A private-by-default expert panel in a box for researchers, independent thinkers, and R&D teams.
Ask a raw research question. The R.A.I.N. Lab assembles multiple expert perspectives, grounds strong claims in papers or explicit evidence, and returns the strongest explanations, disagreements, and next moves.
Most tools help you find papers. R.A.I.N. Lab helps you think with a room full of experts.
James is the assistant inside the R.A.I.N. Lab.
🌐 简体中文 • 日本語 • Русский • Français • Tiếng Việt
The R.A.I.N. Lab turns one question into a structured research conversation.
- It frames the problem from multiple expert angles.
- It separates strong evidence from weak speculation.
- It shows agreements, disagreements, and open questions instead of forcing false certainty.
- It helps you decide what to read next, test next, or ask next.
This is built for work that starts messy: early-stage research, technical due diligence, strategy exploration, independent investigation, and R&D planning.
Start with the example hosted experience:
If you want the fastest way to feel the product, start there first.
Run the local experience on your own machine:
python rain_lab.pyPress Enter for demo mode, or continue into guided setup.
On Windows, you can also double-click INSTALL_RAIN.cmd to create shortcuts.
On macOS/Linux, run ./install.sh for a one-click setup.
- Researchers working through ambiguous questions
- Independent thinkers building evidence-backed views
- R&D teams comparing explanations, risks, and next moves
- Technical operators who want private workflows and inspectable reasoning
| Use case | What R.A.I.N. Lab helps you do |
|---|---|
| Pressure-test a research claim | Compare competing explanations and inspect where the evidence is thin |
| Map a new topic fast | Turn a vague question into viewpoints, sources, disagreements, and next steps |
| Prepare decisions | Surface trade-offs, unresolved risks, and what would change the conclusion |
| Stay private | Keep your local workflow and model setup on your own machine when needed |
Most research tools optimize for retrieval. R.A.I.N. Lab is designed for synthesis, challenge, and judgment.
| Typical tool behavior | R.A.I.N. Lab behavior |
|---|---|
| Returns a list of papers or links | Returns competing interpretations and strongest next moves |
| Treats the first plausible answer as good enough | Preserves disagreements and uncertainty where it matters |
| Hides reasoning behind one-shot summaries | Makes evidence, gaps, and confidence easier to inspect |
| Assumes cloud-first workflows | Supports local and private usage paths |
If you want the product to run from your machine with your own setup:
- Launch the app:
python rain_lab.py- For guided setup, run:
python rain_lab.py --mode first-run- For a first structured prompt, try:
python rain_lab.py --mode beginner --topic "compare the strongest arguments for and against a local-first research workflow"The guided flow can connect to LM Studio or Ollama so your model traffic stays local.
- Multi-perspective research synthesis
- Evidence-aware reasoning with explicit uncertainty
- Guided next steps for reading, testing, and follow-up questions
- Private local workflow options
- Available in 6 languages: 中文, 日本語, Русский, Français, Tiếng Việt, English
- Python 3.10+
- Optional: LM Studio or Ollama for local AI models
- Optional: ZeroClaw/Rust toolchain for the fast runtime layer
Python works without the optional pieces. Adding them expands the local/private path.
Click to expand
If you want to contribute to R.A.I.N. Lab or run the developer setup locally:
git clone https://github.com/topherchris420/james_library.git
cd james_library
# Python setup
uv python install 3.12
uv venv .venv --python 3.12
uv pip sync --python .venv/bin/python requirements-dev-pinned.txt
# Rust setup (optional, for the fast runtime layer)
cargo build --release --locked
# Run
uv run --python .venv/bin/python rain_lab.py --mode first-runRecommended mental model:
- R.A.I.N. Lab is the experience.
- James is the assistant you interact with inside the lab.
- Python handles launcher flows and orchestration.
- ZeroClaw/Rust handles the fast runtime, tool surface, and lower-level infrastructure.
Testing:
ruff check .
pytest -q
cargo fmt --all
cargo clippy --all-targets -- -D warningsSee ARCHITECTURE.md and CONTRIBUTING.md for contributor details.
MIT. Built by Vers3Dynamics, special thanks to ZeroClaw.