Building an Autonomous Research Bot with GitHub Actions and Gemini

March 21, 2026 |

As part of my ongoing work, I wanted a way to automatically track and summarize the latest research related to a specific protein of interest. To solve this, I built an autonomous pipeline that scrapes new papers, summarizes them, and publishes a weekly digest directly to my personal website. Here’s how I set it up.

The Architecture: Python + Gemini API

The core of the bot is a Python script that runs weekly. It relies on two main components:

Europe PMC API: To search for recent publications related to the target protein (for example, “p53” or any other molecule).
Google Gemini API: To synthesize the abstracts and write a cohesive, human-readable narrative.

The fetcher script constructs a query for papers published in the last 7 days. It extracts the titles, authors, and abstracts, passing them to the Gemini 2.5 Flash model with a specific prompt:

“You are a scientific reporter writing a short Weekly Research Digest for our community. Write a CONCISE, flowing narrative of MAXIMUM 300 words that highlights the most important trends and findings… Focus on the 2-3 most impactful themes.”

Gemini returns a well-structured markdown summary containing inline citations, which the script then formats into a Jekyll-compatible markdown file with appropriate YAML frontmatter.

Automation: GitHub Actions Every Week

To make the bot truly autonomous, I leverage GitHub Actions. Here is a look at a simplified workflow configuration (.github/workflows/research-bot.yml):

name: Automated Research Fetcher
on:
  schedule:
    - cron: '0 0 * * 0' # Midnight UTC every Sunday
  workflow_dispatch:

jobs:
  fetch_and_publish:
    runs-on: ubuntu-latest
    permissions:
      contents: write
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.10'
      - name: Run Fetcher Script
        run: python3 scripts/fetch_research.py
      - name: Commit and Push Changes
        run: |
          git config --global user.name "Research Bot"
          git config --global user.email "bot@researchhub.local"
          git add _research/*
          git commit -m "Automated update: Added new research [skip ci]" || echo "No changes to commit"
          git push

Every week, the GitHub Action provisions an Ubuntu runner, sets up Python 3.10, and executes the fetcher script. If the script detects new papers and generates a new markdown digest in the _research/ directory, the workflow automatically commits and pushes the file back to the main branch.

Setting up the Gemini API Key

For the AI summarization to work, the Python script needs a valid GEMINI_API_KEY.

Get an API Key: Head over to Google AI Studio and generate a free API key for the Gemini model.
Add as a Secret: Go to your GitHub repository’s Settings > Secrets and variables > Actions, and click New repository secret.
Name the secret GEMINI_API_KEY and paste your key. The workflow will now automatically pull this securely during runtime!

Presentation: Jekyll Collections

With the markdown files committed to the _research/ directory, my website’s Jekyll configuration handles the rest. By defining a custom collection in _config.yml:

collections:
  research:
    output: true
    permalink: /research/:year/:month/:day/:title.html

The site automatically renders the AI-generated digests alongside my portfolio and blog under the /research/ route. The process is completely hands-off. The bot finds the research, writes the summary, and publishes the page, ensuring that my research hub is always up to date!