v1.1.0

v1.1.0 Release Notes

ds-workspace-mcp v1.1.0 is a minor release that expands the stable analytical MCP surface with safer SQL execution, richer dataset support, stronger modeling diagnostics, report persistence improvements, and reproducible evaluation metadata.

Highlights

• hardens DuckDB and SQLite execution with stricter read-only controls, relation allowlisting, and timeout handling
• adds generalized dataset support for Parquet, JSON, Excel, schema diffing, drift diagnostics, and safe group-by aggregation
• improves profiling with bounded distribution, categorical quality, duplicate-row, key-candidate, identifier, and free-text diagnostics
• upgrades modeling readiness with leakage evidence, executable validation guidance, statistically valid dummy baselines, and transparent forecast baselines
• adds reproducible evaluation manifests to experiment plans, modeling reports, dummy baselines, and forecast baselines
• makes report storage transactional with overwrite controls, content hashes, search, compare, copy, rename, delete, section extraction, and Docker persistence
• refactors server composition into a FastMCP application factory with domain-specific MCP registrations

Compatibility

This release adds MCP tools, CLI commands, and structured response fields in line with the documented minor-release compatibility policy. Existing public names remain supported.

Known Limitations

• analytical guidance remains heuristic and should be reviewed before operational decisions
• forecast baselines initially require regular inferred frequencies
• SQL timeout enforcement is best-effort interruption and still relies on configured operational bounds
• HTTP authentication remains a shared-token layer rather than scoped identity

Validation Summary

Validated with:

• poetry run ruff format --check src tests examples
• poetry run ruff check .
• poetry run mypy .
• poetry run pytest
• poetry build