Launching v2.12.0 — Infographics, Parallel Web Search & Venue Style Guides

New since v2.9.0:

• Infographics skill — AI-generated infographics with 10 types, 8 style presets, colorblind-safe palettes, and Gemini 3 Pro quality review. Use --research to pull live data before generating.
• Parallel web search — All research queries now route through the Parallel Chat API. Faster, smarter routing between Parallel Deep Research and Perplexity based on query type.
• Venue writing style guides — Publication-ready style guides for Nature/Science, Cell Press, medical journals (NEJM/Lancet/JAMA), and ML/CS conferences (NeurIPS, ICML, ACL, CHI). Abstracts now default to flowing paragraphs.
• Model upgrade — All references updated to claude-opus-4-6.
• Citation hardening — BibTeX entries must have volume, pages, and DOI before final compile. Research results now auto-saved to sources/ for reproducibility.
• Slides & posters — Cleaner slide output (no leftover _v1/_v2 files), formatting consistency protocol, and simplified poster layout rules.

Full Changelog: v2.9.0...v2.12.0

v2.9.0

🚀 Release v2.9.0

Major release with enhanced citation verification and research workflows.

🔧 Improvements

• Citation Verification - Enhanced process using WebSearch for metadata validation
• Research Lookup - Upgraded Sonar Pro to Sonar Pro Search for more accurate results
• Documentation - Updated output directory references throughout codebase

📦 Installation

pip install scientific-writer==2.9.0

Or with uv:

uv pip install scientific-writer==2.9.0

📝 Full Changelog

See CHANGELOG.md for complete details.

v2.8.0 - Nano Banana Pro: AI-Powered Scientific Diagrams 🎨

Scientific Writer 2.8.0 - Nano Banana Pro Scientific Schematics 🎨

Release Date: November 20, 2025

🚀 Headline Feature: AI-Powered Scientific Diagrams with Nano Banana Pro

Version 2.8.0 introduces Nano Banana Pro, a revolutionary AI-powered system for generating publication-quality scientific diagrams from natural language descriptions. No coding, no templates, no manual drawing required—just describe what you want, and Nano Banana Pro creates it automatically.

✨ Simply Describe Your Diagram

# CONSORT flowchart
python scripts/generate_schematic.py \
  "CONSORT participant flow: screened n=500, excluded n=150, randomized n=350" \
  -o consort.png

# Neural network architecture
python scripts/generate_schematic.py \
  "Transformer architecture with encoder and decoder, show attention mechanism" \
  -o transformer.png

# Biological pathway
python scripts/generate_schematic.py \
  "MAPK signaling pathway: EGFR → RAS → RAF → MEK → ERK → nucleus" \
  -o mapk.png

That's it! Nano Banana Pro automatically:

• ✅ Generates your diagram from the description
• ✅ Reviews quality and identifies improvements
• ✅ Iteratively refines the diagram (3 iterations by default)
• ✅ Produces publication-ready output

🎯 What Makes Nano Banana Pro Special

1. Automatic Iterative Refinement

Nano Banana Pro doesn't just generate once—it improves your diagram through an intelligent review cycle:

Iteration 1: Initial generation from your description

• AI evaluates: clarity, labels, accuracy, accessibility
• Scores the diagram (0-10) with specific critiques

Iteration 2: Improvements based on feedback

• Addresses specific issues identified in review
• Re-evaluates with updated quality assessment

Iteration 3: Final polished version

• Incorporates all feedback from previous iterations
• Publication-ready output

Example progression:

v1: Score 7.0 - "Good structure, but font size too small, labels overlap"
v2: Score 8.5 - "Much improved readability, minor spacing issues remain"
v3: Score 9.5 - "Excellent. Publication ready. Professional quality."

2. Built-In Scientific Quality Standards

Every diagram automatically follows best practices:

Visual Quality:

• Clean white/light backgrounds
• High contrast for readability
• Sharp, clear lines and text
• Professional appearance

Typography:

• Sans-serif fonts (Arial, Helvetica)
• Minimum 10pt font size
• Consistent sizing throughout
• No overlapping text

Scientific Standards:

• Accurate representation of concepts
• Clear labels for all components
• Appropriate scale bars, legends, axes
• Standard notation and symbols

Accessibility:

• Colorblind-friendly color palettes (Okabe-Ito scheme)
• High contrast ratios
• Grayscale-compatible designs
• WCAG 2.1 compliant

3. Universal Diagram Support

Nano Banana Pro works for any type of scientific diagram:

