feat: high-performance web terminal with asyncio-native I/O

Complete rewrite of the terminal pipeline for VS Code Server-level
responsiveness. Key improvements:

Backend:
- Replace blocking select.select(0.1) with asyncio.add_reader() for
  event-driven PTY reading (eliminates ~110ms polling latency)
- Add output batching (2ms window) to reduce WebSocket frame overhead
- Add flow control: client acks processed bytes, server pauses PTY reads
  at 64KB threshold, resumes at 32KB
- Add 5s ack timeout fallback to prevent stuck sessions

Frontend:
- Switch WebSocket to binary mode (binaryType = 'arraybuffer')
- Eliminate Blob -> arrayBuffer async conversion overhead
- Add flow control ack messages (every 4096 bytes or 100ms)
- Add xterm-addon-webgl with graceful DOM fallback
- Add performance tuning (scrollback=10000, fastScrollSensitivity)

SDD artifacts:
- openspec/explorations/terminal-responsiveness.md
- openspec/proposals/terminal-responsiveness.md
- openspec/specs/terminal-responsiveness.md
- openspec/designs/terminal-responsiveness.md
- openspec/tasks/terminal-responsiveness.md

Quality gates: pytest (19 passed, 1 skipped), tsc --noEmit clean
This commit is contained in:
Alex Blank
2026-06-02 14:40:32 +02:00
parent 906aab3b73
commit c754984df8
11 changed files with 984 additions and 114 deletions
+6 -25
View File
@@ -222,8 +222,7 @@ async def _handle_terminal_websocket(
# Use mutable session reference so loops can survive reset # Use mutable session reference so loops can survive reset
session_ref = SessionRef(session, slot_session_id) session_ref = SessionRef(session, slot_session_id)
# Start I/O loops and heartbeat # Start write loop and heartbeat (read is now event-driven in TerminalSession)
read_task = asyncio.create_task(_read_loop(session_ref, websocket))
write_task = asyncio.create_task( write_task = asyncio.create_task(
_write_loop(session_ref, websocket, instance_id) _write_loop(session_ref, websocket, instance_id)
) )
@@ -232,7 +231,7 @@ async def _handle_terminal_websocket(
# Wait for either task to complete (indicating disconnect or error) # Wait for either task to complete (indicating disconnect or error)
done, pending = await asyncio.wait( done, pending = await asyncio.wait(
[read_task, write_task, heartbeat_task], [write_task, heartbeat_task],
return_when=asyncio.FIRST_COMPLETED, return_when=asyncio.FIRST_COMPLETED,
) )
@@ -267,28 +266,6 @@ async def _handle_terminal_websocket(
) )
async def _read_loop(session_ref: SessionRef, websocket) -> None:
"""Read output from the container and send to WebSocket."""
try:
while True:
session = session_ref.session
if not session.is_alive() or session._closed:
await asyncio.sleep(0.1)
continue
data = await session.read_output()
if data:
try:
await websocket.send_bytes(data)
except WebSocketDisconnect:
break
except Exception:
break
else:
await asyncio.sleep(0.01)
except Exception:
pass
async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> None: async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> None:
"""Read input from WebSocket and send to container.""" """Read input from WebSocket and send to container."""
try: try:
@@ -319,6 +296,10 @@ async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> N
rows, rows,
) )
await session.resize(cols, rows) await session.resize(cols, rows)
elif msg_type == "ack":
char_count = ctrl.get("chars", 0)
if char_count > 0:
session.acknowledge_data(char_count)
elif msg_type == "reset": elif msg_type == "reset":
# Reset terminal session (scoped to current slot) # Reset terminal session (scoped to current slot)
logger.debug( logger.debug(
+219 -65
View File
@@ -1,10 +1,13 @@
"""Terminal session management for tool instances.""" """High-performance terminal session with asyncio-native I/O.
Replaces blocking select.select() with event-driven asyncio.add_reader()
for sub-frame latency. Includes output batching and flow control.
"""
import asyncio import asyncio
import logging import logging
import os import os
import pty import pty
import select
import signal import signal
import struct import struct
import fcntl import fcntl
@@ -17,18 +20,31 @@ logger = logging.getLogger(__name__)
class TerminalSession: class TerminalSession:
"""Manages a single terminal session connected to a docker container. """Manages a single terminal session with event-driven PTY I/O.
