dark-research-lab 0.2.0

Dark Research Lab - autonomous research paper factory

DRL -- Dark Research Lab

Autonomous research paper factory for social science. Turn a git repo into a reproducible academic paper with AI-driven analysis, literature indexing, and LaTeX compilation.

Install

Requires Python 3.10+ and Claude Code.

# Install DRL
uv tool install dark-research-lab

# Install beads (task tracker)
npm install -g beads

Quick Start

# 1. Create a project and scaffold
mkdir my-paper && cd my-paper && git init
drl setup

# 2. Add your data and literature
cp ~/data/*.csv data/input/          # Keep originals safe elsewhere
cp ~/papers/*.pdf literature/pdfs/
drl index                            # Index PDFs for agent search

# 3. Configure for your field
/drl:flavor

# 4. Decompose your research question into epics
/drl:architect

# 5. Run the pipeline on each epic
/drl:cook-it <epic-id>

# Or run autonomously
drl loop --force && screen -dmS loop ./infinity-loop.sh

See docs/drl/ONBOARDING.md for the full setup walkthrough.

How It Works

DRL wraps compound-agent with research-specific skills, agents, and guardrails:

Researcher
    |
    v
 drl CLI (Go binary in a Python wheel)
    |
    +-- Claude Code (executes skills/agents)
    +-- Beads (epic tracking with dependency graphs)
    +-- Literature RAG (PDF extraction + embedding via ca-embed)
    +-- LaTeX toolchain (3-pass pdflatex + bibtex)
    +-- Advisory Fleet (optional: Gemini, Codex reviewers)

Each research question passes through a cook-it cycle:

1. Spec -- research question, hypotheses, literature gap
2. Plan -- methodology, variables, statistical models
3. Work -- analysis, tables, figures, section drafting
4. Review -- methodology audit + external model review
5. Synthesis -- lessons captured, paper section finalized

Every methodological decision is logged to docs/decisions/ for full traceability. A reproducibility package (lockfile + data manifest + run script) is generated at compilation time.

Project Structure

paper/          LaTeX source and compiled outputs
src/            Analysis scripts
literature/     PDFs and indexed knowledge base
docs/           Decisions, specs, agent notes
tests/          Test suite
.claude/        Skills, agents, hooks, commands

Commands

Command	Purpose
drl setup	Initialize or update project templates
drl index	Index literature PDFs for RAG search
drl loop	Run infinity loop over all epics
/drl:compile	Compile LaTeX paper + reproducibility package
/drl:flavor	Customize skills for your research field
/drl:onboard	Guided first-time setup
/drl:architect	Decompose research question into epics

Documentation

• System Specification
• Architecture Decisions
• Agent Configuration

License

MIT