About Features of Ollama-MCP Bridge #11
Description
Could you explain a bit more about the following features?
Tool Router: How does it determine which MCP instance to route a request to? Does it rely on predefined mappings or some form of dynamic detection?
Automatic Tool Detection: How does the system detect the tools required from user prompts? Does it use a predefined schema or an NLP-based approach?
Multi-MCP Support & Dynamic Tool Routing: How does the bridge handle multiple MCP instances? Is there a load-balancing mechanism, or does it select based on specific tool requirements?
Also, while designing the system, did you consider the context window limitations of local LLMs? Large cloud-based LLMs can handle multiple MCP servers efficiently, but local models might struggle with limited memory and processing power. Was this a factor in your approach?