Skip to content

v2.1.5 - Policy Compliant Azure Login and MCP SDK Compatibility

Latest
Latest

Choose a tag to compare

View all tags
@4R9UN 4R9UN released this 03 Aug 05:31
v2.1.5
3042a19
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Release Date: August 2, 2026

Authentication

  • Checks az account get-access-token before attempting login.
  • Runs plain interactive az login only when no usable token is available.
  • Verifies token acquisition after login before allowing KQL execution.
  • Completes authentication at server startup so tool requests never block on interactive login.
  • Never invokes az login --use-device-code.

MCP Runtime

  • Migrated from the third party fastmcp distribution to the official mcp Python SDK.
  • Eliminates the fastmcp and fastmcp-slim mixed installation crash involving PrivateKeyJWTClientAuthenticator.
  • Eliminates the FastMCP 2 AuthlibDeprecationWarning path while preserving tool names and stdio or streamable HTTP transports.

Maintenance

  • Removed unreachable prompt templates, unused analyzers and feature registries, no-op query wrappers, and definition-only utility methods.
  • Replaced the optional PyPI version check's HTTP client fallback with urllib.request.
  • Removed redundant direct dependency declarations while retaining semantic search, schema validation, connection pooling, and compatibility bounds.
  • Removed the direct Pydantic constraint because the server does not import Pydantic; the official MCP SDK now owns that transitive requirement, preventing this package from downgrading Pydantic in shared environments.
  • Enforced read-only Kusto execution, allowing only .show and .help management commands.
  • Added loopback-only HTTP defaults, HTTPS endpoint validation, bounded responses and caches, and nonblocking authentication/query execution.
  • Closed SQLite connections deterministically and evicted expired cache entries during normal use.
  • Added model-aware embedding storage, configurable semantic models, offline ranking quality gates, and current MCP protocol discovery tests.
  • Made package metadata imports side-effect free and lazy, reducing measured cold import time from about 2.94 seconds to 0.11 seconds.
  • Added managed identity authentication and an internal-only streamable HTTP Container Apps deployment path.
  • Extracted bounded DataFrame/result handling from the main MCP tool handler into a dedicated formatter module.
  • Starts MCP transport immediately while Azure CLI authentication initializes in the background.
  • Makes Kusto tool calls wait for the startup authentication result without launching login during execution.
  • Reduces the normal query budget to 120 seconds while retaining the 600 second hard ceiling.
  • Detects timeout causes wrapped by the Azure Kusto SDK and prevents another full retry.
  • Requires lint, package validation, the platform install matrix, and tests before PyPI publishing.
  • Pins Ruff 0.16.1 and explicitly selects the established correctness rules so local and CI lint results remain consistent.

Upgrade

pip install --upgrade mcp-kql-server==2.1.5
Assets 2
Loading