Reference¶

Links to every external resource the project depends on — grouped by role.

Project and data¶

Resource	Link
This project	biplabendu/shiny-chatbot-dictionary-abcd
ABCD Study	abcdstudy.org
NBDCtools (ABCD data access)	nbdc-datahub/ABCDtools

Resource	Link
all-MiniLM-L6-v2	sentence-transformers/all-MiniLM-L6-v2 on HuggingFace
ONNX int8 weights (AVX2)	same repo, path `onnx/model_quint8_avx2.onnx`

Package	Role in the app	GitHub
shiny	Web application framework	rstudio/shiny
reticulate	R–Python bridge	rstudio/reticulate
reactable	Interactive results table	glin/reactable
bslib	Bootstrap UI theming	rstudio/bslib
nanoparquet	Reads the Parquet dictionary	r-lib/nanoparquet
fontawesome	Icons in buttons	rstudio/fontawesome
cicerone	Guided onboarding tour (driver.js wrapper)	JohnCoene/cicerone
dplyr	Data manipulation	tidyverse/dplyr
stringr	String cleaning	tidyverse/stringr
yaml	Reads `config.yml` at app startup	vubiostat/r-yaml
jsonlite	Serializes the hidden-column lists passed to the client-side JS	jeroen/jsonlite
htmlwidgets	Embeds reactable's row-click JS handler	ramnathv/htmlwidgets

These are installed automatically on first deploy via reticulate::py_require() + uv.

Package	Role	GitHub
onnxruntime	Runs the ONNX model at query time	microsoft/onnxruntime
tokenizers	Fast Rust-backed query tokenizer	huggingface/tokenizers
numpy	Embedding arrays and cosine matmul	numpy/numpy

Used by python/build_embeddings.py to produce the pre-assembled artifacts. Not shipped to the deployed app.

Package	Role	GitHub
huggingface_hub	Downloads model weights from HuggingFace	huggingface/huggingface_hub
pandas	Reads the dictionary Parquet at build time	pandas-dev/pandas
fastparquet	Pandas engine used to read the dictionary Parquet	dask/fastparquet
PyYAML	Parses `config.yml` (model + dictionary settings)	yaml/pyyaml

Resource	Link
MkDocs	mkdocs/mkdocs
pymdownx extensions	facelessuser/pymdown-extensions
Mermaid (diagrams)	mermaid-js/mermaid