Skip to content

dialogysx

Renault / Dacia parts catalogue and repair documentation

GitHub · dialogysx.emdzej.pl

About

A reimplementation of Dialogys 7.5.6, reverse-engineered from the shipped Java application. It reads the original data client-side with no backend, and ships without data — bring a disc, or point it at your own static tree.

No warranty

The catalogue data is Renault/Dacia's, is not redistributed here, and nothing in the repository is derived from it. Part applicability is reconstructed and not yet verified against a real vehicle.

Status

The storage format is cracked and validated, and the app works. dialogysx verify walks every index on a real tree:

SetDatasetsKeysFailures
Single language12576,0340
Multi-language 7.5.632583,0350

Record lengths, key ordering and pointer bounds all check out. The applicability condition grammar parses all 41,758 plates with every byte consumed — 423,076 callouts, 762,244 part candidates, three-valued evaluation.

Measured over localhost: a 7.2 MB index preloads in 119 ms, a part-number lookup costs 12 ms, and a depth-3 envelope query returning 18 records costs 21 ms — with the 7.2 MB data file never downloaded.

Working today

  • Browser client — brand → model → vehicle → assembly → plate, with the drawing, clickable callout hotspots, part names, and a parts table filtered by applicability
  • Repair documentation — 2,131 English workshop manuals and technical notes, navigated by topic and filtered to the vehicle
  • Import in the browser — mount an ISO, point at it, repeat; the wizard builds a tree with no Node and no command line, sharing its planner with the CLI
  • dialogysx CLIimport, verify, plates, organes, docs, datasets, keys, get, criteria

Packages

PackageWhat it is
@dialogysx/rafThe storage engine — HTTP Range, a local directory, or Node fs, behind one read(pos, len)
@dialogysx/catalogueVehicle envelope, part search, drawing callouts, criteria vocabulary, date resolution, doc navigation
@dialogysx/importerDisc classification, import planning, and a zip reader that works in a browser as well as in Node

Not yet claimed

That any parts list is correct. 41,758 records prove the grammar's shape and every date view resolves, but that does not prove "this part fits this car". It needs one vehicle whose right answer is known independently.

Stack: TypeScript, Svelte, pnpm + Turborepo