

Avra Core adds [Limits and quotas](/core/limits), the Dashboard adds [API keys](/dashboard/api-keys) and [Workspaces](/dashboard/workspaces), and Resources has a [Glossary](/resources/glossary) and an [FAQ](/resources/faq).

## Figures and examples

Diagrams across the guides show how the pieces fit together, from tables becoming a graph to what a prediction reads and how a version reaches production. The API guides have complete examples in cURL, Python, and TypeScript, and the [Quickstart](/quickstart) starts with a first API call.

## Corrections

Every page was checked against the product. Batch statuses, webhook payloads, error responses, and limits now match what the API returns, so if you built against an earlier version of these pages, check those parts of your integration.

Two pages are gone because they described things Avra doesn't offer: SFTP batch inference, whose old address now leads to [Batch inference](/core/batch-inference), and the API reference's IP addresses guide, which described a service Avra no longer runs.