Clinical & Medical:

• CONSORT participant flowcharts
• Clinical trial designs
• Diagnostic algorithms
• Medical decision trees
• Patient treatment pathways

Computational & AI:

• Neural network architectures (CNNs, Transformers, RNNs)
• Algorithm flowcharts
• System architectures
• Data pipelines
• Software workflows

Biological & Chemical:

• Signaling pathways (MAPK, PI3K/AKT, etc.)
• Metabolic pathways
• Gene regulation networks
• Protein structures
• Chemical reaction schemes

Engineering & Physics:

• Circuit diagrams
• Block diagrams
• System architectures
• Signal processing flows
• Experimental setups

And Many More:

• Study designs
• Conceptual frameworks
• Process diagrams
• Organizational charts
• Timeline diagrams

4. Comprehensive Output

For every diagram, you receive:

your_diagram_v1.png          # First iteration
your_diagram_v2.png          # Second iteration
your_diagram_v3.png          # Final polished version
your_diagram.png             # Copy of final (for convenience)
your_diagram_review_log.json # Detailed quality assessment

Review Log Example:

{
  "user_prompt": "CONSORT participant flow diagram...",
  "iterations": [
    {
      "iteration": 1,
      "image_path": "figures/consort_v1.png",
      "score": 7.0,
      "critique": "Good structure. Issues: font too small at 8pt (need 10pt+), 
                  some labels overlap, arrows could be clearer.",
      "success": true
    },
    {
      "iteration": 2,
      "score": 8.5,
      "critique": "Much improved. Font now readable, labels clear. 
                  Minor spacing issues in exclusion criteria box.",
      "success": true
    },
    {
      "iteration": 3,
      "score": 9.5,
      "critique": "Excellent. Publication ready. Professional quality, 
                  clear hierarchy, excellent readability.",
      "success": true
    }
  ],
  "final_image": "figures/consort_v3.png",
  "final_score": 9.5,
  "success": true
}

📖 Real-World Examples

Example 1: Clinical Research

Prompt:

python scripts/generate_schematic.py \
  "CONSORT participant flow diagram for RCT.
   Assessed for eligibility (n=500).
   Excluded (n=150): age<18 (n=80), declined (n=50), other (n=20).
   Randomized (n=350) into Treatment (n=175) and Control (n=175).
   Lost to follow-up: 15 and 10 respectively.
   Final analysis: 160 and 165." \
  -o figures/consort.png

Result: Professional CONSORT flowchart ready for journal submission

Example 2: Deep Learning Architecture

Prompt:

python scripts/generate_schematic.py \
  "Transformer architecture. Left side: encoder with input embedding, 
   positional encoding, multi-head self-attention, feed-forward layers. 
   Right side: decoder with masked attention, cross-attention, feed-forward. 
   Show attention connections from encoder to decoder. 
   Label all components with dimensions." \
  -o figures/transformer.png --iterations 5

Result: Detailed architecture diagram suitable for conference papers

Example 3: Biological Pathway

Prompt:

python scripts/generate_schematic.py \
  "MAPK signaling pathway showing activation cascade. 
   Start with EGFR receptor at membrane → RAS → RAF → MEK → ERK → nucleus. 
   Label each phosphorylation step. Use different colors for each kinase. 
   Include inhibitor binding sites." \
  -o figures/mapk.png

Result: Publication-quality pathway diagram with proper biological notation

🎓 Prompt Engineering Tips

✅ Effective Prompts (Specific & Detailed)

Good:

"CONSORT flowchart with vertical flow, top to bottom. 
 Screening box at top (n=500), exclusion criteria in middle (n=150), 
 randomization at bottom (n=350 split into two groups)"

Why it works:

• Specifies layout direction (vertical, top-to-bottom)
• Includes all quantitative details
• Describes structure clearly
• Mentions grouping explicitly

❌ Ineffective Prompts (Too Vague)

Bad:

"Make a flowchart"

Why it fails:

• No structural details
• No content specified
• No layout guidance
• Missing quantitative information

🌟 Key Elements of Great Prompts:

1. Specify Layout:

   • "Vertical flow, top to bottom"
   • "Architecture with encoder on left, decoder on right"
   • "Circular pathway around central hub"

2. Include Quantitative Details:

   • "Neural network: input layer (784 nodes), hidden (128), output (10)"
   • "Patient flow: n=500 screened, n=150 excluded, n=350 randomized"
   • "Pathway with 5 phosphorylation steps"

3. Describe Visual Style:

   • "Minimalist block diagram with clean lines"
   • "Detailed biological pathway with protein structures"
   • "Technical schematic with engineering notation"

4. Request Specific Labels:

   • "Label all arrows with activation/inhibition"
   • "Include layer dimensions in each box"
   • "Show molecular weights for each protein"

5. Mention Color Requirements:

   • "Use colorblind-friendly palette"
   • "Grayscale-compatible design"
   • "Color-code by function: blue=input, green=processing, red=output"

🚀 Getting Started

Step 1: Get Your API Key

Visit https://openrouter.ai/keys to get your OpenRouter API key.