Supports persistent sessions that survive WebSocket disconnections. Uses asyncio.add_reader() instead of polling for near-zero read latency.
Multiple WebSocket connections can attach/detach from the same session. Output is batched (2ms window) and sent as binary WebSocket frames.
Flow control prevents memory bloat on fast output.
""" """
# Circular buffer size (10KB) # Circular buffer for replay (10KB)
BUFFER_SIZE = 10 * 1024 BUFFER_SIZE = 10 * 1024
# Idle timeout in seconds (30 minutes) # Idle timeout in seconds (30 minutes)
IDLE_TIMEOUT = 30 * 60 IDLE_TIMEOUT = 30 * 60
# Output batching window in seconds
BATCH_WINDOW_S = 0.002 # 2ms
# Flow control: pause PTY reads when unacknowledged bytes exceed this
FLOW_CONTROL_PAUSE = 64 * 1024
# Flow control: resume PTY reads when unacknowledged bytes drop below this
FLOW_CONTROL_RESUME = 32 * 1024
# Max WebSocket frame size
MAX_FRAME_SIZE = 64 * 1024
# Session number counter per instance_id for auto-naming # Session number counter per instance_id for auto-naming
_instance_counters: dict[str, int] = {} _instance_counters: dict[str, int] = {}
@@ -47,7 +63,6 @@ class TerminalSession:
self.process: asyncio.subprocess.Process | None = None self.process: asyncio.subprocess.Process | None = None
self._closed = False self._closed = False
self._master_fd: int | None = None self._master_fd: int | None = None
self._slave_fd: int | None = None
# Circular buffer for output replay # Circular buffer for output replay
self._output_buffer: deque[bytes] = deque(maxlen=self.BUFFER_SIZE) self._output_buffer: deque[bytes] = deque(maxlen=self.BUFFER_SIZE)
@@ -67,6 +82,20 @@ class TerminalSession:
self.name = name or self._generate_name(str(instance_id)) self.name = name or self._generate_name(str(instance_id))
self.status: str = "active" self.status: str = "active"
# Output batching
self._batch_buffer = bytearray()
self._batch_timer: asyncio.TimerHandle | None = None
self._batch_lock = asyncio.Lock()
# Flow control
self._unacknowledged_bytes = 0
self._paused = False
self._read_handler_set = False
self._flow_control_lock = asyncio.Lock()
# Ack timeout fallback
self._ack_timeout_handle: asyncio.TimerHandle | None = None
@classmethod @classmethod
def _generate_name(cls, instance_id: str) -> str: def _generate_name(cls, instance_id: str) -> str:
"""Generate an auto-incremented session name for the instance.""" """Generate an auto-incremented session name for the instance."""
@@ -77,87 +106,189 @@ class TerminalSession:
async def start(self, startup_command: str | None = None) -> None: async def start(self, startup_command: str | None = None) -> None:
"""Start the docker exec process with a shell using a PTY.""" """Start the docker exec process with a shell using a PTY."""
