Consolidate scattered MCP implementations into unified orchestrator#91
Consolidate scattered MCP implementations into unified orchestrator#91
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
- Consolidate scattered MCP functionality into unified system - Create MCPOrchestrator for centralized task coordination - Create MCPServerRegistry for server management - Add comprehensive type definitions in types.py - Wire MCP orchestrator into service container - Create dependency injection helpers - Add comprehensive documentation Co-authored-by: groupthinking <154503486+groupthinking@users.noreply.github.com>
- Create implementation summary with metrics and testing results - Add step-by-step migration guide for external repositories - Document API usage examples and integration points - Include troubleshooting guide and next steps Co-authored-by: groupthinking <154503486+groupthinking@users.noreply.github.com>
Claude
AI
changed the title
![vercel[bot]](https://avatars.githubusercontent.com/in/8329?s=60&v=4){kind=small}
| self.orchestration_active = False | ||
| if self.orchestration_task: | ||
| self.orchestration_task.cancel() | ||
| try: |
There was a problem hiding this comment.
Spawned execution tasks are not tracked and cannot be properly managed, causing uncontrolled background task execution and inability to gracefully shutdown the orchestrator
|
|
||
| except Exception as e: | ||
| logger.error(f"Task {task_id} failed with error: {e}") | ||
|
|
There was a problem hiding this comment.
server_state variable could be undefined in except block, causing NameError when an exception occurs before assignment
| headers["Authorization"] = f"Bearer {config.auth_token}" | ||
|
|
||
| async with session.get( | ||
| url, headers=headers, timeout=config.timeout |
There was a problem hiding this comment.
| url, headers=headers, timeout=config.timeout | |
| url, headers=headers, timeout=aiohttp.ClientTimeout(total=config.timeout) |
aiohttp.ClientSession.get() receives raw integer timeout instead of ClientTimeout object, causing TypeError at runtime during health checks
| state.uptime_seconds += 10 # Monitoring interval | ||
|
|
||
| except Exception as e: | ||
| logger.error(f"Monitoring loop error: {e}") |
There was a problem hiding this comment.
Uptime calculation uses hardcoded 10-second increments instead of actual elapsed time, and doesn't reset when a server transitions from offline back to online
| created_at: datetime = Field( | ||
| default_factory=datetime.utcnow, description="Creation timestamp" | ||
| ) | ||
| started_at: Optional[datetime] = Field( |
There was a problem hiding this comment.
Pydantic Field default_factory should use lambda: datetime.utcnow() instead of direct datetime.utcnow reference for clarity and best-practice compatibility
2 participants
Problem
MCP functionality was fragmented across 8 external repositories and 3 internal implementations within EventRelay, totaling 2,273 lines of duplicated coordination logic. This violated the "Protocol-First Composability" principle requiring ONE canonical MCP implementation.
Solution
Created unified MCP orchestrator at
src/youtube_extension/services/mcp/consolidating internal implementations (51% code reduction). External repositories require manual migration by owners.Architecture
New unified layer:
orchestrator.py- Task coordination, dependency management, load balancingregistry.py- Server discovery, health monitoring, capability-based routingtypes.py- 15 capabilities, 6 task states, 5 server statesConsolidated from:
Usage
Server registration:
External Repositories (Out of Scope)
Cannot access external repos from this codebase. Migration guide provided in
docs/MCP_CONSOLIDATION_GUIDE.mdfor:Key Features
Files Changed
Created (8):
Modified (1):
Original prompt