Step 2: Set Environment Variable

# One-time setup
export OPENROUTER_API_KEY='sk-or-v1-your_key_here'

# Make it permanent (add to ~/.bashrc or ~/.zshrc)
echo 'export OPENROUTER_API_KEY="sk-or-v1-your_key"' >> ~/.zshrc
source ~/.zshrc

Step 3: Generate Your First Diagram

# Try a simple test
python scripts/generate_schematic.py \
  "Simple flowchart with 3 boxes connected by arrows" \
  -o test.png

# Check the outputs
ls test*.png              # v1, v2, v3 versions
cat test_review_log.json  # Quality scores and feedback

Step 4: Review the Results

Open the generated images:

• test_v1.png - See the initial generation
• test_v2.png - Notice the improvements
• test_v3.png - Final polished version

Read the review log to understand the quality progression.

🔧 Advanced Features

Custom Iteration Count

Need more refinement? Adjust the iteration count:

# 5 iterations for complex diagrams
python scripts/generate_schematic.py \
  "Complex multi-layer neural network architecture" \
  -o complex_nn.png \
  --iterations 5

Verbose Mode for Debugging

See detailed progress and API interactions:

python scripts/generate_schematic.py \
  "diagram description" \
  -o output.png \
  -v  # Verbose mode

Python API Integration

Use Nano Banana Pro programmatically in your own code:

from scripts.generate_schematic_ai import ScientificSchematicGenerator

# Initialize generator
generator = ScientificSchematicGenerator(
    api_key="your_key_here",
    verbose=True
)

# Generate with iterative refinement
results = generator.generate_iterative(
    user_prompt="CONSORT flowchart for clinical trial",
    output_path="figures/consort.png",
    iterations=3
)

# Access results
print(f"Final quality score: {results['fina...

v2.7.0 - Claude Code Plugin Focus

[2.7.0] - 2025-01-22

🎯 Claude Code Plugin Focus

This release emphasizes using Scientific Writer as a Claude Code (Cursor) plugin, making it easier than ever to access scientific writing capabilities directly in your IDE.

✨ Added

Enhanced Plugin Experience

• Streamlined Plugin Installation - Improved documentation and setup process for Claude Code plugin usage

  • Clear step-by-step installation guide
  • Marketplace integration instructions
  • Local development and testing guide
  • Troubleshooting for common plugin issues

• Optimized Plugin Structure - Better organization for plugin usage

  • All 19+ skills automatically available when plugin is installed
  • /scientific-writer:init command creates comprehensive CLAUDE.md configuration
  • Skills accessible directly in IDE without additional setup
  • Template files optimized for plugin context

• Plugin-First Documentation - Enhanced README with prominent plugin section

  • Plugin installation prominently featured at the top
  • Clear examples for using skills within Claude Code
  • Plugin testing guide for developers
  • Troubleshooting section for plugin-specific issues

🔧 Improvements

Better IDE Integration

• Seamless Skill Access - All skills work natively within Claude Code

  • No need to switch between CLI and IDE
  • Skills automatically discoverable via @skill-name syntax
  • Context-aware skill suggestions
  • Direct file editing and creation within IDE

• Improved Initialization Command - Enhanced /scientific-writer:init experience

  • Better handling of existing CLAUDE.md files
  • Backup and merge options for existing configurations
  • Clear feedback on what was installed
  • Summary of available skills and capabilities

• Plugin-Optimized Workflows - Workflows designed for IDE usage

  • File operations work directly in project directory
  • No need for separate data folders - use project structure
  • Skills integrate with IDE's file system
  • Better progress feedback within IDE context

📝 Documentation Updates

• Plugin Quick Start - New quick start guide for plugin users
• Plugin Examples - Real-world examples of using skills in Claude Code
• Skill Reference - Complete list of all 19+ available skills
• Troubleshooting - Common plugin installation and usage issues

