Repolode
Submit
OpenRaiser

OpenRaiser/NanoResearch

An autonomous research pipeline that takes a topic through ideation, coding, GPU/SLURM experiment runs, analysis, figures and a LaTeX paper.

View on GitHub
Stars
1,338
Forks
91
Open issues
8
Last commit
14 days ago

What is it?

NanoResearch is a Python pipeline that automates an AI research project end to end. Given a topic, it searches literature through OpenAlex and Semantic Scholar, designs an experiment plan, generates runnable code, submits training jobs to a local GPU or a SLURM cluster, parses the resulting logs and metrics, produces figures, and drafts a LaTeX paper that is reviewed and revised before export.

The pipeline runs in nine stages and stores every intermediate artifact in a workspace with a manifest, so a failed run can be resumed from the stage where it stopped. Stages can be routed to different OpenAI-compatible models, and experiment failures trigger log analysis and automatic code repair before retrying. Users can drive it from the command line with a full-screen TUI or plain log output, from a Feishu (Lark) bot, or through Claude Code and Codex with project-level instructions.

The project targets researchers and teams who want a working experiment and draft manuscript from a research idea, with the paper's numbers taken from actual runs. The output is meant as a first draft that still needs human review.

Key features

  • 01Searches literature via OpenAlex and Semantic Scholar during ideation.
  • 02Generates experiment code and submits jobs to local GPUs or SLURM.
  • 03Parses training logs and metrics to build analysis and figures.
  • 04Exports a LaTeX paper with references and code.
  • 05Resumes any stage from the saved workspace manifest.
  • 06Routes each pipeline stage to a different OpenAI-compatible model.

Use cases

  • Turning a research idea into a runnable experiment with a draft paper.
  • Batch-running experiments across several topics for reproducible results.
  • Producing a LaTeX first draft grounded in real training outputs.
  • Keeping an auditable record of every step of an automated research run.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/OpenRaiser/NanoResearch.git
  2. 2

    From inside the NanoResearch folder, install the package in editable mode with dev extras

    pip install -e ".[dev]"
  3. 3

    Create ~/.nanoresearch/config.json with your OpenAI-compatible base_url and api_key

  4. 4

    Validate the configuration with a dry run

    nanoresearch run --topic "Adaptive Sparse Attention Mechanisms" --dry-run

Things to know

  • You must supply your own OpenAI-compatible LLM API endpoint and key to run the pipeline.
  • Actual experiment execution expects a local GPU or a SLURM cluster.

In collections

Open-source AI agent frameworks & appsAI agent skill libraries

Topics

#ai-agents#research-automation#llm#latex#slurm#machine-learning#paper-writing#agent-skills#agents#ai

FAQ

What is NanoResearch?
An autonomous research pipeline that takes a topic through ideation, coding, GPU/SLURM experiment runs, analysis, figures and a LaTeX paper. NanoResearch is a Python pipeline that automates an AI research project end to end. Given a topic, it searches literature through OpenAlex and Semantic Scholar, designs an experiment plan, generates runnable code, submits training jobs to a local GPU or a SLURM cluster, parses the resulting logs and metrics, produces figures, and drafts a LaTeX paper that is reviewed and revised before export.
Is NanoResearch free to use? What license does it have?
NanoResearch is open source under the MIT license. Check the license terms before using it in commercial products.
Is NanoResearch still maintained?
Actively maintained. The last commit to NanoResearch was on Oct 9, 2026.
What can I use NanoResearch for?
Turning a research idea into a runnable experiment with a draft paper. Batch-running experiments across several topics for reproducible results. Producing a LaTeX first draft grounded in real training outputs. Keeping an auditable record of every step of an automated research run.
How do I install NanoResearch?
Clone the repository: git clone https://github.com/OpenRaiser/NanoResearch.git From inside the NanoResearch folder, install the package in editable mode with dev extras: pip install -e ".[dev]" Create ~/.nanoresearch/config.json with your OpenAI-compatible base_url and api_key Validate the configuration with a dry run: nanoresearch run --topic "Adaptive Sparse Attention Mechanisms" --dry-run
Where can I get NanoResearch?
The source code and installation instructions are on GitHub: https://github.com/OpenRaiser/NanoResearch

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects