Roadmap (public)¶

v0.1 (implemented public foundation) — additive contextual Material schema, generated JSON Schemas, MP/JARVIS/NOMAD/OPTIMADE plus bounded local CSV/JSONL import, LeMat-Bulk workflow companion, periodic crystal graph, transparent scorecard, ephemeral FastAPI registry, guided demo and local contributor workbench
Workflow layer — richer CandidateSlice.report() output, benchmark-frame presets, more transparent filtering recipes, offline-first analysis helpers
LeMaterial companion — deepen LeMat-Bulk coverage, document LeMat-Traj / LeMat-Synth as future integrations, expand the SQL/EDA cookbook
Connectors — deepen native provider coverage while reusing bounded timeout/retry/Retry-After policy; OQMD remains available through OPTIMADE and native OQMD remains an explicit stub
Schema — evolve additively from Pydantic source models; preserve contextual properties, dataset manifests, source artifacts and simulation-result envelopes
Graphs — More edge/node features, optional DGL/PyG examples while keeping the core library lightweight
Benchmarks — Tighter Matbench / leaderboard examples
Simulation — result-import/parser examples around SimulationResultEnvelope; LAMMPS/QE execution and multi-simulator orchestration remain out of scope
Uncertainty — First-class epistemic/aleatory fields where sources provide them

Seed issues (suggested on GitHub)¶