🎯 Usage Examples

Plugin Installation

# Add marketplace
/plugin marketplace add https://github.com/K-Dense-AI/claude-scientific-writer

# Install plugin
/plugin install claude-scientific-writer

# Initialize in your project
/scientific-writer:init

Using Skills in Claude Code

# Create a paper (skill automatically used)
> Create a Nature paper on CRISPR gene editing

# Use specific skills
> @research-lookup Find recent papers on mRNA vaccines
> @peer-review Evaluate this manuscript
> @clinical-reports Create a case report for this patient

# Generate documents
> Create an NSF grant proposal for quantum computing
> Generate conference slides from my paper
> Create a research poster for NeurIPS

💡 Key Benefits for Plugin Users

• No CLI Required - Everything works directly in Claude Code
• Instant Access - All 19+ skills available immediately after installation
• IDE Integration - Files created and edited directly in your project
• Context Aware - Skills understand your project structure
• Seamless Workflow - No switching between tools

🚀 Migration from CLI to Plugin

For existing CLI users:

• Plugin provides same functionality with better IDE integration
• Skills work identically in both CLI and plugin modes
• Can use both CLI and plugin in the same project
• Plugin is recommended for IDE-based workflows

📦 Plugin Structure

claude-scientific-writer/
├── .claude-plugin/          # Plugin metadata (if exists)
├── commands/                 # Plugin commands
│   └── scientific-writer-init.md
├── skills/                   # All 19+ skills
│   ├── research-lookup/
│   ├── peer-review/
│   ├── clinical-reports/
│   └── ... (16 more)
├── templates/                # CLAUDE.md template
│   └── CLAUDE.scientific-writer.md
└── ... (Python package files)

🎨 Plugin Features

• 19+ Specialized Skills - Research, writing, review, and more
• One-Command Setup - /scientific-writer:init configures everything
• Skill Discovery - Ask "What skills are available?" to see full list
• Direct Integration - Skills work with IDE's file operations
• Template System - Professional templates for all document types

v2.6.0 - Professional Hypothesis Generation Reports

✨ New: Professional Hypothesis Generation Reports

Scientific Hypothesis Generation Framework

Generate publication-ready hypothesis reports with:

• 3-5 Competing Mechanistic Hypotheses - Evidence-based explanations with systematic evaluation
• Beautiful LaTeX Reports - Colored boxes for visual organization (blue, green, purple, teal, orange)
• 7-Dimensional Quality Assessment - Testability, falsifiability, parsimony, explanatory power, scope, consistency, novelty
• Comprehensive Citations - 50+ references (15-20 main text, 40-60+ appendices)
• Detailed Experimental Designs - Ready-to-implement testing strategies

Report Structure

Main Text (Concise, 8-14 pages):

• Executive Summary - One-page overview
• Competing Hypotheses - Each in dedicated colored box
• Testable Predictions - Specific, measurable predictions
• Critical Comparisons - How to distinguish between hypotheses

Appendices (Comprehensive):

• Literature Review (40-60+ citations)
• Detailed Experimental Designs
• Quality Assessment Tables
• Supplementary Evidence

Usage Examples

# Cancer biology
scientific-writer
> Why do some tumors respond to immunotherapy while others don't?

# Neuroscience  
> What mechanisms could explain the therapeutic effect of ketamine in depression?

# Climate science
> Generate hypotheses for accelerated ice sheet melting in Greenland

Scientific Rigor

• ✅ Systematic literature search and synthesis
• ✅ Multiple competing explanations (not just one)
• ✅ Seven-dimensional quality evaluation
• ✅ Detailed experimental test designs
• ✅ Clear, falsifiable predictions
• ✅ Professional LaTeX presentation

Installation

pip install scientific-writer==2.6.0
# or
uv pip install scientific-writer==2.6.0
# or
uvx scientific-writer@2.6.0

Documentation

See CHANGELOG.md for full release notes.

---

Full Changelog: v2.5.0...v2.6.0

v2.5.0 - Scientific Slides & Presentation System 🎨

Scientific Writer v2.5.0 - Scientific Slides & Presentation System 🎨

We're excited to announce v2.5.0, a major update introducing professional scientific presentation generation capabilities!

