Laboratory for Social Connectedness & Health · University of Arizona

Code research papers with AI — then review in Excel

Upload PDFs and your coding manual. Claude fills the same JSON fields your PI defined. You download a CSV for human quality checks — not a replacement for expert review.

Try the interactive demo Setup on your Mac

The live app runs on your computer (not in the browser). This page is a guide + practice UI.

Interactive demo

Click through the same 4 steps the real app uses. Upload buttons are simulated here.

Step 1 — Connect API key

Ask your PI for an Anthropic key, or use one in a .env file.

○ Not done

Step 2 — Upload coding instructions

Your Word file, e.g. Claude_Code_Prompting_v4.docx

No file selected

○ Not done

Step 3 — Upload research PDFs

Up to 400 PDFs per batch. Drag & drop or browse.

  • No papers yet — try adding sample names below

○ Not done

Step 4 — Run & download

Click Start coding papers in the sidebar. When finished, open the CSV in Excel and check flagged rows, plus the change log for any papers that couldn't be coded.

    ○ Run simulation when ready

    How to use (real app)

    1

    Install once

    Clone the repo, install Python packages, add your API key to .env.

    2

    Start the UI

    python3 run_server.py then open http://127.0.0.1:8000

    3

    Upload & run

    Prompt + PDFs → Start → wait (several hours for a full 300+ paper run) → Download CSV + change log.

    4

    Human QA

    Review needs_human_review, eligibility, and any *_ERROR.json files.

    Install on your Mac

    git clone https://github.com/isjustabhi/systematic-review-processor.git
cd systematic-review-processor
pip3 install -r requirements.txt
# Create .env with ANTHROPIC_API_KEY=sk-ant-...
python3 run_server.py

    CLI alternative: python3 run_review.py (after PDFs are in Pilot Papers/)

    Who is this for?