Ragandllm

This MCP server provides a modular Runtime Adaptation for Retrieval-Augmented Generation (RAG) and large language models to handle fish data, recipe generation, and fish record keeping. It exposes a small set of tools that you can invoke from an MCP client to upload images, generate recipes, and register catch data, enabling interactive AI-assisted culinary workflows.

How to use

You interact with the RAG and LLM MCP server from your MCP client by calling the available tools. The server supports three core tools: upload_file to submit a fish image and receive the fish name and ID, get_recipe to request a recipe based on the selected fish and user preferences, and register_choka to log your caught fish with size and count. Use each tool to build end-to-end interactions such as identifying a fish, generating a simple pan-fry recipe, and recording your catch for future reference.

Typical usage flow you can follow: first upload a fish image to identify it, then request a recipe tailored to your local cooking style, and finally register the catch to keep a running log. Your client should pass the required inputs for each step as described in the tool details: fish identifiers from the upload step, user input for recipe generation, and fish size/count for the catch log.

How to install

## Prerequisites
# Ensure you have the MCP runtime available in your environment so you can run stdio-based MCP servers with the uv tool.

## Quickstart configuration (stdio)
# Start the MCP server from your working directory using uv
uv --directory /path/to/RAGandLLM-MCP run RAGandLLM-MCP

Install and prepare the server directory as shown in the quickstart example. Create the directory where the MCP server files reside, then start the server with the standard runtime command. For Windows paths, mirror the example directory structure if you are following a Windows setup. If you switch to a Mac or Linux path, adapt the directory accordingly before launching the server.

Additional notes

Configuration in client tools may involve setting up a local MCP server entry that points to the stdio runtime with the directory containing the server. You can configure your client to invoke uv with the --directory option and the run subcommand followed by the MCP name you provided during startup.

The server exposes three tools for practical AI-assisted workflows: upload_file, get_recipe, and register_choka. Each tool has its own input requirements and produces outputs that you can feed into subsequent steps. Use upload_file to obtain the FishID and FishName, then pass those results to get_recipe together with your desired UserInput. Use register_choka to record your catch by supplying FishID, FishName, FishSize, and FishCount.

Troubleshooting and tips

If the server does not start, verify that the runtime tool uv is installed and accessible in your environment. Confirm that the working directory you point to contains the MCP server files and that you are using the correct MCP command name RAGandLLM-MCP when launching. If you encounter client-side errors, check that your request payload conforms to the expected fields for each tool and that you pass the required identifiers between steps.

Available tools

upload_file

Uploads a fish image to identify the fish. Returns FishID and FishName in the response.

get_recipe

Requests a recipe using the identified FishID, FishName, and UserInput describing cooking preferences or constraints.

register_choka

Registers the catch data, sending FishID, FishName, FishSize, and FishCount to record the catch.