Uploaded image for project: 'LLM AI Integration'
  1. LLM AI Integration
  2. LLMAI-153

MCP architecture refactor

    XMLWordPrintable

Details

    • Improvement
    • Resolution: Done
    • Major
    • None
    • 0.9
    • None
    • Unknown

    Description

      Restructure the MCP server internals ahead of the next tool wave (objects, attachments, history), following an architecture audit that found the shared machinery trapped inside individual tools and already being forked by the index module.

      The tool-support layer (declarative parameters, result/error builders, text helpers) and the access-door roles are promoted to mcp-api as @Unstable public API so contributing modules reuse them instead of copying. Internally, the duplicated plumbing is extracted into shared pieces: reach-conditional tool schemas, the write scaffold, get_document's rendered-HTML composer, and an authorized HQL row-query door the coming tools will query through. The per-wiki space filter is also cached (invalidated on config saves) instead of re-parsed per document checked.

      Behavior-preserving throughout;

      Attachments

        Activity

          People

            ppantiru Paul Pantiru
            ppantiru Paul Pantiru
            Votes:
            0 Vote for this issue
            Watchers:
            0 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: