I’m trying to set up localGPT-Vision on Windows 10 but keep running into dependency issues with packages that require C/C++ or Rust libraries. These problems block document indexing and prevent the demo from running.
Issues Faced
1. python-poppler via docx2pdf
· Installing python-poppler fails with:
ERROR: Dependency "poppler-cpp" not found
· Tried conda install -c conda-forge poppler and choco install poppler, but python-poppler still couldn't find poppler-cpp.
· As a workaround, replaced docx2pdf with:
· mammoth → DOCX → HTML
· WeasyPrint → HTML → PDF
2. vllm dependency (outlines_core)
· pip install -r requirements.txt fails because outlines_core needs the Rust compiler.
· Rust setup on Windows is complicated, so vllm was removed for a CPU-only environment.
3. WeasyPrint GTK+ libraries
· Running python app.py fails with:
OSError: cannot load library 'libgobject-2.0-0': error 0x7e
· Tried choco install gtk3 --yes but installation fails even in admin mode.
· WeasyPrint docs recommend manual GTK+ installation and updating PATH, but this is difficult on Windows.
Impact
Because of these issues:
• Document indexing doesn’t work.
• The demo app can’t run at all.
Request for Help
Could you provide:
1. A reliable Windows setup guide for these dependencies, or
2. Alternative libraries for DOCX/HTML to PDF conversion that are easier to install on Windows, or
3. Confirmation if WSL2/Docker/Linux is the recommended way to run this project for a CPU-only setup.
I’m trying to set up localGPT-Vision on Windows 10 but keep running into dependency issues with packages that require C/C++ or Rust libraries. These problems block document indexing and prevent the demo from running.
Issues Faced
1. python-poppler via docx2pdf
· Installing python-poppler fails with:
ERROR: Dependency "poppler-cpp" not found
· Tried conda install -c conda-forge poppler and choco install poppler, but python-poppler still couldn't find poppler-cpp.
· As a workaround, replaced docx2pdf with:
· mammoth → DOCX → HTML
· WeasyPrint → HTML → PDF
Impact
Because of these issues:
• Document indexing doesn’t work.
• The demo app can’t run at all.
Request for Help
Could you provide:
1. A reliable Windows setup guide for these dependencies, or
2. Alternative libraries for DOCX/HTML to PDF conversion that are easier to install on Windows, or
3. Confirmation if WSL2/Docker/Linux is the recommended way to run this project for a CPU-only setup.