Skip to content

FEATURES.md

Latest commit

 

History

History
539 lines (413 loc) · 14.1 KB

FEATURES.md

File metadata and controls

539 lines (413 loc) · 14.1 KB
Raw

Complete Features Guide

This guide provides a comprehensive overview of all features available in Scientific Writer v2.0.

Table of Contents

  1. Document Generation
  2. AI-Powered Capabilities
  3. Intelligent Paper Detection
  4. Data & File Integration
  5. Document Conversion
  6. Developer Features

Document Generation

Scientific Papers

Generate publication-ready papers in LaTeX with proper IMRaD structure.

Supported Venues:

  • Nature, Science - High-impact journals with specific formatting
  • NeurIPS, ICML, ICLR - Machine learning conferences
  • IEEE, ACM - Engineering and computer science
  • Custom venues - Describe any journal or conference

Example Requests:

> Create a Nature paper on CRISPR gene editing
> Write a NeurIPS paper on transformer architectures
> Generate an IEEE paper on signal processing

Output Structure:

paper_outputs/20241030_<topic>/
├── drafts/
│   ├── v1_draft.tex
│   ├── v1_draft.pdf
│   └── v2_draft.tex
├── final/
│   ├── manuscript.tex
│   └── manuscript.pdf
├── references/
│   └── references.bib
├── figures/
│   └── *.png, *.pdf
├── data/
│   └── *.csv, *.json
└── progress.md

Research Posters

Professional conference posters using LaTeX beamerposter, tikzposter, or baposter.

Features:

  • Full-page layouts with minimal margins
  • Color schemes and visual design
  • Accessibility and colorblind-safe palettes
  • Standard sizes (A0, A1, 36×48")
  • Quality control scripts

Example:

> Create a conference poster about my transformer paper
> Generate an A0 poster for NeurIPS with blue color scheme

Grant Proposals

Agency-specific formatting for major U.S. funding sources.

Supported Agencies:

Agency Focus Areas Key Components
NSF Basic research, education Intellectual Merit + Broader Impacts
NIH Biomedical research Specific Aims (1 page), Research Strategy
DOE Energy, physical sciences TRLs, cost sharing, lab partnerships
DARPA High-risk/high-reward Heilmeier Catechism, PM engagement

Example:

> Write an NSF proposal for quantum computing research
> Generate NIH R01 Specific Aims for cancer immunotherapy
> Create a DOE proposal for renewable energy storage

Key Features:

  • Budget preparation and justification
  • Review criteria alignment
  • Timeline and milestone planning
  • Broader impacts strategies (NSF)
  • Preliminary data integration (NIH)

Literature Reviews

Systematic literature reviews with citation management.

Features:

  • Database search strategies (PubMed, Web of Science)
  • Citation verification and formatting
  • Synthesis and organization
  • Multiple citation styles (APA, IEEE, Nature, etc.)

Example:

> Create a literature review on machine learning in healthcare
> Synthesize recent papers on quantum computing from 2023-2024

Clinical Reports

Comprehensive clinical documentation with regulatory compliance.

Report Types:

  • Case Reports - CARE-compliant reports for medical journals
  • Diagnostic Reports - Radiology, pathology, laboratory reports
  • Clinical Trial Reports - SAE reports, Clinical Study Reports (ICH-E3)
  • Patient Documentation - SOAP notes, H&P, discharge summaries

Features:

  • 12 professional templates based on industry standards
  • HIPAA compliance and de-identification tools
  • Regulatory compliance (FDA, ICH-GCP)
  • Medical terminology validation
  • 8 automated validation scripts

Example:

> Create a clinical case report for a rare disease presentation
> Generate a radiology report template for chest CT
> Write an SAE report for clinical trial adverse event
> Create a discharge summary for heart failure patient

Scientific Schematics

Publication-quality diagrams and visualizations.

Diagram Types:

  • CONSORT flowcharts - Clinical trial participant flow
  • Circuit diagrams - Electrical schematics with CircuitikZ
  • Biological pathways - Signaling cascades, metabolic networks
  • System architecture - Block diagrams, data flow
  • Process flows - Methodology diagrams, decision trees

Features:

  • Vector graphics (TikZ/LaTeX)
  • Colorblind-safe Okabe-Ito palette
  • Programmatic generation with Python
  • Publication-ready PDF/SVG/PNG output

Example:

> Create a CONSORT diagram for my clinical trial
> Generate a circuit diagram for an RC low-pass filter
> Design a biological pathway showing MAPK signaling

AI-Powered Capabilities

Real-Time Research Lookup

Powered by Perplexity Sonar Pro via OpenRouter API.

