I tried to install the openai-agents package directly from the GitHub repository, but ran into a dependency problem. During installation, I got this error message:
Cannot locate a version that meets the requirement griffe<2,>=1.5.6
Installation was not completed successfully
The installation command I used was:
git clone https://github.com/openai/openai-agents-python.git
cd openai-agents-python
pip install .
I’m trying to figure out if this griffe dependency conflict is what’s causing the whole installation to fail. Has anyone else encountered this same issue when installing openai-agents from source? Is there a way to resolve this dependency problem or should I try a different installation method?
This griffe dependency conflict usually happens when you’ve got an incompatible version cached or already installed. Don’t reinstall everything yet - just clear pip’s cache first with pip cache purge then try installing again. I’ve seen this fix similar dependency issues when pip was hanging onto old package metadata. Another thing that worked for me was using the --upgrade-strategy eager flag, which forces pip to upgrade all dependencies to their latest compatible versions. This works great when packages have recently updated their dependency constraints. Also check if you’re mixing conda packages with pip installs - that can create version conflicts that aren’t obvious at first.
You’ve encountered a griffe version conflict due to an incompatible version already installed. You can verify your current griffe version using pip show griffe. If it’s not the correct version, proceed to uninstall it with pip uninstall griffe, and then try re-installing openai-agents. It should automatically resolve to the correct griffe version. I faced a similar problem last month with an outdated griffe installation causing issues. To overcome this, I executed pip install --force-reinstall . in the cloned directory to force pip to reinstall all dependencies, effectively bypassing version conflicts without needing a clean environment.
I encountered the same griffe dependency issue recently when trying to install openai-agents from the GitHub repository. It generally arises from conflicting versions of installed packages. What worked for me was creating a new virtual environment specifically for the installation. After activating it, I followed up with pip install --upgrade pip and then proceeded with the usual installation commands. This approach resolved the griffe conflict effortlessly. If the problem persists, it might be worth checking if you have mkdocs-material or similar documentation tools installed, as they can sometimes interfere with dependency versions. Additionally, a colleague found success by manually installing the griffe package beforehand using the command pip install "griffe>=1.5.6,<2" before attempting to install openai-agents.
Dependency conflicts like this are exactly why I ditched manual package management years ago. The griffe issue you’re hitting is classic Python ecosystem pain.
Stop fighting pip and virtual environments - just automate this whole thing. Set up something that clones the repo, handles dependencies automatically, and manages multiple Python environments if needed.
I’ve built similar setups for our team when we pull and install packages from various GitHub repos. The automation handles dependency checking, creates clean environments, and can retry with different Python versions if the first attempt tanks.
This way you’re not manually troubleshooting griffe versions every time someone installs from source. Plus you can extend it for other repos with similar issues.
The griffe dependency conflict happens because pip gets confused with version constraints when installing from source. Don’t waste time fighting dependency resolution - just use the --pre flag instead. Run pip install --pre . from the cloned directory. This fixed it for me when I hit similar issues with packages that had strict version requirements. Also check if you’ve got any globally installed packages messing things up - system-wide Python packages can cause weird conflicts. You could also try pip install -e . for an editable install, which sometimes handles dependencies differently than regular installation.
had the same issue last week! try pip install --no-deps . first, then manually install missing dependencies one at a time. worked when nothing else did - griffe kept throwing version conflicts.