Public distribution point for the Regional Knowledge Packs used by the Botanist Bot mobile app, an off/online plant-identification app.
This repository holds no application source. It exists so packs can be fetched without authentication; the app itself is private.
One read-only SQLite file per region, downloaded on request by the user and never automatically. Each pack carries, for the taxa of one US state:
taxa— scientific name, family, common names, GBIF taxon keytaxon_embeddings— precomputed BioCLIP ViT-B/16 text-tower vectors, fp16, 512-dknowledge— description, care, toxicity, each with its provenanceoccurrences— coarse ~50 km GBIF occurrence gridattributions— the notices reproduced below
Toxicity text appears only where a cited source exists. Where there is none the field is null and the app says so, rather than guessing.
| Region | Pack | Taxa | Size | Release |
|---|---|---|---|---|
| Florida | region-us-fl |
24 | 104 KB | packs-0.1.0 |
| Connecticut | region-us-ct |
24 | 108 KB | packs-0.2.0 |
| Massachusetts | region-us-ma |
24 | 104 KB | packs-0.3.0 |
| New York | region-us-ny |
24 | 104 KB | packs-0.4.0 |
| New Jersey | region-us-nj |
24 | 104 KB | packs-0.5.0 |
| Vermont | region-us-vt |
24 | 104 KB | packs-0.6.0 |
| Georgia | region-us-ga |
24 | 108 KB | packs-0.7.0 |
| Arizona | region-us-az |
24 | 104 KB | packs-0.8.0 |
| California | region-us-ca |
24 | 108 KB | packs-0.9.0 |
Each release asset is verified by SHA-256 on the device before it is installed; a mismatch aborts the install.
Packs contain material from the sources below, and each pack reproduces these
notices in its own attributions table, which the app surfaces under
Settings → Data attributions.
- USDA NRCS PLANTS Database — CC0 1.0 — https://plants.sc.egov.usda.gov/downloads Every Taxon was checked against the relevant USDA NRCS State Plants List.
- GBIF Backbone Taxonomy — CC BY 4.0 — https://doi.org/10.15468/39omei Names, families, and taxon keys were resolved against GBIF.
- Wikipedia — CC BY-SA 4.0 — https://en.wikipedia.org/ Available cited descriptions are English Wikipedia extracts.
The pack files in every release are licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0). The Wikipedia extracts inside them are CC BY-SA, and share-alike is the strictest term among the sources, so it governs the whole file. The other components are compatible with it:
| Component | Source licence | Within the pack |
|---|---|---|
| Taxon names, families, GBIF keys | GBIF Backbone — CC BY 4.0 | CC BY-SA 4.0 |
| USDA symbols and state membership | USDA NRCS PLANTS — CC0 1.0 | CC BY-SA 4.0 |
| Descriptions | English Wikipedia — CC BY-SA 4.0 | CC BY-SA 4.0 |
| Taxon embeddings | Output of BioCLIP (MIT); no licence attaches to model output | CC BY-SA 4.0 |
| Occurrence grid | Derived from GBIF occurrence records; per-record licences (CC0, CC BY, CC BY-NC) are counted in a non-shipping build snapshot | CC BY-SA 4.0 |
If you redistribute a pack, or anything derived from one, keep the attributions
table intact and license the result under CC BY-SA 4.0.
Embeddings are produced by the BioCLIP ViT-B/16 text tower using the 80-template
OpenAI ImageNet prompt ensemble. A pack is only compatible with an app build whose
model_id and embedding_dim match its pack_meta.
- BioCLIP — MIT — https://huggingface.co/imageomics/bioclip · https://github.com/Imageomics/bioclip · Stevens et al., BioCLIP: A Vision Foundation Model for the Tree of Life, CVPR 2024, https://arxiv.org/abs/2311.18803
- open_clip — MIT — https://github.com/mlfoundations/open_clip — loads the
model and supplies the
OPENAI_IMAGENET_TEMPLATESprompt set - OpenAI CLIP — MIT — https://github.com/openai/CLIP — origin of the 80 ImageNet prompt templates
- SQLite — public domain — https://www.sqlite.org/copyright.html — pack container format