🌟 What's New

Professional Presentation Generation

Transform your research into beautiful, professional presentations with just a simple prompt:

scientific-writer
> Create a conference presentation on The AI Scientist framework

Key Features:

• 🎯 LaTeX Beamer Templates - Modern, professional designs optimized for academic settings
• 📊 Automatic Content Structuring - AI-powered organization for visual delivery
• 🎨 Beautiful Design - Clean aesthetics following academic best practices
• ♿ WCAG 2.1 Compliant - Accessible color schemes and layouts
• 📑 Multiple Formats - Export to PDF and PowerPoint (PPTX)
• 🔄 Paper-to-Slides Conversion - Transform existing papers into presentations

Comprehensive Presentation Skill

The new scientific-slides skill includes extensive resources:

📐 Design Guidelines (663 lines)

• Visual hierarchy and layout principles
• Color theory and accessibility standards
• Typography best practices for presentations
• Data visualization guidelines
• Animation and transition recommendations
• Venue-specific formatting (conferences, seminars, posters)

📦 Complete Resource Library

• Professional LaTeX Beamer templates
• PowerPoint conversion scripts
• Presentation assets (icons, diagrams)
• Example automation scripts
• Reference materials and best practices

PowerPoint Conversion Support

Generate presentations in multiple formats:

• Native LaTeX Beamer (PDF)
• PowerPoint (PPTX) via python-pptx
• Preserves layout, formatting, and design elements
• Supports complex slide structures and animations

🚀 Usage Examples

CLI - Generate Conference Presentation

scientific-writer
> Create a 20-minute conference presentation on CRISPR gene editing

# The system will:
# ✓ Generate professional Beamer slides
# ✓ Structure content for the time limit
# ✓ Include diagrams and figures
# ✓ Compile to PDF
# ✓ Optionally convert to PowerPoint

API - Programmatic Generation

import asyncio
from scientific_writer import generate_paper

async def main():
    async for update in generate_paper(
        "Create a research seminar presentation on machine learning in drug discovery"
    ):
        if update["type"] == "progress":
            print(f"[{update['percentage']}%] {update['message']}")
        else:
            print(f"✓ Presentation: {update['files']['pdf_final']}")

asyncio.run(main())

Convert Existing Papers

# Place your paper in the data folder
cp my_paper.pdf data/

scientific-writer
> Convert this paper into a 15-minute department seminar presentation

# Creates:
# ✓ Structured slides with key findings
# ✓ Visual representations of data
# ✓ Speaker notes
# ✓ Time-optimized content

💡 Why This Matters

For Researchers:

• Save hours creating presentations from scratch
• Ensure professional quality for conferences and seminars
• Focus on content while AI handles design
• Accessible presentations following WCAG guidelines

For Institutions:

• Consistent, professional branding
• Rapid presentation creation for grant proposals
• Enhanced communication of research findings
• Integration with existing paper workflows

🎨 Design Philosophy

Built on evidence-based principles:

• Cognitive Load Theory - Minimize extraneous information
• Dual Coding Theory - Combine verbal and visual elements
• Evidence-Based Presentation - CONSORT/PRISMA standards
• Academic Communication - Nature, Science, Cell guidelines

📊 Real-World Examples

Check out presentations generated with v2.5.0:

• AI Scientist framework presentation (Sakana AI)
• CRISPR applications in agriculture
• Machine learning research seminars
• Clinical trial results presentations

🔧 Installation

Install or upgrade to v2.5.0:

pip install --upgrade scientific-writer==2.5.0
# or
uv pip install scientific-writer==2.5.0
# or run directly
uvx scientific-writer@2.5.0

📝 Documentation

All presentation capabilities are fully documented in:

• scientific_writer/.claude/skills/scientific-slides/ - Complete skill directory
• assets/powerpoint_design_guide.md - 663-line comprehensive guide
• Templates, scripts, and reference materials included

🎯 What's Next

Future enhancements planned:

• Additional presentation themes
• Interactive slide generation
• Enhanced animation support
• Collaboration features

🙏 Acknowledgments

This release maintains backward compatibility with all existing features while adding powerful new presentation capabilities. The CLI and API experiences remain unchanged for existing users.

📋 Full Changelog

See CHANGELOG.md for complete release notes.

---

Quick Links:

• 📦 PyPI Package
• 📖 Documentation
• 🐛 Report Issues
• 💬 Discussions

v2.4.0 - Smart File Routing and Enhanced Editing Workflow

Release v2.4.0 - Smart File Routing and Enhanced Editing Workflow

🎯 Overview

Version 2.4.0 introduces a Smart File Routing System that revolutionizes how files are organized and processed in the Scientific Writer. This release brings intelligent file categorization, a dedicated sources folder for reference materials, and refined editing mode detection for a more predictable and organized workflow.