# Create a pseudo-terminal on the host # Create a pseudo-terminal on the host
self._master_fd, self._slave_fd = pty.openpty() self._master_fd, slave_fd = pty.openpty()
# Set the terminal size initially # Set the terminal size initially
self._set_terminal_size(self._cols, self._rows) self._set_terminal_size(self._cols, self._rows)
logger.debug( logger.debug(
f"Starting terminal session {self.session_id} for container {self.container_id} with initial size {self._cols}x{self._rows}" "Starting terminal session %s for container %s with initial size %sx%s",
self.session_id,
self.container_id,
self._cols,
self._rows,
) )
# Build the shell command # Build the shell command
if startup_command: cmd = startup_command or self.startup_command
shell_cmd = f'bash -c "{startup_command}" || true; exec bash -il' if cmd:
shell_cmd = f'bash -c "{cmd}" || true; exec bash -il'
logger.debug( logger.debug(
f"Using startup command for session {self.session_id}: {startup_command}" "Using startup command for session %s: %s",
self.session_id,
cmd,
) )
else: else:
shell_cmd = "bash -il" shell_cmd = "bash -il"
# Start docker exec with the slave fd as stdin/stdout/stderr # Start docker exec with the slave fd as stdin/stdout/stderr
# Using -it because the slave fd IS a TTY
self.process = await asyncio.create_subprocess_exec( self.process = await asyncio.create_subprocess_exec(
"docker", "docker",
"exec", "exec",
"-it", "-it",
"-e", "-e",
"TERM=xterm", "TERM=xterm-256color",
self.container_id, self.container_id,
"bash", "bash",
"-c", "-c",
shell_cmd, shell_cmd,
stdin=self._slave_fd, stdin=slave_fd,
stdout=self._slave_fd, stdout=slave_fd,
stderr=self._slave_fd, stderr=slave_fd,
) )
# Close slave fd in parent process # Close slave fd in parent process
os.close(self._slave_fd) os.close(slave_fd)
self._slave_fd = None
self.last_activity = time.time() self.last_activity = time.time()
def _set_terminal_size(self, cols: int, rows: int) -> None: # Start event-driven reading
"""Set the terminal size using TIOCSWINSZ.""" self._start_reading()
if self._master_fd is None:
logger.warning("Cannot resize: master_fd is None (session not started)") def _start_reading(self) -> None:
return """Register PTY master fd with asyncio event loop for event-driven reads."""
# TIOCSWINSZ = 0x5414 on Linux if self._read_handler_set or self._master_fd is None or self._closed:
TIOCSWINSZ = 0x5414 return
size = struct.pack("HHHH", rows, cols, 0, 0) try:
try: loop = asyncio.get_event_loop()
fcntl.ioctl(self._master_fd, TIOCSWINSZ, size) loop.add_reader(self._master_fd, self._on_fd_readable)
logger.debug(f"Resized PTY to {cols}x{rows} (fd={self._master_fd})") self._read_handler_set = True
except (OSError, IOError) as e: logger.debug("Started event-driven reading for session %s", self.session_id)
logger.error(f"Failed to resize PTY: {e}") except Exception as exc:
logger.error("Failed to start reading for session %s: %s", self.session_id, exc)
def _stop_reading(self) -> None:
"""Unregister PTY master fd from asyncio event loop."""
if not self._read_handler_set or self._master_fd is None:
return
try:
loop = asyncio.get_event_loop()
loop.remove_reader(self._master_fd)
self._read_handler_set = False
except Exception:
pass
def _on_fd_readable(self) -> None:
"""Callback when PTY master fd has data available (called by event loop)."""
if self._master_fd is None or self._closed:
return
async def read_output(self) -> bytes:
"""Read output from the PTY master and store in buffer."""
if self._master_fd is None or self._closed:
return b""
try: try:
# Use select to check if data is available
readable, _, _ = select.select([self._master_fd], [], [], 0.1)
if readable:
data = os.read(self._master_fd, 4096) data = os.read(self._master_fd, 4096)
if data: except (OSError, IOError):
return
if not data:
return
self._add_to_buffer(data) self._add_to_buffer(data)
self.last_activity = time.time() self.last_activity = time.time()
return data
return b"" # Queue for batching + flow control
except (OSError, IOError, ValueError): self._queue_output(data)
return b""
def _add_to_buffer(self, data: bytes) -> None: def _add_to_buffer(self, data: bytes) -> None:
"""Add data to circular buffer, maintaining size limit.""" """Add data to circular buffer, maintaining size limit."""
self._output_buffer.append(data) self._output_buffer.append(data)
self._buffer_size += len(data) self._buffer_size += len(data)
# Trim if exceeds max size
while self._buffer_size > self.BUFFER_SIZE and self._output_buffer: while self._buffer_size > self.BUFFER_SIZE and self._output_buffer:
removed = self._output_buffer.popleft() removed = self._output_buffer.popleft()
self._buffer_size -= len(removed) self._buffer_size -= len(removed)
def _queue_output(self, data: bytes) -> None:
"""Add output to batch buffer and schedule flush."""
self._batch_buffer.extend(data)
self._unacknowledged_bytes += len(data)
# Check flow control
if self._unacknowledged_bytes > self.FLOW_CONTROL_PAUSE and not self._paused:
self._pause_output()
# Schedule batch flush if not already scheduled
if self._batch_timer is None:
loop = asyncio.get_event_loop()
self._batch_timer = loop.call_later(
self.BATCH_WINDOW_S,
self._flush_batch_sync,
)
def _flush_batch_sync(self) -> None:
"""Synchronous entry point for batch flush (called from event loop)."""
