requirements.txt

# Paperfessor core runtime dependencies.
#
# WARNING: `pip install -r requirements.txt` installs the minimum
# versions below into the ACTIVE environment and MAY UPGRADE OR
# DOWNGRADE packages you already have (litellm, numpy, matplotlib,
# and pydantic in particular pull large dependency trees). To avoid
# disturbing an existing environment, install into a fresh virtualenv
# or conda env first:
#
#   python -m venv .venv && . .venv/bin/activate   # or conda create -n paperfessor python=3.11
#   pip install -r requirements.txt
#
# Or just `pip install paperfessor`, which resolves the same set
# through the package metadata.
#
# Optional extras are NOT listed here — install them explicitly:
#   GUI:            pip install "paperfessor[gui]"     (PyQt6)
#   web/browsing:   pip install "paperfessor[web]"     (playwright, then `playwright install`)
#   development:    pip install "paperfessor[dev]"     (pytest, ruff, mypy)
#
# OPTIONAL SCIENTIFIC PACKAGES the Undergraduate agent may use for
# some research topics (NOT installed by default — kept out of the
# core so a plain install stays lightweight):
#   torch            deep-learning experiments / GPU acceleration
#   opencv-python    image / vision topics
#   pandas           tabular data handling
#   scipy statsmodels statistics / signal processing
#   networkx         graph topics
#
# AUTO-INSTALL WARNING: if the UG has install permission
# (PAPERFESSOR_UG_ALLOW_INSTALLS=true, the DEFAULT), it will pip-install
# any package it needs DURING paper generation — into the active
# environment. Every such install is logged to
# workspace/src/tools/installed.txt. To forbid it, set
# PAPERFESSOR_UG_ALLOW_INSTALLS=false (see the README permissions
# section). Running inside a dedicated venv/conda env is strongly
# recommended so auto-installs never disturb your other projects.

typer>=0.12.0
rich>=13.7.0
pydantic>=2.6.0
pydantic-settings>=2.2.0
httpx>=0.27.0
litellm>=1.40.0
keyring>=24.3.0
platformdirs>=4.2.0
PyYAML>=6.0.1
cryptography>=42.0.0
tenacity>=8.2.3
tiktoken>=0.7.0
requests>=2.31.0
numpy>=1.26.0
scikit-learn>=1.4.0
matplotlib>=3.8.0
pypdf>=4.0.0
pypdfium2>=4.28.0
pillow>=10.0.0