---

✨ Major New Features

🔍 Smart File Routing System

The centerpiece of this release is the new intelligent file categorization system that automatically routes files to the appropriate folders based on their type and purpose:

• 📝 Manuscript files (.tex only) → drafts/ folder

  • Triggers EDITING MODE for seamless editing workflow
  • Only LaTeX files are now treated as editable manuscripts
  • Clear indication that these files will be actively edited

• 📚 Source/Context files (.md, .docx, .pdf) → sources/ folder

  • NEW dedicated folder for reference materials
  • Keep background documents, literature, and context organized
  • Clear separation between reference and editable content

• 📊 Data files (.csv, .json, .xlsx, .txt, etc.) → data/ folder

  • Datasets and structured data files
  • Ready for analysis and integration into papers

• 🖼️ Image files (.png, .jpg, .svg, etc.) → figures/ folder

  • All visual assets organized in one place
  • Automatic integration into paper figures

• 📎 Other files → sources/ folder

  • Any unrecognized file types go to sources for context

📁 New Sources Directory

A brand new sources/ folder is now created for every paper project:

paper_outputs/YYYYMMDD_HHMMSS_topic/
├── drafts/      # Editable manuscripts (.tex)
├── sources/     # ✨ NEW: Reference materials (.md, .docx, .pdf)
├── data/        # Datasets (.csv, .json, .xlsx)
├── figures/     # Images (.png, .jpg, .svg)
├── references/  # Bibliography (.bib)
└── final/       # Published outputs

Benefits:

• Clear distinction between editable manuscripts and reference materials
• Better organization of complex projects with multiple document types
• Easy access to background materials without cluttering other folders
• Improved context management for AI-assisted writing

---

🔧 Key Improvements

Enhanced Editing Mode Detection

EDITING MODE is now more precise and predictable:

• Before: Any manuscript file (.tex, .md, .docx, .pdf) triggered editing mode
• Now: Only .tex files in drafts/ trigger EDITING MODE
• Result: More predictable behavior with clear rules

Why this matters:

• PDF files are often reference materials, not editable sources
• Markdown and Word docs are typically background context
• LaTeX is the primary format for scientific manuscripts
• Users get exactly the behavior they expect

Improved File Processing

Major enhancements to file handling and user feedback:

• Separate Progress Counters - Distinct counts for manuscripts, sources, data, and images
• Clear Indicators - Shows exactly where each file is being copied
• Better Error Handling - More robust error handling during file operations
• Enhanced Feedback - Informative CLI output throughout the process

Example output:

✓ Copied 1 .tex manuscript(s) to drafts/ [EDITING MODE]
✓ Copied 2 source/context file(s) to sources/
✓ Copied 3 data file(s) to data/
✓ Copied 5 image(s) to figures/
✓ Deleted original files from data folder

Documentation Updates

Comprehensive updates to system instructions and user documentation:

• Updated .claude/WRITER.md with new file routing rules
• Enhanced CLI help text explaining file categorization
• Improved welcome message with clear file handling examples
• Better workflow documentation with practical use cases

---

🗑️ Cleanup

Removed Redundant Files

• Removed CLAUDE.md from project root
  • All system instructions now centralized in .claude/WRITER.md
  • Reduces confusion and maintenance overhead
  • Single source of truth for agent instructions

---

💡 Key Benefits

For Users

✅ Better Organization - Files automatically go to the right place every time

✅ Predictable Behavior - Clear rules make the system easy to understand and use

✅ Enhanced Clarity - No more guessing where files will end up

✅ Improved Workflow - Manage complex projects with multiple file types effortlessly

✅ Better Context - Reference materials clearly separated from editable content

For Developers

✅ Cleaner Code - New helper functions for file categorization

✅ Better Extensibility - Easy to add new file types and routing rules

✅ Improved Testing - More testable file processing logic

✅ Enhanced Maintainability - Centralized file type definitions

---

🎯 Usage Example

Here's how the smart file routing works in practice:

# Place various files in the data folder
cp my_paper.tex data/           # → drafts/ (EDITING MODE)
cp background.pdf data/          # → sources/ (REFERENCE)
cp literature_review.docx data/  # → sources/ (REFERENCE)
cp dataset.csv data/             # → data/
cp results.json data/            # → data/
cp figure1.png data/             # → figures/
cp diagram.svg data/             # → figures/

