Lithops-Storage

This guide provides two methods for running and interacting with the application.

---

Method 1: Manual AWS Inbound Rule Setup

This method involves running the application directly and connecting to it from your local machine, which may require configuring a network rule in your AWS security group.

1. Execute App.py

   On your PyRun VM, start the application server. Note that for this method to work, the application must be configured to use the HTTP transport.

   # Ensure this line is in the code: mcp.run(transport="http", host="0.0.0.0", port=8080)
   python3 /work/app.py

2. Open Inspector from your personal PC

   From your personal PC, launch the Model-Context-Protocol Inspector.

   npx @modelcontextprotocol/inspector

3. Connect to the PyRun VM

   In the Inspector UI, configure the connection with the following details:

   • Transport Type: streamable http
   • URL: http://<IP_OF_VM>:8080/mcp

   Note: You may need to add an Inbound Rule to your VM's AWS Security Group to allow HTTP traffic on port 8080 from your IP address.

4. Example Usage

   a. Execute lithops_storage without parameters.

   b. Run a download_file operation.

   lithops_download_file(
         bucket="mcp-testing",
         key="customers.csv",
         file_name="/work/app.py",
   )

---

Method 2: Using Ollama

This method uses mcphost to integrate with a locally running Ollama model, communicating via standard I/O instead of HTTP.

Prerequisites: You have to install ollama and the model you wan't to use.

First, you must modify a line in the application code.

Substitute this line in the code: mcp.run(transport="http", host="0.0.0.0", port=8080)

For this one: mcp.run(transport="stdio")

Steps

1. Install MCPHOST

   go install github.com/mark3labs/mcphost@latest

2. Export Go to your PATH

   export PATH=$PATH:$(go env GOPATH)/bin

3. Execute MCPHOST

   Run mcphost and point it to the application and your desired Ollama model.

   mcphost --config /work/mcphost.json --model {your_model}

   More info about MCPHOST can be found here: https://github.com/mark3labs/mcphost

4. Prompt Example

   Can you please download the file with the key {KEY} from the bucket {BUCKET} where save as file_name /work/{filename}, keep in my mind maybe you have to execute first lithops_storage without config since we are in pyrun.