This pull request introduces several updates to streamline the development workflow, enhance documentation, improve the manifest.json configuration, and add new features for building and distributing the Lobster DXT extension. Below is a summary of the most important changes:

Development Workflow Enhancements:

• Added .dxtignore file to exclude unnecessary files from the package, such as development, build, test, and large data files. This ensures a cleaner and more efficient distribution.
• Removed the .github/workflows/build-dxt.yml file, transitioning to a new approach for building and releasing the DXT extension. Builds are now handled using the official DXT CLI and documented in docs/DXT_DISTRIBUTION.md.

Documentation Updates:

• Updated docs/DXT_DISTRIBUTION.md to reflect the new build and distribution process, including instructions for using the DXT CLI (dxt pack) and manual GitHub Actions triggers for testing. [1] [2] [3]

Configuration Improvements:

• Enhanced manifest.json with additional metadata, including display_name, long_description, and support fields. Updated the server entry point to use a bash script (run_mcp_server.sh) instead of a direct Python file. Added new tools for sequence analysis, concept interventions, and naturalness scoring. Changed the license to Apache-2.0. [1] [2]

Feature Additions:

• Added a new Slurm script slurm/scripts/eval_ume.sh for GPU-based evaluation of biological sequences using the Lobster DXT extension. Includes configurations for memory, batch size, and sequence length.

Setup Improvements:

• Updated src/lobster/mcp/setup.py to check for the availability of the uv package and provide alternative instructions for using the DXT extension. Added warnings and fallback options for users without uv. [1] [2] [3]