mcp-polaris-ai-datainsight

Polaris AI DataInsight is an API service that easily converts documents in various formats into structured data (such as JSON).

This tool supports the extraction of text, images, and other elements from various document formats (e.g. .docx, .pptx, .xlsx, .hwp, .hwpx).

For more details, please refer to the documentation.

Features

1. Extract content from document

Extract text, images, and other elements from various document formats.

• Supported document formats : docx, xlsx, pptx, hwpx, hwp
• Supported elements types : text, table, image, chart, shape, header, footer, caption
• Images in the document are stored on local storage, and the corresponding image paths are included in the JSON output.
• Tables are represented in JSON format, as illustrated in this example.

2. List files in allowed directories

List the files in the directory path set in the POLARIS_AI_DATA_INSIGHT_RESOURCES_DIR environment variable (in the case of Docker, the path mounted to /app/readable).

• Used to retrieve file paths to be passed as arguments to the "Extract content from document" Tool.

Installation and Setup

Prerequisites

To use this server, follow these steps:

1. Generate an API key.

   • Refer to this guide to generate an API key.

2. Install uv.

3. Create a writable directory for storing extraction-result resources(e.g. images in files), and set its path in the POLARIS_AI_DATA_INSIGHT_RESOURCES_DIR environment variable.

After that, choose one of the installation methods below and start the server.

Method 1: Manual Configuration

If you prefer a manual setup, add the following configuration to your IDE's MCP config file:

{
  "mcpServers": {
    "datainsight": {
      "command": "uvx",
      "args": [
        "--no-cache", 
        "mcp-polaris-ai-datainsight@latest",
        "/abs/path/to/input-docs-1",
        "/abs/path/to/input-docs-2",
        "/abs/path/to/input-docs-3"
      ],
      "env": {
        "POLARIS_AI_DATA_INSIGHT_API_KEY": "your-api-key",
        "POLARIS_AI_DATA_INSIGHT_RESOURCES_DIR": "/abs/path/to/output/assets"
      }
    }
  }
}

Method 2: Docker Container

1. Clone repository

   git clone --branch main https://github.com/PolarisOffice/PolarisAIDataInsight.git
   

   If you want to clone only mcp-polaris-ai-datainsight directory:

   # Git Version >= 2.25
   git clone --filter=blob:none --sparse --branch main https://github.com/PolarisOffice/PolarisAIDataInsight.git
   

   cd PolarisAIDataInsight
   

   git sparse-checkout set mcp-polaris-ai-datainsight
   

2. Build Docker image:

   cd mcp-polaris-ai-datainsight
   
   docker build --no-cache -t mcp-polaris-ai-datainsight .
   

3. Use this MCP Server config: Note: All readable files must be mounted to /app/readable by default.

   {
     "mcpServers": {
       "datainsight": {
         "command": "docker",
         "args": [
           "run",
           "-i",
           "--rm",
           "-e", "POLARIS_AI_DATA_INSIGHT_API_KEY=your-api-key",
           "--mount", "type=bind,src=/abs/path/to/input-docs-1,dst=/app/readable/dir_1,ro",
           "--mount", "type=bind,src=/abs/path/to/input-docs-2,dst=/app/readable/dir_2,ro",
           "mcp-polaris-ai-datainsight"
         ]
       }
     }
   }
   

Method 3: Clone git repository

[!] Important: uv and poetry must be pre-installed.

1. Clone git repository
2. Install python dependencies in virtual environment

   cd mcp-polaris-ai-datainsight
   

   uv venv .venv
   
   # Linux
   source .venv/bin/activate
   # Windows
   .venv\bin\activate
   
   
   poetry install --no-root
   

3. Set API Key and Resources Directory as environment values and Run server

   # Linux
   export POLARIS_AI_DATA_INSIGHT_API_KEY="your-api-key"
   export POLARIS_AI_DATA_INSIGHT_RESOURCES_DIR="/abs/path/to/output/assets"
   # Windows
   set POLARIS_AI_DATA_INSIGHT_API_KEY="your-api-key"
   set POLARIS_AI_DATA_INSIGHT_RESOURCES_DIR="/abs/path/to/output/assets"
   

   python -m mcp_polaris_ai_datainsight.server /abs/path/to/input-docs-1 /abs/path/to/input-docs-1 ...
   

Output

• Refer to this example for a sample output.
• Alternatively, you can test our API using the playground.