self._batch_timer = None
if not self._batch_buffer or not self._websockets:
self._batch_buffer.clear()
return
payload = bytes(self._batch_buffer)
self._batch_buffer.clear()
# Send to all websockets (asyncio.create_task for async send)
dead_sockets = set()
for ws in list(self._websockets):
try:
asyncio.create_task(self._send_bytes(ws, payload))
except Exception:
dead_sockets.add(ws)
if dead_sockets:
self._websockets -= dead_sockets
async def _send_bytes(self, ws: Any, payload: bytes) -> None:
"""Send bytes to a single websocket, catching errors."""
try:
await ws.send_bytes(payload)
except Exception:
self._websockets.discard(ws)
def acknowledge_data(self, char_count: int) -> None:
"""Client acknowledges processing char_count bytes.
Called from the WebSocket handler when the client sends an 'ack' message.
"""
self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
if self._paused and self._unacknowledged_bytes < self.FLOW_CONTROL_RESUME:
self._resume_output()
# Reset ack timeout
if self._ack_timeout_handle:
self._ack_timeout_handle.cancel()
loop = asyncio.get_event_loop()
self._ack_timeout_handle = loop.call_later(
5.0, self._ack_timeout_fallback
)
def _ack_timeout_fallback(self) -> None:
"""If no ack received for 5s, assume client is dead and resume."""
logger.warning(
"Flow control ack timeout for session %s, resuming output",
self.session_id,
)
self._unacknowledged_bytes = 0
if self._paused:
self._resume_output()
def _pause_output(self) -> None:
"""Pause reading from PTY due to flow control."""
self._paused = True
self._stop_reading()
logger.debug("Paused output for session %s (%d unacked)", self.session_id, self._unacknowledged_bytes)
def _resume_output(self) -> None:
"""Resume reading from PTY."""
self._paused = False
self._start_reading()
logger.debug("Resumed output for session %s", self.session_id)
def get_buffer(self) -> bytes: def get_buffer(self) -> bytes:
"""Get buffered output for replay.""" """Get buffered output for replay."""
return b"".join(self._output_buffer) return b"".join(self._output_buffer)
@@ -172,38 +303,41 @@ class TerminalSession:
except (OSError, IOError): except (OSError, IOError):
pass pass
def _set_terminal_size(self, cols: int, rows: int) -> None:
"""Set the terminal size using TIOCSWINSZ."""
if self._master_fd is None:
logger.warning("Cannot resize: master_fd is None (session not started)")
return
TIOCSWINSZ = 0x5414
size = struct.pack("HHHH", rows, cols, 0, 0)
try:
fcntl.ioctl(self._master_fd, TIOCSWINSZ, size)
logger.debug("Resized PTY to %sx%s (fd=%s)", cols, rows, self._master_fd)
except (OSError, IOError) as e:
logger.error("Failed to resize PTY: %s", e)
async def resize(self, cols: int, rows: int) -> None: async def resize(self, cols: int, rows: int) -> None:
"""Resize the terminal.""" """Resize the terminal."""
if self._closed: if self._closed:
logger.warning("Cannot resize: session is closed") logger.warning("Cannot resize: session is closed")
return return
# Only resize if dimensions actually changed
if cols == self._cols and rows == self._rows: if cols == self._cols and rows == self._rows:
return return
self._cols = cols self._cols = cols
self._rows = rows self._rows = rows
logger.debug(f"resize() called for session {self.session_id}: {cols}x{rows}") logger.debug("resize() called for session %s: %sx%s", self.session_id, cols, rows)
self._set_terminal_size(cols, rows) self._set_terminal_size(cols, rows)
