Fix #19: add dataset collection support and robust collection detection

[ldalcolmo]

Description

This PR fixes Issue #19 by adding native support for dataset collections in Galaxy MCP, enabling agent workflows to correctly discover, inspect, and navigate histories that contain collections.

It introduces:

• Proper exposure of dataset collections in get_history_contents()
• A new MCP tool get_collection_details() for inspecting collection structure and members
• A robust, API-based guard in get_dataset_details() to prevent agents from treating collections as datasets

These changes make Galaxy MCP fully usable in agent-native environments where histories frequently contain list, paired, or nested dataset collections.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📚 Documentation update
• ⬆️ Dependency update
• 🧰 Maintenance/chore

Checklist

• I have performed a self-review of my code
• I have added tests for my changes
• I have updated the documentation accordingly
• My changes generate no new warnings

Related Issues

Closes #19

[ldalcolmo]

This PR fixes Issue #19 and enables robust handling of dataset collections in Galaxy MCP, which is critical for agent-native workflows and LLM integrations.

What it does:

• get_history_contents() now returns both datasets and dataset collections with a clear history_content_type field.
• Added the new get_collection_details() tool, which returns normalized collection and member dataset info for agent traversal.
• get_dataset_details() now includes an API-based guard to detect collection IDs and provide a helpful error message, avoiding fragile text parsing.

Tests included:

• Mixed history contents
• Truncation for large collections
• Correct guard behavior when passing a collection ID
• All existing tests remain untouched and pass

This change is backward-compatible and purely additive, and should help unblock agent workflows that need to navigate collections.

Happy to make any adjustments if needed! 🙌

[bgruening]

Welcome @ldalcolmo! I started the CI :)

[ldalcolmo]

Welcome @ldalcolmo! I started the CI :)

Thanks @bgruening!
I cleaned up all ruff/E501 and formatting issues in a few follow-up commits so CI should be green now.
The functional changes are in the initial commit (collections + guards + new MCP tool), the others are formatting only.
Happy to adjust anything if you have suggestions.

[dannon]

Thanks for this contribution! Adding collection support is a welcome improvement!

A couple of things I noticed:

1. Performance consideration: The change from gi.datasets.get_datasets() (server-side pagination) to gi.histories.show_history(contents=True) means we now fetch all items and paginate client-side. For large histories this could be slower, though I understand it's necessary to include collections. Might be worth a comment noting this tradeoff.

2. Unused parameter: The details parameter in get_history_contents() is no longer used in the new implementation

3. Minor test nit: A few test methods have their docstrings after the first line of code rather than immediately after the signature (e.g., test_get_collection_details_list_collection).

The new get_collection_details() tool looks well-structured, and the error handling that detects when someone passes a collection ID to get_dataset_details() is a nice touch for agent UX.

[ldalcolmo]

Thanks a lot for the review and the kind words! 🙏
Great points.

Performance: agreed — switching from gi.datasets.get_datasets() (server-side pagination) to show_history(contents=True) does fetch all items and paginates client-side. I’ll add an explicit comment/docstring note explaining the tradeoff and why it’s needed to include dataset collections.

Unused details parameter: good catch — it’s now redundant in the new implementation. I’ll remove it (and update call sites/tests accordingly).

Test docstring placement: also agreed — I’ll move those docstrings to be the first statement in the test methods.

I’ll push a small follow-up commit with these changes shortly. Will ping you once CI is green again.

Thanks for this contribution! Adding collection support is a welcome improvement!

A couple of things I noticed:

1. Performance consideration: The change from gi.datasets.get_datasets() (server-side pagination) to gi.histories.show_history(contents=True) means we now fetch all items and paginate client-side. For large histories this could be slower, though I understand it's necessary to include collections. Might be worth a comment noting this tradeoff.
2. Unused parameter: The details parameter in get_history_contents() is no longer used in the new implementation
3. Minor test nit: A few test methods have their docstrings after the first line of code rather than immediately after the signature (e.g., test_get_collection_details_list_collection).

The new get_collection_details() tool looks well-structured, and the error handling that detects when someone passes a collection ID to get_dataset_details() is a nice touch for agent UX.

[dannon]

Thank you!

[ldalcolmo]

Thanks a lot for the review and approval!
Really appreciate the feedback.

If there are other MCP or Galaxy integration issues where I can help, feel free to point me to them — happy to contribute.