Research & Development

Scope

• Objective: prismAId leverages Large Language Models (LLMs) for systematic scientific literature reviews, making them accessible and efficient without coding.
• Speed: Faster than traditional methods, prismAId provides high-speed software for systematic reviews.
• Replicability: Addresses the challenge of consistent, unbiased analysis, countering the subjective nature of human review.
• Cost: More economical than custom AI solutions, with review costs typically between $0.0025 and $0.10 per paper.
• Audience: Suitable for scientists conducting literature reviews, meta-analyses, project development, and research proposals.

Mechanism

LLM Basics

• How LLMs Work:
  • Large Language Models (LLMs) are AI trained on extensive text data to understand and generate human-like text.
  • These models handle various language tasks like text completion, summarization, translation, and more.
• Data Flow and Processing:
  • Modern LLMs offer subscription-based API access, with prismAId focusing on prompt engineering to extract targeted information.
  • prismAId enables structured, replicable prompt creation for systematic reviews, simplifying rigorous data extraction.

Data Flow

• prismAId’s workflow embeds protocol-based approaches:
  • Literature Selection: Based on defined protocols, ensuring replicability.
  • Content Classification: prismAId handles paper classification, parsing selected literature to extract user-defined information.
  • API Calls & Cost Management: prismAId sends single-shot prompts for each paper, processes AI-generated JSON files, and provides token-based cost estimates for informed decision-making.

Contributing

How to Contribute

We welcome contributions to improve prismAId, whether you’re fixing bugs, adding features, or enhancing documentation:

• Branching Strategy: Create a new branch for each set of related changes and submit a pull request via GitHub.
• Code Reviews: All submissions undergo thorough review to maintain code quality.
• Community Engagement: Connect with us through GitHub issues and discussions for feature requests, suggestions, or questions. - Discuss in the Matrix prismAId Support Room or follow the prismAId Announcements Room for the latest updates and release notifications.

Guidelines

For detailed contribution guidelines, see our CONTRIBUTING.md and CODE_OF_CONDUCT.md.

Software Stack

prismAId is developed in Go, selected for its simplicity and efficiency with concurrent operations. We prioritize the latest stable Go releases to incorporate improvements.

Technical Foundation

To support consistent development, we provide templates for VSCodium or Visual Studio Code that are located in the root directory of our repository.

Using the Templates:

1. Clone Repository: Clone the prismAId repository.
2. Open in VSCodium/VSCode.
3. Copy JSON Files: Place in a vscode directory at the project root.
4. Remove .template Extension and follow inline instructions.
5. Ignore in GIT: Add to .gitignore for privacy.

Architecture

• Go Module: Core logic and API access are implemented in Go.
• Python Package: Python wrapper around a C shared library compiled from the Go code.
• R Package: Contains a C shared library with an intermediate C wrapper, enabling R interaction with the shared library.
• Self-Contained Binaries: Simplifies setup by packaging all dependencies within the binaries.
• Cross-Platform Compatibility: Fully operational across Windows, macOS, and Linux.

Development Philosophy

• Open Source: We value community contributions and transparency.
• CI/CD Pipelines: Automated testing and deployment maintain quality and reliability.

Open Science Support

prismAId actively supports Open Science principles through:

1. Transparency and Reproducibility
   • prismAId enhances transparency, making analyses understandable and reproducible, with consistent results across systematic reviews.
   • Detailed logs and records improve reproducibility.
2. Accessibility and Collaboration
   • An open-source, openly licensed tool fostering collaboration and participation.
   • Long-term accessibility through Zenodo.
3. Efficiency and Scalability
   • Efficient data handling enables timely, comprehensive reviews.
4. Quality and Accuracy
   • Explicit prompts define information clearly, ensuring consistent, reliable reviews.
5. Ethics and Bias Reduction
   • Transparent design minimizes biases, with community oversight supporting ethical standards.
6. Scientific Innovation
   • Standardized, reusable methods facilitate innovation, cumulative knowledge, and rapid knowledge dissemination.