DrugLib Reviews Query Skill

Search patient drug reviews by drug name or medical condition. Auto-detects entity type and routes to the appropriate field.

Input Pattern	Detected As	Match Logic
single/two-word term (e.g. lamictal)	drug name	substring on urlDrugName
multi-word or medical keyword (e.g. bipolar disorder)	condition	substring on condition

Condition keywords that trigger condition routing: disease, disorder, syndrome, infection, pain, cancer, diabetes, hypertension, depression, anxiety, asthma, arthritis, migraine, allergy, insomnia, nausea, obesity, acne, gerd, copd.

API

Function	Input	Returns
load_reviews()	—	list[dict] (cached)
search(entity)	single entity string	list[dict]
search_batch(entities)	list of entity strings	dict[str, list[dict]]
summarize(hits, entity)	list[dict] + label	compact LLM-readable text
to_json(hits)	list[dict]	list[dict] (JSON-serialisable)

Usage

See if __name__ == "__main__" block in 16_DRUGLIB_REVIEWS.py for runnable examples covering: drug name search, condition search, batch search, and JSON output.

Data

• Source: Drug Review Dataset (Druglib.com), UCI ML Repository #461
• Citation: Kallumadi, S. & Gräßer, F. (2018). Drug Reviews (Druglib.com) [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C55G6J
• License: CC BY 4.0
• Files: drugLibTrain_raw.tsv, drugLibTest_raw.tsv (TSV, merged at load)
• Path: DATA_DIR variable in 16_DRUGLIB_REVIEWS.py
• Columns:

Column	Type	Description
urlDrugName	str	Drug name (lowercase, URL-style)
rating	int 0–9	Overall patient satisfaction
effectiveness	str	Highly / Considerably / Moderately / Marginally Effective, Ineffective
sideEffects	str	No / Mild / Moderate / Severe / Extremely Severe Side Effects
condition	str	Medical condition being treated
benefitsReview	str	Free-text review of benefits
sideEffectsReview	str	Free-text review of side effects
commentsReview	str	Free-text general comments