pip install hangs or takes forever with 'INFO: pip is looking at multiple versions of ... to determine which version is compatible' or 'This could take a while' (dependency backtracking): how do I speed it up?

Cause: pip's resolver picks versions, later finds a conflict, throws that work away and tries other versions. This is called backtracking and is expected behaviour, not a bug, but with loose requirements and many releases it can take a very long time and download many old versions. pip prints messages such as "INFO: pip is looking at multiple versions of <package> to determine which version is compatible with other requirements. This could take a while." Related failures are ResolutionImpossible (the requirements truly conflict) and ResolutionTooDeepError (search depth exceeded). Fix (best first): 1. Find which package it is backtracking on from the repeated Downloading lines, and add a lower bound for it, for example python -m pip install tea "cup>=3.13". Lower bounds exclude old releases and shrink the search. 2. Put limits for indirect dependencies in a constraints file and pass it: python -m pip install -r requirements.txt --constraint constraints.txt. Constraints limit versions but do not cause packages to be installed. 3. Loosen over-tight exact pins in top-level requirements; mutually incompatible pins create conflicts. Use upper bounds sparingly. 4. For repeatable deployments, resolve once and install from a lockfile or fully pinned requirements generated with a tool like pip-tools (pip-compile), then install with pip so no resolution is needed at deploy time. 5. If it ends in ResolutionImpossible, read the conflict summary, then relax or replace the requirement named there. Consider a newer pip, since resolver behaviour keeps improving, and a clean virtual environment. 6. If you hit ResolutionTooDeepError, raise lower bounds and use constraints as above; the pip docs ask for such cases to be reported on pip's issue tracker. Verify: re-run the install; the "looking at multiple versions" messages stop or the install completes quickly, and python -m pip check reports no broken requirements. Sources: https://pip.pypa.io/en/stable/topics/dependency-resolution/ https://pip.pypa.io/en/stable/user_guide/