Skip to content

INSTALL.md

Latest commit

 

History

History
174 lines (123 loc) · 5.46 KB

INSTALL.md

File metadata and controls

174 lines (123 loc) · 5.46 KB
Raw

Astro-AI Environment Setup Guide

Based on CAM-SUST Summer School Installation Guide

Option 1: Using pip (Current approach)

pip install -r requirements.txt

Option 2: Using conda (Recommended for astronomical packages)

# Create environment
conda create -n astro-ai python=3.10
conda activate astro-ai

# Install conda packages first
conda install -c conda-forge numpy pandas matplotlib scipy astropy h5py
conda install -c conda-forge streamlit seaborn

# Install pip packages
pip install bagpipes
pip install openai>=1.0.0
pip install pytest>=7.0.0 black>=23.0.0

Option 3: Full astronomical setup (Advanced)

# Create environment for full astronomy stack
conda create -n astro-ai-full python=3.10
conda activate astro-ai-full

# Install system dependencies (if on Linux/WSL)
# sudo apt-get install libfftw3-dev libgsl-dev build-essential

# Install conda packages
conda install -c conda-forge cython numpy pandas matplotlib scipy astropy h5py
conda install -c conda-forge streamlit seaborn

# Install specialized astronomy packages
pip install bagpipes
pip install tools21cm
pip install nautilus-sampler

# Optional: Install 21cmFAST (Linux/WSL only)
pip install git+https://github.com/21cmfast/21cmFAST.git

# Optional: Install JWST pipeline (requires C++ compiler)
# conda install -c conda-forge jwst

# Install remaining packages
pip install openai>=1.0.0
pip install pytest>=7.0.0 black>=23.0.0

Environment Files

For reproducible installations, you can also create:

  • environment.yml (for conda)
  • requirements.txt (for pip)

Deployment Guidance & Eliminating Runtime Warnings

The application supports optional heavy scientific dependencies. In production you can choose a capability tier:

Tier Intended Use Required Packages
Minimal Demo, AI summaries only core requirements (see requirements.txt), NO py21cmfast, NO jwst
Science-Light Add Bagpipes SED fitting + bagpipes (already in requirements)
Full 21cm Simulated reionization + py21cmfast (+ tools21cm optional)
JWST + SED JWST pipeline reduction + jwst + bagpipes
Complete Everything all above

Recommended Strategy

  1. Build a base Docker / conda image for the Minimal or Science-Light tier (fast CI).
  2. Create a separate image for Full science features (longer build) used only where needed.
  3. Use environment variable ASTRO_AI_STRICT=1 (you can add logic) to enforce mandatory capabilities.

Strict Mode (in app)

The sidebar "Environment Status" expander offers a checkbox for Strict Mode which raises an error if required modules are missing. Adjust required list in app.py:

all_required_or_raise(["py21cmfast", "bagpipes", "jwst_pipeline", "astropy"])

Edit this list per deployment tier.

Suppressing Warnings Cleanly

All module availability messages now use logging. Configure log level via environment:

export ASTRO_AI_LOG_LEVEL=WARNING  # Linux / WSL

Or adapt _configure_logging() in app.py to read that.

Installing Heavy Dependencies

Package Notes
py21cmfast Prefer Linux/WSL; needs FFTW, GSL. Use conda for libs then pip install.
jwst Large; adds CRDS downloads. Set CRDS_PATH & CRDS_SERVER_URL.
bagpipes Pure Python + some scientific stack; LaTeX optional for nicer plots.
tools21cm Utility library; optional for extra analysis.
nautilus-sampler Alternative nested sampler if PyMultiNest unavailable.

JWST Pipeline Environment Variables

export CRDS_PATH="$HOME/.crds"
export CRDS_SERVER_URL="https://jwst-crds.stsci.edu"

On Windows PowerShell:

[System.Environment]::SetEnvironmentVariable('CRDS_PATH', "$Env:USERPROFILE\.crds", 'User')
[System.Environment]::SetEnvironmentVariable('CRDS_SERVER_URL', 'https://jwst-crds.stsci.edu', 'User')

LaTeX (Bagpipes Plot Quality)

Install a lightweight TeX distro (MiKTeX on Windows) or disable TeX in code:

import matplotlib as mpl
mpl.rcParams['text.usetex'] = False

Docker (Example Skeleton)

FROM mambaorg/micromamba:1.5.8
COPY environment.yml /tmp/environment.yml
RUN micromamba env create -f /tmp/environment.yml -y && \
	micromamba clean --all --yes
ENV MAMBA_DOCKERFILE_ACTIVATE=1
ENV OPENAI_API_KEY=changeme
WORKDIR /app
COPY . /app
EXPOSE 8501
CMD ["python", "-m", "streamlit", "run", "app.py", "--server.port", "8501", "--server.address", "0.0.0.0"]

Preflight CI Check (Optional)

Add a small script to assert required tier before deployment:

# scripts/preflight.py
from utils.feature_flags import detect_capabilities, all_required_or_raise
all_required_or_raise(["bagpipes"])  # minimal tier example
print("Preflight passed")

Run in CI:

python scripts/preflight.py

OpenAI Client Version

The codebase expects openai>=1.0.0. Ensure the deployed image doesn’t pin an older cached layer.