# Run scientific writer
scientific-writer

# The system automatically:
# ✓ Routes .tex to drafts/ and activates EDITING MODE
# ✓ Copies .pdf and .docx to sources/ as reference material
# ✓ Copies .csv and .json to data/ folder
# ✓ Copies .png and .svg to figures/ folder
# ✓ Provides clear feedback for each operation

> "Improve the introduction using the background material from sources/"

---

📝 Technical Details

Files Modified

• scientific_writer/cli.py - Enhanced file routing and user feedback (140+ lines changed)
• scientific_writer/core.py - New file categorization functions (52+ lines added)
• scientific_writer/utils.py - Added sources/ directory scanning
• .claude/WRITER.md - Updated file routing documentation
• scientific_writer/.claude/WRITER.md - Updated file routing rules

New Functions

• get_source_extensions() - Returns source/context file extensions
• get_data_extensions() - Returns data file extensions
• Enhanced process_data_files() - Implements smart routing logic
• Updated scan_paper_directory() - Includes sources/ folder

API Compatibility

✅ Fully backward compatible - No breaking changes to public API

✅ CLI unchanged - All existing commands work identically

✅ Programmatic API preserved - No changes to function signatures

---

🚀 Installation

Install from PyPI

# Install latest version
pip install scientific-writer==2.4.0

# Or with uv
uv pip install scientific-writer==2.4.0

# Or run directly with uvx
uvx scientific-writer@2.4.0

Upgrade from Previous Version

pip install --upgrade scientific-writer

---

📊 Release Statistics

• 8 commits since v2.3.2
• 6 files modified
• 188 insertions, 813 deletions
• Net reduction: 625 lines (improved code efficiency)
• 1 major feature added (Smart File Routing)
• 3 core improvements implemented
• 1 cleanup completed

---

🎓 What's Next?

Looking ahead to future releases:

• Enhanced citation management with automatic DOI lookup
• Improved template system for journals and conferences
• Better integration with reference managers
• Enhanced peer review feedback system
• More output format options

---

🙏 Acknowledgments

Thank you to all users who have provided feedback and suggestions. Your input directly shaped the improvements in this release, particularly around file organization and editing workflows.

---

📚 Additional Resources

• Documentation: docs/FEATURES.md
• API Reference: docs/API.md
• Troubleshooting: docs/TROUBLESHOOTING.md
• Changelog: CHANGELOG.md
• PyPI Package: https://pypi.org/project/scientific-writer/2.4.0/

---

🐛 Bug Reports & Feature Requests

Found a bug or have a feature request? Please open an issue on GitHub:
https://github.com/K-Dense-AI/claude-scientific-writer/issues

---

Full Changelog: v2.3.2...v2.4.0

v2.3.2

[2.3.2] - 2025-11-06

🔧 Improvements

• Package maintenance and version update

---

Installation:

pip install scientific-writer==2.3.2
# or
uv pip install scientific-writer==2.3.2
# or
uvx scientific-writer@2.3.2

PyPI: https://pypi.org/project/scientific-writer/2.3.2/

v2.3.0 - Edit Papers Anywhere 📝

Scientific Writer v2.3.0 - Edit Papers Anywhere 📝

🎉 Write & Edit Scientific Papers Anywhere

This release introduces Manuscript Editing Mode - a powerful new workflow that lets you edit existing papers from anywhere! Simply place your manuscript file (.tex, .md, .docx, .pdf) in the data/ folder, and the system automatically recognizes it as an editing task.

✨ What's New

Automatic Editing Mode Detection

The system now intelligently routes files based on their type:

📄 File Type	📁 Destination	🎯 Purpose
Manuscript files (.tex, .md, .docx, .pdf)	drafts/	Editing existing manuscripts
Image files (.png, .jpg, .svg, etc.)	figures/	Figures for your paper
Data files (.csv, .txt, .json, etc.)	data/	Data for analysis

How It Works

# 1. Place your manuscript in the data folder
cp my_research_paper.tex data/

# 2. Run scientific writer
scientific-writer

# The system automatically:
# ✓ Detects it's a manuscript file
# ✓ Copies it to drafts/ folder
# ✓ Displays [EDITING MODE] indicator
# ✓ Treats this as an editing task

# 3. Request your edits
> "Improve the introduction and add 5 more citations to the methods section"

# Result:
# - Original manuscript preserved
# - New version created (v2_my_research_paper.tex)
# - All changes documented in revision_notes.md

Clear Visual Feedback

When editing mode is active, you'll see clear indicators:

⚠️  EDITING MODE - Manuscript files detected!

