MCP and NumPy

Two distinct topics for this TIL: processing data using NumPy and wiring up a basic MCP server.

NumPy and Events #

I was reading an article about NumPy for high performance sensor data analytics. It made me think back to my days dealing with SCADA (Supervisory Control and Data Acquisition) systems for municipal water utilities, and in some sense for in-vehicle sensor data analytics.

At first I was just curious how NumPy was doing vectorized operations, but then I thought ‘where does sensor data come from?’ and ‘What is a simple example of an end-to-end system without spinning up Kafka?’

The result is a codebase and Docker Compose setup demonstrating real-time sensor data collection, storage, and analysis using NumPy and modern data tools. The dashboard is showing data in Redis:

a dashboard driven by data in Redis

MCP #

Part of my AI Engineering course was about tool-calling. Since model context prototocol (MCP) is one key way that this can be implemented, I wanted to explore how to do this in the simples possible way.

Go and get the code at dvhthomas/mcp_demo.

Here’s what a session looks like:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
You: What's the weather in Tokyo?


============================================================
REACT LOOP - Agent Reasoning Steps (Real-time)
============================================================

[Step 1] User Question:
------------------------------------------------------------
What's the weather in Tokyo?

[Step 2] AIMessage (ai):
------------------------------------------------------------


  Tool Calls:
    - get_weather: {'city': 'Tokyo'}

[Step 3] ToolMessage (tool):
------------------------------------------------------------
{
  "city": "Tokyo",
  "country": "Japan",
  "temperature": 7.1,
  "temperature_unit": "\u00b0C",
  "wind_speed": 4.6,
  "wind_speed_unit": "km/h",
  "weather_code": 1,
  "coordinates": {
    "latitude": 35.6895,
    "longitude": 139.69171
  }
}

[Step 4] AIMessage (ai):
------------------------------------------------------------
The current weather in Tokyo is mostly cloudy with a temperature of 7.1°C (°C). There is a gentle breeze with a wind speed of 4.6 km/h. The weather code indicates that it's a partly sunny day. If you'd like to know more about the forecast or conditions, I can try to help!

============================================================

============================================================
Final Answer: The current weather in Tokyo is mostly cloudy with a temperature of 7.1°C (°C). There is a gentle breeze with a wind speed of 4.6 km/h. The weather code indicates that it's a partly sunny day. If you'd like to know more about the forecast or conditions, I can try to help!
============================================================

You:

I learned about server-sent events (SSE), which is a core part of MCP:

What it is: A web standard that enables a server to push data to a client in real-time over a single, long-lived HTTP connection.
How it works: After the client initiates a connection, the server can send updates whenever new data is available without the client having to repeatedly request it.
Purpose: To create live data streaming applications, such as live sports scores, stock ticker updates, and real-time notifications, making it easier to push information to a web browser.

Thankfully FastMCP handles this well:

1
2
3
4
5
6
7
@mcp_server.tool(
    name=weather_definition["name"], description=weather_definition["description"]
)
async def get_weather(city: str) -> str:
    """Get current weather information for a city."""
    result = await weather_adapter.execute({"city": city})
    return json.dumps(result, indent=2)

NumPy and Events #

MCP #

See Also