Discovery with Zotero Connector

Navigate to Google Scholar, PubMed, arXiv, or any academic database. When a relevant paper appears, click the Zotero Connector extension in the browser toolbar.

Zotero automatically:

• Extracts metadata (title, authors, publication date, DOI)
• Downloads the PDF if available
• Files the paper in the selected collection
• Creates a proper citation entry

Expected time: 2 minutes per paper

AI Summarization with Gemini

Run the summarization script on the newly added PDF:

./gemini-summarize-paper.sh \
  ~/Zotero/storage/ABC123/smith_2024.pdf \
  smith_2024

The script generates a structured summary at ~/Obsidian/Literature Notes/smith_2024_summary.md containing:

• Research question and objectives
• Methodology (3-sentence overview)
• Key findings (bullet points)
• Study limitations
• Future research directions
• Suggested citations

Expected time: 1 minute (mostly automated)

Synthesis in Obsidian

Open the generated summary in Obsidian and review for accuracy. Add personal insights at the bottom:

## Personal Notes

**Relevance to my research**: This methodology could be adapted for...

**Critical observation**: The authors assume X, but recent work by [[jones-2023]] suggests...

**Next steps**: Test this approach on my dataset. See [[experiment-design-notes]].

Create bidirectional links to related papers using [[paper-name]] syntax. Add tags for organization:

#literature #machine-learning #active-learning #smith2024

Expected time: 2 minutes

Install the Batch Script

Copy the batch processing script to a directory in your PATH:

# Create scripts directory if needed
mkdir -p ~/scripts

# Copy and make executable
cp 02-code-examples/gemini-scripts/batch-summarize-zotero.sh ~/scripts/
chmod +x ~/scripts/batch-summarize-zotero.sh

Verify the script location:

which batch-summarize-zotero.sh

Expected output: /Users/yourname/scripts/batch-summarize-zotero.sh

Configure Automated Execution

Set up a cron job to run the script every 6 hours:

crontab -e

Add this line to the crontab file:

# Run batch summarization every 6 hours
0 */6 * * * ~/scripts/batch-summarize-zotero.sh >> ~/.gemini-batch-log 2>&1

Save and exit the editor (:wq in vim, Ctrl+X then Y in nano).

Verify the cron job:

crontab -l

Understanding the Automation

The script performs these operations automatically:

• Scans Zotero storage for PDFs added in the last 24 hours
• Generates AI summaries using Gemini API (respects free tier limits)
• Saves summaries to Obsidian Literature Notes folder with standardized naming
• Enforces 15-second delay between API calls (prevents rate limiting)
• Logs all activity to ~/.gemini-batch-log for monitoring
• Stops at 25 papers per day (free tier maximum)

Monitor progress:

tail -f ~/.gemini-batch-log

Expected behavior: Wake up each morning to find summaries of all papers added the previous day.

Aggregate Papers with Dataview

Create a Dataview query in Obsidian to collect all relevant papers:

TABLE author, year, key-finding
FROM "Literature Notes"
WHERE contains(tags, "#systematic-review")
SORT year DESC

This query displays a table of all papers tagged for inclusion in the systematic review, sorted by publication year.

For a more detailed view:

LIST
FROM "Literature Notes"
WHERE contains(tags, "#systematic-review")
SORT year DESC

Review the results to ensure all relevant papers are included and properly tagged.

Export Aggregated Notes

Export the collected notes to a single text file for AI processing. Use either the manual copy method or automated concatenation:

Manual method: Copy the Dataview results from Obsidian

Automated method (recommended):

# Concatenate all literature notes
cat ~/Obsidian/Literature\ Notes/*.md > literature-export.txt

# Verify file size
wc -l literature-export.txt

Expected output: Several hundred lines containing all note content.

For more control, use grep to filter specific tags:

grep -l "#systematic-review" ~/Obsidian/Literature\ Notes/*.md | \
  xargs cat > literature-export.txt

Generate AI Synthesis

Use Gemini to synthesize the exported notes into a structured literature review:

gemini << EOF
Synthesize these literature notes into a structured academic review:

$(cat literature-export.txt)

Organize the synthesis using these sections:

1. Theoretical frameworks - Identify dominant theories and conceptual models
2. Methodological approaches - Compare research designs, data collection methods, analytical techniques
3. Key findings - Highlight consensus areas and contradictions in the literature
4. Research gaps - Identify unanswered questions and understudied areas
5. Future directions - Suggest promising avenues for further investigation

Write in academic style with clear section headers. Target length: 1500 words.
Include in-text references to specific papers when discussing their contributions.
EOF

The output provides a comprehensive synthesis suitable for adaptation into a literature review chapter.

