Summary

SQLite JSON tag filtering builds SQL text and random bind parameter names per call. This adds CPU overhead in SQLite deployments.

Evidence (current code)

• mcpgateway/utils/sqlalchemy_modifier.py: SQLite branch in json_contains_expr uses uuid for bind names and builds text(...) per call.

Impact

• Extra CPU for tag-filtered list endpoints on SQLite.

Proposed fix

• Use deterministic bind naming by position; cache SQL text templates by len(values) and match_any.

Acceptance criteria

• SQLite tag filtering avoids random bind names and reuses cached SQL templates.
• Behavior unchanged for any/all matching.