Niamoto
Français

Roadmap

What's shipping, what's next, what's on the horizon.

Niamoto Roadmap

Last updated: 2026-04-20 — reviewed at each minor release.

Vision

Niamoto is a generic ecological data platform. It turns heterogeneous datasets into publishable web portals through a configurable Import → Transform → Export pipeline, running locally with no cloud dependency. The desktop app (currently Tauri, under evaluation) and the CLI share the same engine and the same plugins.

Recently shipped

Highlights from the last ~3 months:

  • macOS signing & notarization pipeline(v0.15.5, April 2026)
  • In-app feedback system
  • Rich multi-source enrichmentGBIF, CoL, iNaturalist, BHL, GN TaxRef, Tropicos, spatial v1
  • Sources dashboard & mission control redesign
  • Transform & export parallelization
  • Frontend architecture refactorto src/app, src/features, src/shared
  • Release automationvia the niamoto-release skill

GBIF Ebbe Nielsen Challenge 2026 ⭐

Deadline: 26 June 2026.

Pitch — Niamoto: Local-First Intelligence. Combine local intelligent curation (classical ML + fuzzy matching + optional SLM) with automatic generation of web portals from GBIF data, with no cloud dependency. The differentiating angle from 2024–2025 winners (who rely on remote LLMs).

Core deliverables activated for the challenge:

  1. GBIF rich enrichmenttaxonomic and spatial enrichment via GBIF APIs
  2. GBIF registry publicationpublish Niamoto as a referenced tool
  3. Challenge presentation pagededicated landing for submission

Optional bonuses (time-permitting):

  • BDQ validation (12 Tier-1 tests — covers ~60% of real data-quality issues)
  • Local schema detection for Darwin Core imports
  • Optional local SLM (Liquid AI LFM2 or Qwen3 via Ollama)
  • GBIF portal auto-configuration, powered by the generic transform/export auto-config described in Now → Configuration intelligence

References:

Now (April – June 2026)

In flight or imminent. Excludes GBIF-specific deliverables listed above.

Desktop & distribution

  • Desktop update harness & auto-updater
  • Binary-size audit & reduction
  • Desktop shell evaluation — Tauri vs Electrondecision-grade analysis driven by Linux stability issues, cross-platform rendering consistency, and packaging size. Informs whether the shell stays on Tauri or migrates to Electron.

Documentation

  • Desktop-first documentation overhaul — user guide, in-app docs, public docs, team/partners page

Institutional site

  • Landing page refresh + hybrid teaser video

UI polish

  • Enrichment tab UX redesign
  • UI density compaction & rendering smoothness

Configuration intelligence

  • Auto-config for collection transform/export pairsextend the auto-configuration approach already available for imports and for index pages to the transform/export pairs inside each collection. Fully generic across datasets, taxonomies, and domain values. Feeds directly into the GBIF portal auto-configuration used for the challenge demo, but remains a general capability for any data source.

Soon (Summer 2026)

Planned but not yet started.

  • Niamoto Doctorunified diagnostics (CLI + GUI)
  • Starter project templatesniamoto init --template with real starter kits
  • Export contract packJSON schema on the export side
  • ML model regeneration pipeline
  • Transform parallelization phase 2

Later (H2 2026 and beyond)

Identified in ideation, not yet planned.

  • Example & fixture certification pipelinedocs and fixtures as executable contracts
  • Suggestion explainability layerattached evidence for every auto-suggestion (matched fields, confidence band, override paths)
  • Transform provenance explorerdependency graph across import → transform → export
  • Desktop v1.0iteration round after the GBIF submission
  • Plugin platform overhaulthree related workstreams: (a) a real plugin marketplace for discovering, installing, and updating community plugins, (b) R-language plugin support alongside the current Python plugins, and (c) an in-app plugin creator that walks users through scaffolding a new plugin without leaving the desktop app.
  • Niamoto MCP serverexpose Niamoto's capabilities (data analysis, entity registry, import/transform/export configuration, plugin scaffolding) through a Model Context Protocol server so an AI agent can configure a complete instance end-to-end: inspect the user's raw data, propose a project structure, wire imports and transforms, generate portals, and — when needed — create a dedicated plugin for that instance. Builds on the plugin platform overhaul above. Secondary benefit: MCP tool-use traces become a training and evaluation harness for the local SLMs explored in the GBIF track — structured agent traces feed fine-tuning, ground-truth configurations anchor benchmarks, and each tool doubles as a verifiable eval task. This makes the local-first intelligence direction self-reinforcing.
  • Hosted Niamoto (exploratory)optional dynamic deployment to serve the Niamoto API (and, if needed, the generated pages) online. Scenarios being considered: self-hosted multi-instance portals, optional hosting on niamoto.org for users who want it, publication of static sites to niamoto.org with sync back to the local instance, and a reproducible self-host setup for teams that prefer their own infrastructure. Large scope, exploratory — the desktop app and static publication remain the primary path.

Not planned

These directions are not on the roadmap. Not hard "nevers", but not actively planned:

  • Native mobile appgenerated portals are responsive; no iOS/Android client planned.
  • Real-time multi-user collaborationoutside the "one analyst, one instance" model.
  • Replacing DuckDBDuckDB remains the core engine.

How to contribute

  • Read CONTRIBUTING.md.
  • Open an issue or a discussion before large changes.