MSD

English | 简体中文

MSD (Micro Strategy Daemon) is a AI Friendly, High-Performance time-series database built on top of RocksDB, written in pure Rust. It is engineered to handle massive amounts of time-series data with speed and efficiency.

Purpose & Domain

The primary goal of MSD is to provide a easy-to-use, AI-aware time-series database for quantitative finance, help your do financial analysis and trading strategy development in 'one person studio' with AI assistance.

Key Features

• 🤖 AI Friendly:

  • MCP Integration: The built-in MCP service allows AI to understand the data, write correct data queries and analysis code.
  • Simple API: The API is designed to be simple, let AI write correct code easily.

• 🛠️ Easy Deployment:

  • Zero Dependencies: compiled as a single, self-contained executable with all dependencies (including RocksDB) statically linked.
  • Simplified Ops: No complex installation scripts, library conflicts, or container orchestrations required—just copy the binary and run.
  • Easy Upgrades: Updating the database is as simple as replacing the executable file.

• 🐍 Python First: Python is the first citizen language for MSD bindings. Both performance and ease of use are prioritized.

  • Zero-Copy NumPy Transformation: Leveraging Rust's memory safety and PyO3, the bindings allow for near instant transformation of msd tables into NumPy arrays (using from_vec / from_slice), enabling ultra-fast data analysis without serialization overhead.
  • Easy integration with pandas/polars: The bindings provide easy integration with popular data analysis libraries, such as pandas and polars, API return grouped, joined, aggregated DataFrames directly, no need to write complex SQL queries.

• 🚀 High Performance: We know that prepare data is a time-consuming process, and know there are huge data read in 'write-restart-verify' iterations when doing data analysis. In a typical workstation hardware(8C 16G SSD), MSD can

  • Insert total 100M OHLCVA rows within 10K stocks less than 20s at 5M/s speed
  • Query total 100M OHLCVA rows within 10K stocks less than 10s at 10M/s speed
  • See bench-of-msd for more details

Usage Guide

prerequisites

Download the pre-built binary from releases with your platform, unpack it to any PATH aware location.

Running the Server

Start the MSD server to accept connections and requests.

# Run with default settings (listens on 127.0.0.1:50510)
msd server

# Custom configuration
msd server --listen 0.0.0.0:8080 --workers 16 --db ./my_data

# Help
msd server --help

Interactive Shell

Connect to a running server and interact with it using the built-in shell.

# Run with default settings (connects to 127.0.0.1:50510)
msd shell

# Custom configuration
msd shell -s http://192.168.1.100:50510

# Execute a command without entering the shell
msd shell "select * from stock_kline_1d where obj='SH600000' limit 10"

# Help
msd shell --help

Python Bindings

1. Install the bindings

pip install pymsd

2. Install your favorite data analysis library and http client library, for example pandas and requests

pip install pandas requests

3. Use the bindings

import pymsd

MSD_URL = "http://127.0.0.1:50510"
client = pymsd.create_msd_pandas(MSD_URL)

data = client.load(
  objs=["SH600000", "SZ000001"],
  tables=["stock_kline_1d", "stock_dividend", "stock_shares"],
  start="2025-01-01",
  join={"stock_dividend": "zero", "*": "backward"},
)

sh600000 = data["SH600000"]
print(sh600000[(sh600000["dividend"] > 0) | (sh600000.index < 5) | (sh600000.index > 240)])

# results:
# - result is a pandas DataFrame as stock_kline_1d date order
# - stock_dividend is joined by zero, can be do compute meaningful, natural in next step
# - stock_shares is joined by backward, can be do compute meaningful, natural in next step

"""
            ts   open   high    low  close       volume  ...  transfers  dividend  rightShare  rightPrice         total      tradable
0   2025-01-02  11.73  11.77  11.39  11.43  181959699.0  ...        0.0      0.00         0.0         0.0  1.940557e+10  1.940557e+10
1   2025-01-03  11.44  11.54  11.36  11.38  115468044.0  ...        0.0      0.00         0.0         0.0  1.940557e+10  1.940557e+10
2   2025-01-06  11.38  11.48  11.22  11.44  108553630.0  ...        0.0      0.00         0.0         0.0  1.940557e+10  1.940557e+10
3   2025-01-07  11.42  11.53  11.37  11.51   74786288.0  ...        0.0      0.00         0.0         0.0  1.940557e+10  1.940557e+10
4   2025-01-08  11.50  11.63  11.40  11.50  106238601.0  ...        0.0      0.00         0.0         0.0  1.940557e+10  1.940557e+10
104 2025-06-12  11.56  11.70  11.52  11.68  129022000.0  ...        0.0      3.62         0.0         0.0  1.940557e+10  1.940557e+10
187 2025-10-15  11.34  11.42  11.26  11.40  127106100.0  ...        0.0      2.36         0.0         0.0  1.940560e+10  1.940560e+10
241 2025-12-30  11.53  11.56  11.45  11.48   58258400.0  ...        0.0      0.00         0.0         0.0  1.940560e+10  1.940560e+10
242 2025-12-31  11.48  11.49  11.40  11.41   59062000.0  ...        0.0      0.00         0.0         0.0  1.940560e+10  1.940560e+10
243 2026-01-05  11.42  11.51  11.41  11.50   87549100.0  ...        0.0      0.00         0.0         0.0  1.940560e+10  1.940560e+10
244 2026-01-06  11.50  11.68  11.48  11.67  130464800.0  ...        0.0      0.00         0.0         0.0  1.940560e+10  1.940560e+10

[11 rows x 14 columns]
"""