An agentic AI ecosystem for human-led, natural-language software development.
You describe what you want in plain English; an organism of AI agents plans, builds, tests and releases it — with you making every judgment call. PyAutoLabs is where that ecosystem, PyAutoScientist, develops open-source astronomy software for Bayesian model-fitting, galaxy morphology and strong gravitational lensing, every day.
PyAutoScientist is a working reference implementation of a human-led AI development organism: a set of repositories through which plain-English intent becomes tested, released software, with a human directing the work and checkpointing every decision that matters. It is not a framework you install — it is the live system that develops everything below, documented so you can fork it and lead your own.
| Organ | Repo | Role |
|---|---|---|
| Mind | PyAutoMind | Where you lead: every piece of work starts here as a plain-English markdown file saying what to do. |
| Brain | PyAutoBrain | The reasoning layer that works out how — classifying, planning and routing each task through specialist agents. |
| Heart | PyAutoHeart | The health monitor whose GREEN/YELLOW/RED verdict is the authoritative "is it safe to release?" gate. |
| Hands | PyAutoBuild | The executor that packages, tags and releases the libraries to PyPI, nightly. |
| Memory | PyAutoMemory | Long-term scientific knowledge — cross-linked literature wikis the agents consult. |
| Nerves | PyAutoConf | The configuration and serialization layer (autoconf) connecting the organism's conventions to every library. |
The software the organism develops is organised as three families, each a library plus the repos a working scientist needs around it.
Probabilistic programming: model composition, non-linear search and Bayesian inference (pip install autofit).
| Repo | Role |
|---|---|
| PyAutoFit | The instrument — the model-fitting and statistical inference library itself. |
| autofit_workspace | Where the scientist works — example scripts, pipelines and configuration. |
| HowToFit | The classroom — narrative lectures teaching model fitting from first principles. |
| autofit_assistant | The AI research assistant — point Claude or ChatGPT at it and ask. |
| autofit_workspace_test | The referee — regression checks that every result still reproduces. |
Strong gravitational lens modeling, from Hubble to Euclid and JWST (pip install autolens).
| Repo | Role |
|---|---|
| PyAutoLens | The instrument — the strong-lens modeling library itself. |
| autolens_workspace | Where the lensing scientist works — example scripts, pipelines and datasets. |
| HowToLens | The classroom — narrative lectures teaching lens modeling from first principles. |
| autolens_assistant | The AI research assistant — point Claude or ChatGPT at it and ask. |
| autolens_workspace_test | The referee — regression checks that every result still reproduces. |
| autolens_workspace_developer | The back-room workbench — developer scripts and experiments. |
| autolens_profiling | The stopwatch — JAX likelihood performance runs and results. |
Multi-wavelength modeling of galaxy light, mass and morphology (pip install autogalaxy).
| Repo | Role |
|---|---|
| PyAutoGalaxy | The instrument — the galaxy structure and morphology library itself. |
| autogalaxy_workspace | Where the galaxy scientist works — example scripts, pipelines and datasets. |
| HowToGalaxy | The classroom — narrative lectures teaching galaxy modeling from first principles. |
| autogalaxy_workspace_test | The referee — regression checks that every result still reproduces. |
(An autogalaxy_assistant is on the way.)
Under the hood sit PyAutoArray (data structures, grids and inversions) and PyAutoReduce (reducing archival telescope imaging into modeling-ready datasets).
pip install autolens
git clone https://github.com/PyAutoLabs/autolens_workspace
cd autolens_workspace
python welcome.pyDocumentation for every project is collected at pyautolabs.github.io.
If you use PyAutoLabs software in your research, please cite:
Nightingale, J. W. et al. (2021). PyAutoLens: Open-Source Strong Gravitational Lensing. JOSS, 6(58), 2825.
All PyAutoLabs packages are released under the MIT License.