Features:

  • Live internet search during paper generation
  • Recent publications and preprints
  • Fact-checking and verification
  • Up-to-date statistics and data

Setup:

# Add to .env file
echo "OPENROUTER_API_KEY=your_key" >> .env

Automatic Usage: The research lookup is automatically invoked when:

  • You request recent research (e.g., "papers from 2024")
  • Claude needs to verify facts or statistics
  • Literature search is needed
  • Current events or recent developments are mentioned

Example:

> Create a paper on recent advances in quantum computing (2024)
# Automatically searches for latest research

> What are the current success rates for CAR-T therapy?
# Looks up latest clinical data

Peer Review with ScholarEval

Systematic quantitative evaluation framework based on research (arXiv:2510.16234).

8 Evaluation Dimensions:

  1. Problem Formulation - Clarity, significance, novelty
  2. Literature Review - Coverage, synthesis, gaps
  3. Methodology - Rigor, validity, reproducibility
  4. Data Collection - Quality, appropriateness, ethics
  5. Analysis - Statistical rigor, interpretation
  6. Results - Clarity, completeness, visualization
  7. Writing Quality - Structure, clarity, grammar
  8. Citations - Relevance, recency, completeness

Scoring:

  • 1-5 scale per dimension
  • Overall score thresholds:
    • 4.5+ - Exceptional (top-tier ready)
    • 4.0-4.4 - Strong (minor revisions)
    • 3.5-3.9 - Good (major revisions)
    • 3.0-3.4 - Acceptable (significant revisions)
    • <3.0 - Needs major rework

Example:

> Evaluate this paper using the ScholarEval framework
> Assess publication readiness for Nature Machine Intelligence
> Review the methods section for rigor and completeness

Iterative Editing

Context-aware revision suggestions that understand your paper's structure and content.

Features:

  • Maintains consistency across sections
  • Preserves your writing style
  • Tracks changes between versions
  • Suggests improvements based on venue requirements

Example:

> Improve the introduction to better motivate the problem
> Make the discussion more concise
> Add transition sentences between paragraphs in the results

Intelligent Paper Detection

Automatically identifies when you're referring to existing papers without needing to specify paths.

How It Works

The system analyzes your input for:

  1. Continuation keywords: "continue", "update", "edit", "the paper"
  2. Search keywords: "find", "look for", "show me", "where is"
  3. Topic matching: Keywords from paper directory names
  4. Temporal context: Defaults to most recent paper when ambiguous

Continuation Keywords

Automatically resumes work on the current or most recent paper:

> continue
> update the paper
> edit my paper
> add a conclusion section
> fix the references
> compile the poster
> generate the PDF

Search Keywords

Finds specific papers by topic:

> find the acoustics paper
> look for the quantum computing paper
> show me the CRISPR paper
> where is the transformer paper

Topic Matching

Matches based on directory names (format: YYYYMMDD_HHMMSS_topic):

# Finds 20241027_090109_acoustics_vinayak_agarwal/
> update the acoustics paper

# Finds 20251029_130950_transformers_ai_paper/
> continue working on the transformers paper

Starting a New Paper

Explicitly start fresh:

> new paper on climate change
> start fresh with a different topic
> create a new paper about quantum computing

Data & File Integration

Automatic Data Handling

Simply drop files into the data/ folder at the project root.

File Routing:

  • Images (png, jpg, svg, etc.) → figures/
  • Data files (csv, json, txt, xlsx) → data/
  • Original files automatically deleted after copying

Supported Image Formats: .png, .jpg, .jpeg, .gif, .bmp, .tiff, .svg, .webp, .ico

Example Workflow:

# 1. Copy your files
cp experiment_results.csv ~/Documents/claude-scientific-writer/data/
cp performance_graph.png ~/Documents/claude-scientific-writer/data/

# 2. Start the CLI
scientific-writer

# 3. Files are automatically detected and processed
> Create a paper analyzing the experimental results
# ✓ Files copied to paper's data/ and figures/
# ✓ Original files deleted from data/

Programmatic Data Files

Use the API to specify data files explicitly:

async for update in generate_paper(
    query="Analyze the experimental results",
    data_files=[
        "./experiment_results.csv",
        "./figures/performance.png",
        "./supplementary_data.json"
    ]
):
    # Files are processed and included
    pass

Note: When using the API, original files are not deleted (for safety).

