R-Package Knowledge Base (RPKB)

Project Page | Paper | GitHub

This database is the official pre-computed ChromaDB vector database for the paper: DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval.

It contains 8,191 high-quality R functions meticulously curated from CRAN, complete with extracted statistical metadata (Data Profiles) and pre-computed embeddings generated by the DARE model.

📊 Database Overview

• Database Engine: ChromaDB
• Total Documents: 8,191 R functions
• Embedding Model: Stephen-SMJ/DARE-R-Retriever
• Primary Use Case: Tool retrieval for LLM Agents executing data science and statistical workflows in R.

🚀 Quick Start (Zero-Configuration Inference)

You can easily download and load this database into your own Agentic workflows using the huggingface_hub and chromadb libraries.

1. Installation

pip install huggingface_hub chromadb sentence-transformers torch

2. Run the DARE Retriever

The following script automatically downloads the DARE model and the RPKB database from Hugging Face and performs a distribution-aware search.

from huggingface_hub import snapshot_download
from sentence_transformers import SentenceTransformer
import chromadb
import torch
import os

# 1. Load DARE Model
device = "cuda" if torch.cuda.is_available() else "cpu"
model = SentenceTransformer("Stephen-SMJ/DARE-R-Retriever", trust_remote_code=False)
model.to(device)

# 2. Download and Connect to RPKB Database
db_dir = "./rpkb_db"
if not os.path.exists(os.path.join(db_dir, "DARE_db")):
    print("Downloading RPKB Database from Hugging Face...")
    snapshot_download(repo_id="Stephen-SMJ/RPKB", repo_type="dataset", local_dir=db_dir, allow_patterns="DARE_db/*")

client = chromadb.PersistentClient(path=os.path.join(db_dir, "DARE_db"))
collection = client.get_collection(name="inference")

# 3. Perform Search
query = "I have a sparse matrix with high dimensionality. I need to perform PCA."
query_embedding = model.encode(query, convert_to_tensor=False).tolist()

results = collection.query(
    query_embeddings=[query_embedding],
    n_results=3,
    include=["documents", "metadatas"]
)

# Display Results
for rank, (doc_id, meta) in enumerate(zip(results['ids'][0], results['metadatas'][0])):
    print(f"[{rank + 1}] Package: {meta.get('package_name')} :: Function: {meta.get('function_name')}")

📖 Citation

If you find DARE, RPKB, or RCodingAgent useful in your research, please cite our work:

@article{sun2026dare,
      title={DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval}, 
      author={Maojun Sun and Yue Wu and Yifei Xie and Ruijian Han and Binyan Jiang and Defeng Sun and Yancheng Yuan and Jian Huang},
      year={2026},
      eprint={2603.04743},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2603.04743}, 
}