SLR Magic: A Complete Workflow for Automating Systematic Literature Reviews Using Large AI Models

SLR Magic is an open-source tool based on Google Apps Script that uses large language models such as Gemini and Qwen3 to automate the entire workflow of systematic literature reviews (SLR), including screening, evaluation, and extraction. Its core value lies in significantly accelerating research literature processing, reducing human bias, and being deeply integrated into the Google Workspace ecosystem to lower the barrier to use.

Systematic literature review is the cornerstone of scientific research, but the traditional process is time-consuming and labor-intensive (taking weeks/months), and prone to human errors and subjective biases due to fatigue. With the development of large language model (LLM) technology, automated processing of massive text has become possible. SLR Magic was born in this context to address the efficiency and reliability issues of traditional SLR.

SLR Magic is an AI-driven SLR acceleration tool developed based on Google Apps Script. It acts both as an "accelerator" (processing thousands of papers in minutes) and a "gatekeeper" (ensuring the relevance and quality of literature). Fully integrated with Google Sheets and Drive, it requires no complex local deployment, making it easy for researchers to use.

SLR Magic adopts a multi-agent architecture with four roles:

1. Abstract Screening Agent: Performs initial screening based on titles/abstracts, quickly judges relevance and provides reasons;
2. Full-text Review Agent: Reads full PDF texts, verifies actual relevance, and eliminates "packaged" papers;
3. Quality Assessment Agent: Evaluates the scientific rigor and methodological reliability of research;
4. Data Extraction Agent: Extracts structured JSON data from qualified papers to support subsequent analysis.

Technical Architecture: Follows Clean Code principles, divided into controller layer (e.g., ScreeningController), UI layer (HTML/JS interaction), service layer (e.g., GeminiAdapter), and tool layer (e.g., SheetUtils). Supports multiple models: native Google Gemini support, private vLLM endpoints, and Ollama local deployment (for sensitive data processing). Workflow: One-click environment initialization (create multiple worksheets) → Configure API/model/prompt → Import CSV (Scopus/Web of Science) and PDF → Parallel processing → Generate visual charts (Sankey diagram, pie chart, etc.).

SLR Magic has built-in token tracking and cost estimation functions to help control the budget. It follows the FAIR principles (Findability, Accessibility, Interoperability, Reusability). Core principle: "Human-in-the-loop"—AI decisions need to be reviewed and verified by experts; the tool is an accelerator, not a replacement.

Suitable scenarios: Large-scale literature reviews, rapid evidence synthesis, repetitive review updates, multi-center collaboration, and teaching demonstrations. Value: Lower technical barriers, improve efficiency (weeks → days), maintain academic rigor, and support real-time team collaboration.

SLR Magic represents an important direction of AI-assisted research: freeing humans from repetitive work to focus on critical thinking. Its open-source nature allows community contributions for improvement and adaptation to more disciplinary scenarios. As LLM capabilities improve and costs decrease, such tools will play a more important role in the research workflow.