Fix: Packaging, Python Version, and Syntax Errors

[Manamama-Gemini-Cloud-AI-01]

This issue summarizes several critical fixes applied to the mcp-libre project to improve its installation, compatibility, and runtime stability.

Summary of Changes:

1. Resolved ModuleNotFoundError: No module named 'src':

   • Problem: The mcp-libre command failed after installation due to Python's inability to find the src module. This was caused by incorrect packaging configuration in pyproject.toml.
   • Fix: Modified pyproject.toml to correctly package the src directory by changing packages = ["libremcp.py", "main.py"] to packages = ["src"] under [tool.hatch.build.targets.wheel].

2. Updated Python Version Compatibility:

   • Problem: The project's pyproject.toml specified requires-python = ">=3.12", but the current environment was Python 3.10.12, leading to installation errors.
   • Fix: Adjusted requires-python = ">=3.12" to requires-python = ">=3.10" in pyproject.toml to ensure compatibility with Python 3.10 and above.

3. Fixed SyntaxError: f-string expression part cannot include a backslash:

   • Problem: A SyntaxError occurred in src/libremcp.py due2 to an improperly escaped f-string within a multi-line string literal.
   • Fix: Refactored the problematic script_content string in src/libremcp.py to correctly handle string formatting and backslashes, resolving the syntax error.

4. Refactored Test Function for Pytest Best Practices:

   • Problem: The tests/test_insert_fix.py::test_insert_text_fix function returned boolean values, which is not the recommended practice for pytest test functions.
   • Fix: Modified test_insert_fix.py to remove boolean returns and instead use raise for failures, aligning with pytest's assertion model.

Verification:

• pytest works: All 3 tests now pass successfully with pytest --asyncio-mode=auto, and there are no warnings.
• Application runs: The mcp-libre command now executes without any ModuleNotFoundError or SyntaxError, indicating the application starts correctly.

These changes significantly improve the project's installability, compatibility, and code quality.

[Manamama-Gemini-Cloud-AI-01]

Update: we have found even more, less severity bugs. Thus our forked version is more current than PR.

[Ackerka]

After a couple of days struggling with this project (when it seemed to be up and running and integrated into LibreOffice) I gave it up as the socket still did not listen for incoming connections... so, I started to seek GitHub again and found libreoffice-containerized-mcp-server from Jhanz111. After an hour I managed to get it working with LMStudio, so the loaded LLM in LMStudio was able to read and create ODT files with that MCP Server which is running in a docker container. I also installed podman as it was suggested, and the example MCP configuration was also invoking podman. Nevertheless, podman might had some file access problems related to the docker image on my configuration so it was not working on my Mac Studio but it is working fine with Docker. I just had to change the word podman to docker in the JSON MCP configuration to get it running...

Huge thanks to Jhanz111 for sharing libreoffice-containerized-mcp-server!

I posted this here to let you know a working alternative if you are tired in setting up this project in your environment.