mcp-fun-stuff

Fun stuff related to Kn-Functions and MCP

Implementations of MCP and integrating MCP into Knative Functions

• using python-sdk

Directories explained

• initial contains some initial MCP efforts using python-sdk lib with a function integrating this mcp
• mc-lama-mash contains mash of ollama and mcp creating mcp server as a function which exposes ollama inference model as one of it's tools. You can communicate with the MCP server (function) via MCP client (this can be further moved into having MCP client as another function and you could communicate w/ that function)
• rag-tool simply pulls a raw text via url and uses RAG to "enhance" the answers of the model

General How to use

• This might not apply exactly to all the directories
• install python venv
• pip install requirements.txt if available (usually just mcp,ollama)
• choose a function/project
• either:
  • run python main.py or similar in non-functions dir
  • run func deploy or func run if in a functions dir
• optionally
  • port-forward your function (pod) if in cluster. eg: kubectl port-forward <pod name in cluster> 8080:8080
• Run the client to connect
  • run client.py to communicate with the MCP server