What dependencies are required to use parquet_mcp_server?
Requires Ollama server running for embedding generation, PostgreSQL with pgvector extension for PostgreSQL conversion, and Python with `uv` for installation.
Parquet MCP Server
@DeepSpringAI
Basic information
About Parquet MCP Server
No overview available yet
Parquet MCP Server config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"parquet_mcp_server": {
"command": "npx",
"args": [
"-y",
"@smithery/cli",
"install",
"@DeepSpringAI/parquet_mcp_server",
"--client",
"claude"
]
}
}
}Parquet MCP Server tools
13Path to input Parquet file
Path to save the output
Column containing text to embed
Name for the new embedding column
Number of texts to process in each batch (for better performance)
Path to the Parquet file to analyze
Path to the input Parquet file
Directory to save the DuckDB database (defaults to same directory as input file)
Path to the input Parquet file
Name of the PostgreSQL table to create or append to
Path to the markdown file to process
Path to save the output parquet file
Preserves document structure and links
Overview
What is parquet_mcp_server?
A Model Context Protocol (MCP) server that provides tools for manipulating and analyzing Parquet files, designed for use with Claude Desktop. It supports text embedding generation, Parquet file analysis, conversion to DuckDB or PostgreSQL (with pgvector), and markdown file processing.
How to use parquet_mcp_server?
Install via Smithery (npx -y @smithery/cli install @DeepSpringAI/parquet_mcp_server --client claude) or clone the repository and install with uv pip install -e .. Configure environment variables in a .env file (embedding URL, Ollama URL, embedding model, PostgreSQL credentials). Add the server to Claude Desktop configuration, then invoke tools using natural language prompts or the provided client functions.
Key features of parquet_mcp_server?
- Generate text embeddings from Parquet columns
- Extract Parquet file schema and metadata
- Convert Parquet to DuckDB databases
- Convert Parquet to PostgreSQL tables with pgvector
- Process markdown files into structured chunks
Use cases of parquet_mcp_server?
- Analyze large Parquet datasets for data science workflows
- Generate vector embeddings from text in Parquet files
- Convert Parquet to DuckDB for fast in-memory querying
- Convert Parquet to PostgreSQL for vector similarity search
- Chunk markdown documents for retrieval‑augmented generation (RAG)
Frequently asked questions
How do I configure the server?
Set environment variables for embedding URL, Ollama URL, embedding model, and PostgreSQL credentials in a `.env` file.
What if embedding generation fails?
Check that the Ollama server is accessible and the specified model is available, and that the text column exists in the input Parquet file.
What if PostgreSQL conversion fails?
Verify PostgreSQL connection settings, ensure the server is running, and that the pgvector extension is installed and you have table creation
More Data & Analytics MCP servers
FERS: structural finite-element analysis
FERS CloudStructural finite-element analysis for engineers and AI agents: beams, 3D frames, trusses and plates, with deflections, forces, reactions, buckling, frequencies and Eurocode 3 steel checks with a hand-calc trace, validat
Social Growth Engineers
Top In ItThe research behind short-form growth for your AI assistant: viral TikTok and Reels videos, the hooks that made them work, app growth playbooks, articles and reports.
MarketCode
RoboSapien LimitedUK property intelligence for AI agents — value any address, read any market, and source land and off-market stock through one MCP endpoint.
TikTok MCP Server
HasDataHosted TikTok MCP server for AI agents: read public profiles, account videos, comments and replies, or search videos and creators. Structured JSON for creator research and content analysis.
3
YouTube MCP Server
HasDataHosted YouTube MCP server for search, video metadata, channel data and timestamped transcripts. Connect Claude, Cursor or other MCP clients with HasData; no YouTube API key required.
6
Comments