# Docker exec -it creates its own PTY inside the container, # Send SIGWINCH to docker exec process
# so host PTY resize doesn't propagate to the container shell.
# Send SIGWINCH to the docker exec process on the host.
# Docker exec forwards signals to the container process, which should
# cause the container's shell to re-read its terminal size.
if self.process and self.process.pid: if self.process and self.process.pid:
try: try:
os.kill(self.process.pid, signal.SIGWINCH) os.kill(self.process.pid, signal.SIGWINCH)
logger.debug(
f"Sent SIGWINCH to docker exec process {self.process.pid} for session {self.session_id}"
)
except ProcessLookupError: except ProcessLookupError:
logger.warning( logger.warning("docker exec process %s not found", self.process.pid)
f"docker exec process {self.process.pid} not found for session {self.session_id}"
)
except Exception as e: except Exception as e:
logger.warning(f"Failed to send SIGWINCH: {e}") logger.warning("Failed to send SIGWINCH: %s", e)
async def reset(self) -> None: async def reset(self) -> None:
"""Reset the session by killing the process and clearing state.""" """Reset the session by killing the process and clearing state."""
@@ -213,9 +347,13 @@ class TerminalSession:
self._output_buffer.clear() self._output_buffer.clear()
self._buffer_size = 0 self._buffer_size = 0
self._websockets.clear() self._websockets.clear()
self._batch_buffer.clear()
self._batch_timer = None
self._unacknowledged_bytes = 0
self._paused = False
self._read_handler_set = False
self.process = None self.process = None
self._master_fd = None self._master_fd = None
self._slave_fd = None
self.status = "active" self.status = "active"
async def close(self) -> None: async def close(self) -> None:
@@ -225,11 +363,21 @@ class TerminalSession:
self._closed = True self._closed = True
self.status = "closed" self.status = "closed"
self._stop_reading()
if self._batch_timer:
self._batch_timer.cancel()
self._batch_timer = None
if self._ack_timeout_handle:
self._ack_timeout_handle.cancel()
self._ack_timeout_handle = None
if self._master_fd is not None: if self._master_fd is not None:
try: try:
os.close(self._master_fd) os.close(self._master_fd)
except OSError: except OSError:
pass # noqa: S110 pass
self._master_fd = None self._master_fd = None
if self.process is not None: if self.process is not None:
@@ -265,14 +413,20 @@ class TerminalSession:
return len(self._websockets) > 0 return len(self._websockets) > 0
async def send_to_all(self, data: bytes) -> None: async def send_to_all(self, data: bytes) -> None:
"""Send data to all attached WebSockets.""" """Send data to all attached WebSockets (used for control messages)."""
dead_sockets = set() dead_sockets = set()
for ws in self._websockets: for ws in self._websockets:
try: try:
await ws.send_bytes(data) await ws.send_bytes(data)
except Exception: except Exception:
dead_sockets.add(ws) dead_sockets.add(ws)
# Clean up dead sockets
for ws in dead_sockets: for ws in dead_sockets:
self._websockets.discard(ws) self._websockets.discard(ws)
async def read_output(self) -> bytes:
"""Legacy method: read output synchronously.
With event-driven I/O, output is automatically sent to websockets.
This method returns any buffered data for callers that poll.
"""
return b""
+1 -3
View File
@@ -90,9 +90,7 @@ class WorkspaceManager:
# Remove stale directory from previous failed/aborted clone # Remove stale directory from previous failed/aborted clone
if os.path.exists(path): if os.path.exists(path):
logger.warning( logger.warning("Removing stale workspace directory: %s", path)
"Removing stale workspace directory: %s", path
)
shutil.rmtree(path, ignore_errors=True) shutil.rmtree(path, ignore_errors=True)
# Load SSH key if repo has one # Load SSH key if repo has one
+12 -12
View File
@@ -16,11 +16,11 @@