File Context

All included files are:

  • Copied to appropriate directories
  • Made available as context to Claude
  • Can be referenced in the paper
  • Listed in the data files message

Document Conversion

MarkItDown - Universal File Converter

Convert 15+ file formats to Markdown for LLM processing.

Supported Formats:

  • Documents: PDF, DOCX, PPTX, XLSX
  • Media: Images (with AI descriptions), Audio (transcription)
  • Web: HTML, YouTube (video transcription)
  • Data: CSV, JSON, XML
  • Code: Various programming languages

Features:

  • AI-enhanced image descriptions using Claude Sonnet 4.5
  • OCR for scanned documents
  • Speech-to-text for audio files
  • Batch processing with parallel execution
  • Scientific literature metadata extraction

Example:

> Convert all PDFs in the literature folder to Markdown
> Extract data from this Excel spreadsheet
> Transcribe the interview audio file
> Convert this PowerPoint to Markdown with AI descriptions

Document Manipulation

DOCX (Word Documents)

  • Create and edit Word documents programmatically
  • Manage comments and track changes
  • Validate document structure
  • Work with templates

PDF Documents

  • Extract text and metadata
  • Analyze PDF layout and bounding boxes
  • Fill PDF forms
  • Convert PDFs to images

PPTX (PowerPoint)

  • Create and modify presentations
  • Convert HTML to PowerPoint
  • Generate thumbnails
  • Manage slides and layouts

XLSX (Excel)

  • Read and write Excel files
  • Recalculate formulas
  • Handle complex spreadsheet operations

Developer Features

Programmatic API

Full async Python API with comprehensive type hints.

Core Function:

from scientific_writer import generate_paper

async def generate_paper(
    query: str,
    output_dir: Optional[str] = None,
    api_key: Optional[str] = None,
    model: str = "claude-sonnet-4-20250514",
    data_files: Optional[List[str]] = None,
    cwd: Optional[str] = None,
) -> AsyncGenerator[Dict[str, Any], None]

Type-Safe Models:

from scientific_writer import (
    ProgressUpdate,  # Progress information
    PaperResult,     # Final result with all paper info
    PaperMetadata,   # Paper metadata (title, date, word count)
    PaperFiles,      # All file paths (PDF, TeX, BibTeX, etc.)
)

Progress Streaming

Real-time updates during generation:

async for update in generate_paper("Create a paper"):
    if update["type"] == "progress":
        # Progress update
        stage = update["stage"]        # initialization|research|writing|compilation|complete
        message = update["message"]    # Human-readable message
        details = update.get("details")  # Optional: tool name, files created, etc.
        
        print(f"[{stage}] {message}")

Comprehensive Results

Final result includes everything about the generated paper:

if update["type"] == "result":
    # Status
    status = update["status"]  # success|partial|failed
    
    # Files
    pdf = update["files"]["pdf_final"]
    tex = update["files"]["tex_final"]
    bib = update["files"]["bibliography"]
    figures = update["files"]["figures"]  # List of figure paths
    
    # Metadata
    title = update["metadata"]["title"]
    word_count = update["metadata"]["word_count"]
    
    # Citations
    citation_count = update["citations"]["count"]
    
    # Compilation
    compiled = update["compilation_success"]
    
    # Errors (if any)
    errors = update["errors"]

Error Handling

Graceful error handling with detailed information:

try:
    async for update in generate_paper(query):
        if update["type"] == "result":
            if update["status"] == "failed":
                print(f"Errors: {update['errors']}")
            elif update["status"] == "partial":
                print("TeX created but PDF compilation failed")
            else:
                print("Success!")
except ValueError as e:
    print(f"Configuration error: {e}")

Custom Configuration

Override defaults for your use case:

async for update in generate_paper(
    query="Create a paper",
    output_dir="./my_custom_directory",
    api_key="sk-ant-custom-key",
    model="claude-sonnet-4-20250514",
    data_files=["data.csv"],
    cwd="/path/to/project"
):
    pass

See Also