Add Virtual MCP server support

[ptelang]

• Added virtual_mcp field to workload server model and database schema
• Implemented intelligent group filtering: prioritizes virtual MCP servers when present in a group, falls back to all servers when absent
• Updated tool discovery, search, and token counting to respect virtual MCP hierarchy
• Enhanced k8s client to detect VirtualMCPServer resources and configure transport/URLs correctly
• Added RBAC permissions for reading virtualmcpservers and mcpgroups CRDs
• Created vmcp-github-fetch.yaml example with complete MCPGroup, backend servers, and VirtualMCPServer configuration
• Created mcpserver_mcp-optimizer-vmcp.yaml for deploying MCP Optimizer with vMCP support
• Updated documentation with vMCP setup instructions and usage examples

🤖 Generated with Claude Code

[claude]

Code Review - PR #293: Add Virtual MCP server support

Summary

This PR adds Virtual MCP (vMCP) support, enabling intelligent filtering that prioritizes virtual MCP servers when present in a group. The implementation includes database schema changes, K8s client enhancements, and comprehensive documentation.

Issues Found

1. Performance: Inefficient double query in _get_group_server_filter() (base_tool_ops.py:751-827)

The method executes a query to identify groups with virtual MCPs, then builds filter logic. This is called every time find_similar_tools() or get_all_tools() is invoked, adding significant overhead.

Recommendation: Cache the group→virtual_mcp mapping or use a single query with conditional logic (e.g., using CASE statements).

2. Missing Tests

No tests verify the new virtual MCP filtering logic. The existing test_group_filtering.py doesn't cover:

• Virtual MCP server prioritization
• Fallback to all servers when no virtual MCPs exist
• Token counting with virtual MCP filtering
• Mixed groups (some with vMCP, some without)

Recommendation: Add tests to test_group_filtering.py covering these scenarios.

3. Migration doesn't handle existing data (migrations/versions/...py:71)

The migration adds virtual_mcp INTEGER NOT NULL DEFAULT 0 but doesn't verify existing rows handle the default correctly.

Recommendation: Consider using ALTER TABLE ... ADD COLUMN virtual_mcp INTEGER DEFAULT 0 NOT NULL explicitly or verify the migration works on existing databases.

4. Type inconsistency in virtual_mcp field

• Database: INTEGER (migrations:71)
• Model: bool (models.py:119)
• API: bool | None (api_models/core.py:61)

SQLite stores as 0/1, Python uses True/False. This works but could be clearer.

Recommendation: Add comment explaining the INTEGER→bool mapping, or use consistent integer types.

5. Verbose logging could impact performance (base_tool_ops.py:793-797)

Debug logging serializes sets to lists on every call, even when debug logging is disabled.

Recommendation: Use lazy evaluation or check log level first.

Positive Aspects

✓ Comprehensive documentation with working examples
✓ Consistent application across all tool operations
✓ Proper parameter handling prevents SQL injection
✓ Clear code structure and naming
✓ Backward compatible (defaults to virtual_mcp=False)

Minor Suggestions

• k8s_client.py:200-203 - Package naming vmcp:{name} is clever but consider if collisions are possible
• ingestion.py:735 - Consider extracting the new_virtual_mcp=workload.virtual_mcp or False pattern since it's repeated
• Documentation - Excellent! Very clear setup instructions

Verdict

Strong feature implementation with good documentation. Primary concern is the performance issue with repeated queries. Consider addressing before merge or in immediate follow-up.

Estimated reading time: 1.5 minutes

[therealnb]

I think the code is ok.

What is the use case for this? It looks redundant to me, if we have optimizer included in vmcp. Is it a temporary thing?