DAP Client: Debug Adapter Protocol Client for Python

DAP Client is a generic client-side implementation of the Debug Adapter Protocol (DAP) for Python. It provides a clean, strongly-typed API for interacting with debug adapters but is decoupled from the transport layer (IO), allowing integration into any framework (synchronous, asynchronous, GUI, etc.).

For a comprehensive example of how to build a debugger UI with this client, see sandbox.py implementation.

Key Features

• Protocol-First: Implements the DAP state machine and message parsing without forcing a specific IO model.
• Strongly Typed: Uses Pydantic models for request arguments and event bodies.
• Easy Integration: Can be used with subprocess, socket, asyncio, or any other transport mechanism.

Installation

pip install dap-python

If you are developing/contributing to the code, use pip install -e . instead. I also support using poetry.

Quick Start

The DAPClient requires an IO handler to send and receive bytes from the debug adapter. The library acts as a translation layer:

• You feed it raw bytes from the adapter -> It yields high-level Event objects.
• You call methods like client.launch() -> It buffers raw bytes for you to send to the adapter.

Example: minimal synchronous runner

This example assumes you have an IO handler similar to sandbox/dap_io.py.

import time
from dap.client import DAPClient
from dap.events import InitializedEvent

# Import your IO handler (see sandbox/dap_io.py for a reference implementation)
from sandbox.dap_io import IO 

# 1. Start the debug adapter (e.g., debugpy)
adapter_cmd = "python -m debugpy.adapter"
io = IO(adapter_cmd)
io.start()

# 2. Initialize the DAP Client
client = DAPClient(
    clientID="my-debugger",
    clientName="My Custom Debugger",
    locale="en-US",
    pathFormat="path"
)

# 3. Send the Initialize Request
# client.send() returns the raw bytes that need to be sent to the adapter
io.write(client.send())

# 4. Event Loop
try:
    while True:
        # Read raw data from the adapter's stdout
        data = io.read()
        if data:
            # Feed data to the client to parse events
            events = client.recv(data)
            
            for event in events:
                print(f"Received: {event}")
                
                if isinstance(event, InitializedEvent):
                    print("Adapter Initialized!")
                    # Once initialized, we can configure and launch
                    client.launch(program="script.py", console="internalConsole")
                    client.configuration_done()
                    
                    # Flush requests generated by the above calls
                    io.write(client.send())

        time.sleep(0.01)
except KeyboardInterrupt:
    io.stop()

Architecture

The library is designed around the DAPClient class in dap.client.

Data Flow

[ Debug Adapter ] <== bytes ==> [ IO Handler ] <== bytes ==> [ DAPClient ] <== Objects ==> [ Your Application ]

1. Send: You call client.step_in(threadId=1). The client updates its internal state and queues the JSON-RPC message. You call client.send() to get the bytes and write them to your transport.
2. Receive: You read bytes from your transport and call client.recv(bytes). The client parses the buffer and yields Event, Response, or Request objects.

API Reference

DAPClient

The main entry point. Common methods include:

• launch(...) / attach(...)
• set_breakpoints(...) / configuration_done()
• continue_execution(...) / step_in(...) / next(...)
• stack_trace(...) / scopes(...) / variables(...)
• disconnect()

See dap/client.py for the full list of supported methods and arguments.