Skip to content

v1.2.0

Choose a tag to compare

View all tags
@mbentham mbentham released this 13 Feb 13:30
· 44 commits to main since this release
v1.2.0
fd6b1d7

What's Changed

Strips JSON overhead from all tool responses to reduce token usage. LLMs already know the server/database they asked about, can count array elements, and can infer column types from values — so redundant metadata was removed across the board.

Breaking Changes

  • list_servers — Returns plain comma-separated text (production, staging) instead of a JSON array
  • list_databases — Returns plain comma-separated text (master, AdventureWorks) instead of a JSON envelope with server, databaseCount, and per-database metadata
  • read_data — Response now contains only truncated and rows. Removed server, rowCount, and columns (type metadata)
  • get_query_plan — Now writes XML to a .sqlplan file and returns a confirmation message instead of returning inline JSON with the plan XML. New required outputPath parameter
  • All stored procedure tools (sp_Blitz, sp_BlitzFirst, sp_BlitzCache, sp_BlitzIndex, sp_BlitzWho, sp_BlitzLock, sp_WhoIsActive, sp_PressureDetector, sp_QuickieStore, sp_HealthParser, sp_LogHunter, sp_HumanEventsBlockViewer, sp_IndexCleanup, sp_QueryReproBuilder) — Response is now a JSON array of {truncated, rows} objects (one per result set). Removed server, procedureName, columns, and rowCount

Internal Improvements

  • Extracted ToolHelper.SaveToFileAsync shared helper for file-write validation (extension check, directory creation, line count confirmation) — used by both get_plantuml_diagram and get_query_plan
  • File extension validation now uses Path.GetExtension instead of EndsWith for precision
  • Tightened tool descriptions for diagram and schema overview tools
  • Updated README with release install and clone-from-source setup options

Full Changelog: v1.1.0...v1.2.0

Assets 2
Loading