# ChaokunHong/MetaScreener

AI-powered tool for efficient abstract and PDF screening in systematic reviews.

Repository: https://github.com/ChaokunHong/MetaScreener
Canonical: https://ross.abutalabs.com/products/metascreener
Homepage: https://www.metascreener.net/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: literature-screening, llm, systematic-review, deepseek, pdf-screening, metaanalysis, screening, screening-api, deepseek-llm, llama4scout, mistral, qwen3
Last push: 2026-06-11T17:36:39+00:00

## Health v2 (maintenance only)
Score: 71/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 87, release rhythm 71, longevity 34
- inputs: {"age_days": 482, "days_push": 83, "days_rel": 117, "gap_med": 71, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1328, forks 49 (observed 2026-08-28T04:04:23.727537+00:00)

## What it is
MetaScreener is an open-source AI tool that automates title/abstract and full-text PDF screening for systematic reviews using an ensemble of multiple open-source LLMs with calibrated confidence scoring. It ingests RIS/BibTeX/CSV search results, applies PICO/PEO/SPIDER criteria, auto-decides high-confidence cases, and routes uncertain ones to human review with full audit trails.

## Use cases
- screen abstracts for a systematic review with AI
- automate title and abstract screening of PubMed search results
- apply PICO inclusion criteria to thousands of papers
- screen full-text PDFs with OCR for a meta-analysis
- reduce manual literature screening time in evidence synthesis
- flag uncertain studies for human review during screening
- export screening decisions to RIS or CSV for a review manager

## When to choose
- you are conducting a systematic review or meta-analysis and need to screen large volumes of citations
- you want reproducible, auditable AI screening with confidence scores rather than a single black-box model
- you need human-in-the-loop triage where only uncertain cases reach a reviewer
- you want a low-cost, privacy-conscious option using open-source LLMs and your own API keys

## When to avoid
- you need fully automated screening with no human verification, as the tool is designed to escalate uncertain cases
- your review requires strict journal or Cochrane compliance that prohibits LLM-assisted screening
- you lack API access to any of the supported LLM providers
- you need a simple single-model classifier rather than a multi-LLM ensemble pipeline

## Facets
- artifact type: application
- maturity: active
- function: llm-inference, rag, ocr, pdf, machine-learning, web-framework, api-framework
- domain: healthcare, large-language-models, artificial-intelligence, education, data-science
- platform: python, self-hosted, cross-platform, cli
- tags: systematic-review, literature-screening, meta-analysis, evidence-synthesis, multi-llm-ensemble, pico-criteria, human-in-the-loop, confidence-calibration, active-learning, ris-import, natural-language-processing, docker, web-server

## Member repositories
- ChaokunHong/MetaScreener (main) score 71

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:23.727537+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:45:47.709577+00:00, confidence not recorded.
  - readme: https://github.com/ChaokunHong/MetaScreener (fetched 2026-08-28T04:04:23.727537+00:00, sha 2eb201b8f497)
  - homepage: https://www.metascreener.net/ (fetched 2026-08-29T12:04:49.878148+00:00, sha 6ff5985f56ee)
  - registry_pypi: https://pypi.org/pypi/metascreener/json (fetched 2026-08-29T12:04:49.888217+00:00, sha 2b799ca5ca1f)
- Data as of 2026-08-30T08:39:29.467469+00:00.