LatteReview (Systematic Review Automation)

Systematic literature screening is automated using AI-driven multi-agent workflows that evaluate papers against customizable inclusion and exclusion criteria across title/abstract and full-text stages.

Data abstraction from scientific papers is performed by dedicated AI agents that extract structured data points and assign relevance scores to streamline synthesis for biomedical research.

A local LM Studio endpoint is required, as the multi-agent review workflow expects LatteReview infrastructure to execute its title/abstract, scoring, and abstraction reviewers.

Set up a literature dataset by loading a CSV file such as data/papers.csv, defining inclusion/exclusion criteria, and running the screening workflow to produce a results CSV.

Multi-agent AI full-text screening works by deploying specialized reviewers that perform in-depth evaluation of candidate studies, applying customizable criteria to filter literature for evidence synthesis.

Limitations include dependency on local LM Studio infrastructure and the need for clearly defined inclusion criteria, as automated data abstraction and scoring are constrained by the provided model's capabilities.