# RAG World > A self-updating registry of retrieval-augmented generation technologies. Every record carries a configuration over a 28-dimension schema, a maturity level from L0 to L6 derived by a deterministic rule with no language model involved, and the evidence that level stands on. Data is collected weekly from arXiv, OpenAlex, GitHub, PyPI and Papers with Code. Do not scrape the pages. Every page on this site is rendered from the JSON files below, and those files are the same data without the markup. Start from https://ragworld.org/data/index.json, which lists them all. - Records: 76 - Evidence records: 771 - Freshest evidence: 2026-10-19 - Built: 2026-10-19 - Licence: CC BY 4.0, https://creativecommons.org/licenses/by/4.0/ - Attribution: RAG World, https://ragworld.org ## Data - [Index of all datasets](https://ragworld.org/data/index.json): what each file holds, how many records, where the schema is described. - [Registry](https://ragworld.org/data/registry.json): Every technology in the registry with its configuration over the 28 dimensions, maturity level, the rule output behind that level, and the evidence records the level stands on. - [Map](https://ragworld.org/data/map.json): Maturity level against attention for every record, as plotted on the maturity map. - [Changes](https://ragworld.org/data/changes.json): Append-only chronicle: every level change with its date and the evidence that caused it. - [Stats](https://ragworld.org/data/stats.json): Counts by kind, level, and stratum, plus the date of the freshest evidence. - [Residuals](https://ragworld.org/data/residuals.json): Mechanisms recorded as not expressible in the current schema, with the records that raised each one. Three mentions make a mechanism a candidate for a new dimension. - [Candidates](https://ragworld.org/data/candidates.json): Works found by weekly discovery and awaiting a human verdict, each with a deterministic fit score from 0 to 10. - [Digest](https://ragworld.org/data/digest.json): Digest issues, generated from the chronicle without a language model. ## Following changes - [Feed, English](https://ragworld.org/data/feed.xml): digest issues and level changes, RSS. - [Feed, Russian](https://ragworld.org/data/feed.ru.xml): the same in Russian. A feed declares one language for the whole channel, so the two languages are two feeds rather than two fields. ## Citing - [Releases](https://ragworld.org/data/releases/index.json): dated, immutable snapshots. Cite a release, not the live file: the live file changes every week. - [How to cite](https://ragworld.org/about): citation formats and the licence. ## Source - [Repository](https://github.com/laputski/rag-world): the data lives in git as one JSON file per technology, plus append-only evidence and level journals. Cloning gives the full history; the artefacts above are derived from it.