"react-dom": "^18.2.0", "react-dom": "^18.2.0",
"react-router-dom": "^6.20.0", "react-router-dom": "^6.20.0",
"react-simple-code-editor": "^0.14.1", "react-simple-code-editor": "^0.14.1",
"sonner": "^1.7.4",
"tailwindcss": "^3.3.0", "tailwindcss": "^3.3.0",
"xterm": "^5.3.0", "xterm": "^5.3.0",
"xterm-addon-fit": "^0.8.0", "xterm-addon-fit": "^0.8.0",
"xterm-addon-web-links": "^0.9.0" "xterm-addon-web-links": "^0.9.0",
"xterm-addon-webgl": "^0.16.0"
}, },
"devDependencies": { "devDependencies": {
"@testing-library/jest-dom": "^6.9.1", "@testing-library/jest-dom": "^6.9.1",
@@ -5469,16 +5469,6 @@
"node": ">=8" "node": ">=8"
} }
}, },
"node_modules/sonner": {
"version": "1.7.4",
"resolved": "https://registry.npmjs.org/sonner/-/sonner-1.7.4.tgz",
"integrity": "sha512-DIS8z4PfJRbIyfVFDVnK9rO3eYDtse4Omcm6bt0oEr5/jtLgysmjuBl1frJ9E/EQZrFmKx2A8m/s5s9CRXIzhw==",
"license": "MIT",
"peerDependencies": {
"react": "^18.0.0 || ^19.0.0 || ^19.0.0-rc",
"react-dom": "^18.0.0 || ^19.0.0 || ^19.0.0-rc"
}
},
"node_modules/source-map-js": { "node_modules/source-map-js": {
"version": "1.2.1", "version": "1.2.1",
"resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.1.tgz", "resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.1.tgz",
@@ -6324,6 +6314,16 @@
"xterm": "^5.0.0" "xterm": "^5.0.0"
} }
}, },
"node_modules/xterm-addon-webgl": {
"version": "0.16.0",
"resolved": "https://registry.npmjs.org/xterm-addon-webgl/-/xterm-addon-webgl-0.16.0.tgz",
"integrity": "sha512-E8cq1AiqNOv0M/FghPT+zPAEnvIQRDbAbkb04rRYSxUym69elPWVJ4sv22FCLBqM/3LcrmBLl/pELnBebVFKgA==",
"deprecated": "This package is now deprecated. Move to @xterm/addon-webgl instead.",
"license": "MIT",
"peerDependencies": {
"xterm": "^5.0.0"
}
},
"node_modules/yallist": { "node_modules/yallist": {
"version": "3.1.1", "version": "3.1.1",
"resolved": "https://registry.npmjs.org/yallist/-/yallist-3.1.1.tgz", "resolved": "https://registry.npmjs.org/yallist/-/yallist-3.1.1.tgz",
+2 -1
View File
@@ -22,7 +22,8 @@
"tailwindcss": "^3.3.0", "tailwindcss": "^3.3.0",
"xterm": "^5.3.0", "xterm": "^5.3.0",
"xterm-addon-fit": "^0.8.0", "xterm-addon-fit": "^0.8.0",
"xterm-addon-web-links": "^0.9.0" "xterm-addon-web-links": "^0.9.0",
"xterm-addon-webgl": "^0.16.0"
}, },
"devDependencies": { "devDependencies": {
"@testing-library/jest-dom": "^6.9.1", "@testing-library/jest-dom": "^6.9.1",
+47 -5
View File
@@ -8,6 +8,7 @@ import React, {
import { Terminal } from "xterm"; import { Terminal } from "xterm";
import { FitAddon } from "xterm-addon-fit"; import { FitAddon } from "xterm-addon-fit";
import { WebLinksAddon } from "xterm-addon-web-links"; import { WebLinksAddon } from "xterm-addon-web-links";
import { WebglAddon } from "xterm-addon-webgl";
import "xterm/css/xterm.css"; import "xterm/css/xterm.css";
import { import {
@@ -107,6 +108,7 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
// WebSocket connection established // WebSocket connection established
const ws = new WebSocket(wsUrl); const ws = new WebSocket(wsUrl);
ws.binaryType = "arraybuffer";
wsRef.current = ws; wsRef.current = ws;
ws.onopen = () => { ws.onopen = () => {
@@ -137,14 +139,35 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
}, 30000); }, 30000);
}; };
// Flow control: accumulate processed bytes and send ack
let ackAccumulator = 0;
const ACK_THRESHOLD = 4096;
let ackTimeout: ReturnType<typeof setTimeout> | null = null;
const flushAck = () => {
if (ackAccumulator > 0 && ws.readyState === WebSocket.OPEN) {
ws.send(JSON.stringify({ type: "ack", chars: ackAccumulator }));
ackAccumulator = 0;
}
};
ws.onmessage = (event) => { ws.onmessage = (event) => {
if (!termRef.current) return; if (!termRef.current) return;
if (event.data instanceof Blob) { if (event.data instanceof ArrayBuffer) {
event.data.arrayBuffer().then((buffer) => { const data = new Uint8Array(event.data);
