About AI4Meta
AI-assisted systematic review & meta-analysis workflow
What is AI4Meta?
AI4Meta is an end-to-end platform for systematic reviews and meta-analyses. It combines AI-assisted literature search, dual-reviewer screening, structured data extraction, statistical analysis, and report generation into a single reproducible workflow. Every decision is logged, every analysis is traceable, and every report is exportable to PRISMA-compliant formats.
Large Language Models
AI4Meta runs a multi-agent orchestration engine backed by MiniMax M2.7 (8-bit) as the default coordinator, with per-task routing across OpenAI, Anthropic, and Google provider families. Users can configure fallback chains and per-agent model overrides in project settings.
- MiniMax-M2.7-8bit — default orchestrator & coordinator backend
- MiniMax M2.5 — high-throughput screening & extraction tasks
- GPT-4o — high-capability reasoning & multimodal fallback
- GPT-4o-mini — fast, cost-efficient routine operations
- Claude Sonnet (4.5) — balanced speed & quality for deep analysis
- Claude Haiku — ultra-fast pre-screening & triage
- Gemini 2.5 Pro — advanced reasoning & long-context synthesis
- Gemini 2.0 Flash — high-throughput batch processing
How It Was Built
AI4Meta was built through a carefully coordinated multi-agent collaboration. Rather than traditional waterfall planning, the codebase grew through iterative exploration between human architects and autonomous coding agents: feature ideas were prototyped, tested, and refined in tight feedback loops. The platform itself runs a multi-agent orchestration engine — a coordinator agent delegates tasks to specialist agents (screening, extraction, analysis, report generation) based on intent classification, with each agent selecting the most capable model for its specific task.
The frontend is a Next.js 15 + React 19 application with Tailwind CSS, the backend is FastAPI + SQLAlchemy, and the deployment is fully containerised with Docker Compose. Every change is validated against live running containers before being considered complete.
Open Source
AI4Meta is built in the open. The source code, issue tracker, and roadmap are available on GitHub. Contributions - whether bug reports, feature requests, or pull requests - are welcome.