Unstructured API

🚀 Unstructured

This project provides various tools and configurations for interacting with different components, including environment variable settings, debugging tools, and client - side operations.

🚀 Quick Start

✨ Features

• Environment Variable Configuration: Allows users to control the behavior of the application through environment variables, such as suppressing debug output and disabling confirmation prompts.
• Debugging Tools: Offers a debugging UI through the MCP Inspector tool provided by Anthropic.
• Terminal Access: Enables terminal access to the minimal client using an external package.
• Tool Subset Usage: Provides guidelines for using a subset of tools in the client.

📦 Installation

Add Terminal Access to Minimal Client

We'll use @wonderwhy - er/desktop - commander to add terminal access to the minimal client. It's built on an MCP file system server. Note that the client (including the LLM) now has access to private files.

Execute the following command to install the package:

npx @wonderwhy-er/desktop-commander setup

Then start the client with additional parameters:

uv run python minimal_client/client.py "http://127.0.0.1:8080/sse" "@wonderwhy-er/desktop-commander"
# Or
make sse-client-terminal

💻 Usage Examples

Environment Variables

• LOG_LEVEL="ERROR": Set this environment variable to suppress debug output and only show clear messages to the user.
• CONFIRM_TOOL_USE='false': Disable the confirmation prompt before using the tool. Use with caution, especially during the development phase, as the LLM may execute expensive workflows or delete data.

# There is no actual Python code here, just for the template format
# If there were related Python code for setting env vars, it would be here

Debugging Tools

Anthropic provides the MCP Inspector tool to debug/test your MCP server. Run the following command to start a debugging UI. From here, you can add environment variables (pointing to your local environment) in the left pane. Include your personal API key as an environment variable there. Go to tools, and you can test the functions you've added to the MCP server.

mcp dev uns_mcp/server.py

If you need to log UnstructuredClient call parameters, set the environment variable DEBUG_API_REQUESTS=false. The logs will be saved as a file in the format unstructured - client - {date}.log, which you can view to debug the call parameters of the UnstructuredClient function.

📚 Documentation

Using a Subset of Tools

If your client only supports using a subset of tools, you need to note the following:

• The update_workflow tool needs to be loaded into the context along with the create_workflow tool, as it contains detailed descriptions of how to create and configure custom nodes.

Known Issues

• update_workflow - The workflow configuration (provided by the user or obtained by calling the get_workflow_info tool) needs to be included in the context, as this tool does not apply changes like patch but completely replaces the workflow configuration.

🔧 Technical Details

The project is based on the MCP file system server and uses various environment variables to control the behavior of the application. The MCP Inspector tool provided by Anthropic helps in debugging the MCP server.

📄 License

No license information is provided in the original document.

📝 Changelog

Any newly developed features/fixes/enhancements will be added to CHANGELOG.md. Before the stable release, use the 0.x.x - dev pre - release format.

⚙️ Troubleshooting

• If you encounter the Error: spawn <command> ENOENT error, it means <command> is not installed or not in the PATH:
  • Ensure it is installed and in your PATH.
  • Or provide the absolute path of the command in the command field. For example, replace python with /opt/miniconda3/bin/python in the configuration.

⚠️ Important Note

When setting CONFIRM_TOOL_USE='false', use it with caution, especially during the development phase, as the LLM may execute expensive workflows or delete data.

💡 Usage Tip

When using the MCP Inspector tool, remember to include your personal API key as an environment variable in the left pane of the debugging UI.