📦 Processing files...
   ✓ Copied 1 manuscript file(s) to drafts/ [EDITING MODE]
   ✓ Copied 2 image(s) to figures/
   ✓ Deleted original files from data folder

🔧 TASK: This is an EDITING task, not creating from scratch.

🚀 Key Features

1. Works with Multiple Formats

• LaTeX (.tex) - Your complete paper
• Markdown (.md) - Research notes or drafts
• Word (.docx) - Documents from collaborators
• PDF (.pdf) - Existing publications to revise

2. Smart File Organization

• Manuscripts automatically go to drafts/ folder
• Images automatically go to figures/ folder
• Data files stay in data/ folder
• No manual organization needed!

3. Version Control Built-In

• Original manuscript preserved
• New versions created (v2, v3, etc.)
• Changes documented in revision_notes.md
• Full audit trail of all edits

4. Works in Both CLI and API

• Same behavior in interactive CLI mode
• Same behavior when using programmatic API
• Consistent experience everywhere

💡 Use Cases

Edit a Paper from a Collaborator

# Collaborator sends you draft.docx
cp draft.docx data/
scientific-writer
> "Convert to LaTeX and add citations for all claims in the introduction"

Improve an Existing LaTeX Paper

# Working on your submission
cp neurips_submission.tex data/
scientific-writer
> "Address reviewer comments: strengthen methods section and add ablation study"

Revise Based on Feedback

# Got feedback from advisor
cp thesis_chapter.tex data/
scientific-writer
> "Rewrite the discussion section to address concerns about generalizability"

🔧 Technical Details

Files Modified

• scientific_writer/.claude/WRITER.md - Added manuscript editing workflow instructions
• scientific_writer/core.py - Added manuscript detection and routing logic
• scientific_writer/cli.py - Updated UI with editing mode indicators
• scientific_writer/api.py - Enhanced progress reporting for manuscripts

New Functions

• get_manuscript_extensions() - Defines manuscript file types
• Enhanced process_data_files() - Routes files intelligently
• Enhanced create_data_context_message() - Provides editing mode context

Backward Compatibility

✅ 100% backward compatible

• All existing functionality preserved
• New behavior only activates for manuscript files
• No breaking changes to API or CLI
• Existing workflows continue to work

📦 Installation

# Update to the latest version
cd claude-scientific-writer
git pull origin main
uv sync

# Verify installation
scientific-writer --version  # Should show 2.3.0

📚 Documentation

For detailed information, see:

• CHANGELOG.md - Complete version history
• WRITER.md - System instructions
• README.md - Full documentation

🎯 Example Workflow

Here's a complete workflow showing the power of editing mode:

# You receive a paper draft from a colleague
cp colleague_draft.docx data/

# Start scientific writer
scientific-writer

# Request comprehensive edits
> "Please:
1. Convert this Word document to LaTeX
2. Improve the introduction with better flow
3. Add 10 more recent citations (2023-2024)
4. Strengthen the methods section
5. Create a new figure summarizing the results
6. Format for Nature journal submission"

# The system will:
# ✓ Recognize this as an editing task
# ✓ Read the existing document
# ✓ Apply all requested changes
# ✓ Create properly versioned outputs
# ✓ Document all changes
# ✓ Generate publication-ready LaTeX

🙏 What This Means for You

Before v2.3.0

• Had to manually specify you're editing
• Files could go to wrong folders
• No clear indication of editing vs. creation
• Manual file organization required

After v2.3.0

• ✨ Automatic detection of editing tasks
• ✨ Smart file routing by type
• ✨ Clear visual feedback throughout
• ✨ Zero manual organization needed
• ✨ Edit papers from anywhere!

🎊 Summary

Version 2.3.0 makes it effortless to edit scientific papers from any source:

• Drop manuscript files in data/ folder
• System automatically recognizes editing mode
• Smart routing to correct folders
• Clear feedback and version control
• Works with LaTeX, Markdown, Word, and PDF

Write and edit scientific papers anywhere! 🚀

---

Full Changelog: CHANGELOG.md

Questions or Issues? Open an issue on GitHub or check the documentation.

v2.2.1 - Bug fixes and stability improvements

🔧 Improvements

• Minor bug fixes and stability improvements
• Documentation updates
• Enhanced error handling

Installation

pip install scientific-writer==2.2.1

or

uv add scientific-writer==2.2.1