Apps
Poetry Cache
Poetry keeps a local cache of package index metadata and downloaded distribution artifacts so dependency resolution and installs do not need to fetch the same information every time. Those folders typically contain cached JSON responses from package sources plus wheel and source distribution archives such as .whl and .tar.gz files. Kudu clears Poetry's metadata and artifact cache while leaving project files, lockfiles, configured repositories, and any virtualenvs untouched.
Why clean Poetry Cache?
- Corrupted wheel or sdist archives in Poetry's artifacts cache can make installs fail with hash mismatches, unzip errors, or repeated retries for the same package
- Stale package index metadata can leave Poetry resolving against outdated version information, so you keep seeing 'could not find a matching version' until the cache is refreshed
- Large artifacts from repeated dependency updates accumulate under the cache directory and show up as unexpectedly high disk usage even though old versions are no longer needed
- Switching package sources or fixing a private index often still uses cached metadata from the previous state, which can cause 404 or authorization-related install failures until the cache is cleared
- Interrupted downloads may leave partial archives behind, leading to installs that repeatedly fail on one dependency while fresh downloads succeed after cleaning
- Resolver behavior can look inconsistent across machines when one system is using older cached repository responses, causing confusing differences in selected versions and lock refresh results
What gets cleaned
Cache paths Kudu targets
Windows
%LocalAppData%/pypoetry/Cache/cache |
%LocalAppData%/pypoetry/Cache/artifacts |
macOS
~/Library/Caches/pypoetry/cache |
~/Library/Caches/pypoetry/artifacts |
Linux
~/.cache/pypoetry/cache |
~/.cache/pypoetry/artifacts |
Frequently asked
Common questions about Poetry Cache
Related cleaners
Free & open source
Download Kudu and reclaim your disk space.
Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.