Refine and Verify

Import the AI-generated synthesis into Obsidian for refinement:

• Review for factual accuracy by cross-checking claims against original papers
• Verify all citations reference actual papers in your library
• Add critical analysis and your interpretive perspective
• Restructure sections if needed for better logical flow
• Format with proper Markdown headers and emphasis
• Add transitional sentences between sections

Save the refined synthesis as a separate note with appropriate tags and links to original literature notes.

Write in Obsidian with Citation Keys

Compose the manuscript in Markdown using simple citation syntax:

# Introduction

Recent studies demonstrate significant advances in AI-assisted research [@smith_2024; @jones_2023].
Multiple approaches have been proposed, though methodological differences remain [@martinez_2024].

## Background

### AI in Research Workflows

The integration of AI tools has transformed academic workflows. Smith (2024) found productivity
gains of 300-500% in literature review tasks [@smith_2024]. Jones et al. (2023) documented
similar improvements in data analysis [@jones_etal_2023, p. 156]. However, limitations exist
in areas requiring domain expertise [-@wilson_2024].

### Automation Challenges

While automation offers efficiency gains, several challenges persist:
- Technical barriers for non-programming researchers [@garcia_2023]
- Quality control in AI-generated summaries [@thompson_2024]
- Integration across heterogeneous tool ecosystems [@lee_2024]

These challenges require systematic investigation.

Citation syntax guide:

• [@author_year] - Standard citation in parentheses
• @author_year - Author name in text, year in parentheses
• [-@author_year] - Suppress author name (year only)
• [@author_year, p. 42] - Include page number
• [@author1_year; @author2_year] - Multiple citations

Compile to Word with APA Citations

Use the compilation script to generate a formatted Word document:

./compile-paper-with-citations.sh paper.md docx

The script executes Pandoc with these parameters:

• Input: paper.md (Markdown source)
• Bibliography: ~/Documents/library.bib (exported from Zotero)
• Citation style: apa.csl (American Psychological Association 7th edition)
• Output format: .docx (Microsoft Word)

Generated file: paper.docx with properly formatted in-text citations and reference list.

Verify the output:

• Open paper.docx in Microsoft Word
• Check citations match APA format
• Verify reference list is alphabetically sorted
• Confirm all cited works appear in references

Switch Citation Formats

Generate the same manuscript in Chicago style by modifying the CSL file:

pandoc paper.md \
  --bibliography ~/Documents/library.bib \
  --citeproc \
  --csl ~/.pandoc/csl/chicago-author-date.csl \
  -o paper-chicago.docx

For other formats, download the appropriate CSL file from the Zotero Style Repository and reference it:

# MLA format
pandoc paper.md \
  --bibliography ~/Documents/library.bib \
  --citeproc \
  --csl ~/.pandoc/csl/modern-language-association.csl \
  -o paper-mla.docx

# IEEE format
pandoc paper.md \
  --bibliography ~/Documents/library.bib \
  --citeproc \
  --csl ~/.pandoc/csl/ieee.csl \
  -o paper-ieee.docx

Expected result: Identical content formatted according to different citation standards.

Advanced: LaTeX Output for Journal Submission

Many academic journals require LaTeX format. Generate LaTeX output with proper citation formatting:

pandoc paper.md \
  --bibliography ~/Documents/library.bib \
  --citeproc \
  --csl ~/.pandoc/csl/nature.csl \
  -o paper.tex

Compile the LaTeX to PDF:

pdflatex paper.tex
bibtex paper
pdflatex paper.tex
pdflatex paper.tex

Or use a single Pandoc command for direct PDF generation:

pandoc paper.md \
  --bibliography ~/Documents/library.bib \
  --citeproc \
  --csl ~/.pandoc/csl/nature.csl \
  -o paper.pdf

Expected output: Publication-ready PDF with journal-specific citation formatting.