Full Text Search MCP

A full-text search server using Tantivy that can be used as both an MCP (Model Context Protocol) server and a Python library.

Installation

pip install -e .

Usage

Combined FastAPI + MCP Server (one process)

fts-mcp-plus-api --config example_config.yaml
# MCP served at /mcp, REST API at the root paths (e.g., /search, /fetch)

As an MCP Server (CLI)

python -m full_text_search.mcp \
    --data-file data.jsonl \
    --id-column id \
    --text-column content \
    --searchable-columns title content tags \
    --description "My document search index" \
    --host 0.0.0.0 \
    --port 8000

As a Python Library

The FullTextSearchMCP class can be imported and used programmatically:

from full_text_search.mcp import FullTextSearchMCP

# Create and initialize the search server
server = FullTextSearchMCP()
server.initialize(
    data_file="data.jsonl",
    id_column="id", 
    text_column="content",
    searchable_columns=["title", "content", "tags"],
    description="My search index"
)

# Use search functionality directly
results = server.search("python programming", limit=10)
print(results)

# Retrieve specific documents
docs = server.read_documents(["doc1", "doc2"])
print(docs)

# Or run as MCP server programmatically
server.run_server(host="localhost", port=8080)

Creating an MCP Server Instance

You can also create a FastMCP server instance for integration with other applications:

from full_text_search.mcp import FullTextSearchMCP

server = FullTextSearchMCP()
server.initialize(
    data_file="data.jsonl",
    id_column="id",
    text_column="content", 
    searchable_columns=["title", "content"],
    description="My search index"
)

# Get the FastMCP server instance
mcp_server = server.create_mcp_server()
# Use mcp_server with your preferred transport

Data Format

The input data should be in JSONL format (one JSON object per line):

{"id": "1", "title": "Python Basics", "content": "Introduction to Python programming..."}
{"id": "2", "title": "Advanced Python", "content": "Advanced concepts in Python..."}

Features

• Full-text search using Tantivy search engine
• Modular design - can be used as a library or standalone server
• No global state - all functionality encapsulated in the FullTextSearchMCP class
• MCP protocol support for integration with AI assistants
• Flexible column mapping - specify which columns to search and how to identify documents
• Document retrieval by ID for full content access

API Methods

FullTextSearchMCP.initialize()

Initialize the search index with your data.

Parameters:

• data_file: Path to JSONL file containing documents
• id_column: Name of column containing unique document IDs
• text_column: Name of main text column for content
• searchable_columns: List of column names to make searchable
• description: Description of what this search index contains
• index_path: Optional path for index storage (defaults to {data_file}_index)

FullTextSearchMCP.search(query, limit=5)

Search documents and return previews.

Parameters:

• query: Search query string
• limit: Maximum number of results to return

Returns: Formatted string with search results and previews

FullTextSearchMCP.read_documents(document_ids)

Retrieve full content for specific document IDs.

Parameters:

• document_ids: List of document IDs to retrieve

Returns: Formatted string with full document content

FullTextSearchMCP.create_mcp_server()

Create a FastMCP server instance with search tools.

Returns: FastMCP server instance

FullTextSearchMCP.run_server(host="0.0.0.0", port=8000)

Run the MCP server.

Parameters:

• host: Host address for the server
• port: Port number for the server