What features does TiDB MCP server provide?

Q: What features does TiDB MCP server provide?

Automatic Embedding: Embeds text fields automatically and stores the vector embeddings in a vector field, enabling seamless retrieval.. Vector Search: Performs semantic similarity-based search by embedding queries and returning the most relevant records with optional filters.. Full-text Search: Supports keyword-based search by tokenizing queries and matching keywords, with options to return results as structured models.. Hybrid Search: Combines exact keyword matching with semantic search to deliver more relevant results in a single query.. Image Search: Enables multi-modal search using image embeddings for image_uri and image_vec fields, including text-to-image and image-to-image retrieval.. Advanced Filtering & Reranking: Provides flexible operators (e.g., $eq, $gt, $in, $and, $or) and optional reranker models to fine-tune results.. Join Structured and Unstructured Data: Supports joining tabular data with unstructured embeddings via Session-based queries to combine results from multiple sources.. Transaction Support: Offers transaction management for consistent data operations, including commit and rollback control.

TiDB MCP Server 2025: Features & Setup Guide

Overview

TiDB MCP Server is a MCP-enabled interface within the TiDB Python AI SDK that enables AI-first integration with TiDB data. It provides unified search modes—vector, full-text, and hybrid—along with automatic embedding of text fields into vector representations and storing them in vector columns. Developers can define schemas with embedding functions (e.g., OpenAI text embeddings) and perform end-to-end workflows from data ingestion to search and retrieval. The server supports image search by integrating multi-modal embeddings, enabling image_uri and image_vec fields, and performing text-to-image or image-to-image retrieval. It offers advanced filtering and reranking operators for refined results, and supports joining structured and unstructured data using Session-based queries. Transactions are supported to ensure data consistency across operations. It integrates with TiDB Cloud for hosted deployments and demonstrates how to connect via TiDBClient, create tables with embedded fields, bulk insert data, and perform vector/ hybrid/ full-text searches. The MCP extension is highlighted as built-in, ensuring a streamlined path to deploy MCP-powered AI capabilities on TiDB data.

Features

Automatic Embedding

Embeds text fields automatically and stores the vector embeddings in a vector field, enabling seamless retrieval.

Vector Search

Performs semantic similarity-based search by embedding queries and returning the most relevant records with optional filters.

Full-text Search

Supports keyword-based search by tokenizing queries and matching keywords, with options to return results as structured models.

Hybrid Search

Combines exact keyword matching with semantic search to deliver more relevant results in a single query.

Image Search

Enables multi-modal search using image embeddings for image_uri and image_vec fields, including text-to-image and image-to-image retrieval.

Advanced Filtering & Reranking

Provides flexible operators (e.g., $eq, $gt, $in, $and, $or) and optional reranker models to fine-tune results.

Join Structured and Unstructured Data

Supports joining tabular data with unstructured embeddings via Session-based queries to combine results from multiple sources.

Transaction Support

Offers transaction management for consistent data operations, including commit and rollback control.