Roadmap
Reviewed: 2026-09
Research direction
Section titled “Research direction”The main question is how to make product data collection more scalable, collaborative, and reusable. R9lab focuses on open, FAIR product-level disassembly data. Automation stays human-verified: software can suggest, people confirm.
The live work list is tracked in the public GitHub issues.
Near term
Section titled “Near term”- improve the data collection guides and keep the rest of the docs current
- strengthen tests, especially around component hierarchies
- simplify CSV and JSON export from live records toward dataset releases
- improve admin and reference-data maintenance workflows
- make camera-assisted capture easier to operate in repeated lab workflows
- re-apply account and upload erasures after a backup restore, so a restored snapshot stays in step with the deletions made since it was taken (added 2026-10-08)
- publish curated, versioned dataset releases separate from the live database, with stable identifiers and documented scope and licensing
- improve API guidance and exports for analysis scripts, reproducible research, and LCA work
- explore human-verified assistance for label reading, component suggestion, and quality control, with provenance records for assisted values
- make reference-data search multilingual and meaning-based: embedding search over categories
and product types next to the current full-text and fuzzy matching (accent-insensitive since
2026-09)
- pin a self-hosted multilingual model (bge-m3 is the first candidate) rather than a hosted one whose output can change under us
- store the model id with each vector, so a model change forces a full re-embed instead of mixing vectors from two models
- explore interoperability with semantic vocabularies and Digital Product Passport identifiers
- first candidate: CEON, the Circular Economy Ontology Network (CC BY 4.0), which separates a product model from a physical item and follows the ISO 59040 Product Circularity Data Sheet; map products and components onto its product and material modules for the JSON-LD export