gregory-ai

Search the GregoryAI research database (https://api.brain-regeneration.com) read-only: find scientific articles and clinical trials on a medical or neuroscience topic, run boolean title/abstract searches, filter by DOI, registry ID (NCT/EudraCT/EUCT/CTIS), journal, date, or AI-relevance, look up authors, browse research subjects and categories, and get RSS feeds. Use whenever someone wants to find papers or trials, explore this research corpus, or pull results as JSON/CSV. Read-only — no data is ever created or modified.

Build and maintain REST API endpoints for GregoryAI's Django REST Framework application. Use this skill whenever someone asks to create a new API endpoint, add a serializer, write a DRF filter, register a URL route, add query parameters to an existing endpoint, fix an API bug, optimise a slow endpoint, add CSV export support, or work with the API layer in any way. Also trigger when someone asks about how the API works, what endpoints exist, what filters are available, or how to consume the GregoryAI API. If the task touches views.py, serializers.py, filters.py, pagination.py, or urls.py inside the api/ or admin/ app, use this skill.

Train, evaluate, debug, and extend GregoryAI's machine learning pipeline for article relevance prediction. Use this skill whenever someone asks about training models, running predictions, debugging ML issues, adding new algorithms, understanding prediction scores, pseudo-labeling, model versioning, or the ML consensus system. Triggers include questions like "train models for team X", "why is recall low", "add a new algorithm", "how does pseudo-labeling work", "run predictions", "check model metrics", "what does the ML pipeline do", or any work involving the gregory.ml module, train_models / predict_articles management commands, or the files_repo_PBL_nsbe research notebooks.

Generate read-only database queries for GregoryAI's PostgreSQL database — Django ORM or raw SQL. Use this skill whenever someone asks to query, count, filter, aggregate, export, or analyse data in the GregoryAI database. This includes questions like "how many articles were added last month", "show me trials by subject", "which sources haven't fetched anything recently", "articles with high ML prediction scores", "list authors with ORCID", or any ad-hoc data exploration against the gregory, subscriptions, or sitesettings models. Also trigger when someone asks for data exports, CSV dumps, or reporting queries. If the question could be answered by querying the database, use this skill.