const data = new Uint8Array(buffer); termRef.current.write(data);
termRef.current?.write(data);
}); // Flow control: accumulate processed bytes
ackAccumulator += data.length;
if (ackAccumulator >= ACK_THRESHOLD) {
flushAck();
} else if (!ackTimeout) {
ackTimeout = setTimeout(() => {
ackTimeout = null;
flushAck();
}, 100);
}
} else if (typeof event.data === "string") { } else if (typeof event.data === "string") {
try { try {
const msg = JSON.parse(event.data); const msg = JSON.parse(event.data);
@@ -251,6 +274,11 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
lineHeight: 1.2, lineHeight: 1.2,
letterSpacing: 0, letterSpacing: 0,
allowTransparency: false, allowTransparency: false,
scrollback: 10000,
ignoreBracketedPasteMode: false,
fastScrollSensitivity: 5,
scrollSensitivity: 1,
smoothScrollDuration: 0,
theme: { theme: {
background: "#1e1e1e", background: "#1e1e1e",
foreground: "#d4d4d4", foreground: "#d4d4d4",
@@ -282,6 +310,20 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
term.loadAddon(fitAddon); term.loadAddon(fitAddon);
term.loadAddon(new WebLinksAddon()); term.loadAddon(new WebLinksAddon());
// Load WebGL renderer for GPU acceleration, fall back to DOM
try {
const webglAddon = new WebglAddon();
term.loadAddon(webglAddon);
webglAddon.onContextLoss(() => {
console.warn("WebGL context lost, falling back to DOM renderer");
webglAddon.dispose();
// Trigger a refit since cell dimensions may differ
requestAnimationFrame(() => fitTerminal());
});
} catch (e) {
console.warn("WebGL renderer failed to load, using DOM renderer", e);
}
const container = terminalRef.current; const container = terminalRef.current;
// Define fitTerminal before connectWebSocket so it's available in onmessage // Define fitTerminal before connectWebSocket so it's available in onmessage
+227
View File
@@ -0,0 +1,227 @@
# Design: High-Performance Web Terminal
## Component Diagram
```
┌─────────────────────────────────────────────────────────────┐
│ Browser │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────────┐ │
│ │ Terminal │ │ WebSocket │ │ xterm.js │ │
│ │ Component │──│ Client │──│ + WebGL addon │ │
│ │ │ │ (binary) │ │ + DOM fallback │ │
│ └─────────────┘ └─────────────┘ └─────────────────────┘ │
│ │ │ │
│ │ Flow control ack │ │
│ └──────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
│ WebSocket
│ (binary frames)
┌─────────────────────────────────────────────────────────────┐
│ API Container │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────────┐ │
│ │ Terminal │ │ WebSocket │ │ TerminalSession │ │
│ │ Manager │──│ Endpoint │──│ (new) │ │
│ │ (lifecycle) │ │ (router) │ │ - asyncio fd reader │ │
│ └─────────────┘ └─────────────┘ │ - output batcher │ │
│ │ - flow control │ │
│ └─────────────────────┘ │
│ │ │
│ ┌────────┴────────┐ │
│ │ docker exec -it │ │
│ │ (subprocess) │ │
│ └────────┬────────┘ │
│ │ │
│ ┌────────┴────────┐ │
│ │ Tool Container │ │
│ │ (bash shell) │ │
│ └─────────────────┘ │
└─────────────────────────────────────────────────────────────┘
```
## File Changes
### New Files
- `apps/api/src/services/terminal_session_v2.py` — New TerminalSession implementation
### Modified Files
- `apps/api/src/services/terminal_session.py` — Delete (or keep as legacy, user said no legacy needed)
- `apps/api/src/services/terminal_manager.py` — Update to use new TerminalSession
- `apps/api/src/api/terminal.py` — Update WebSocket handler for binary frames + flow control
- `apps/web/src/components/terminal.tsx` — Binary mode, WebGL, flow control ack
- `apps/web/package.json` — Add `xterm-addon-webgl`
## TerminalSession Implementation
```python
class TerminalSession:
"""High-performance terminal session with asyncio-native I/O."""
BUFFER_SIZE = 10 * 1024
IDLE_TIMEOUT = 30 * 60
BATCH_WINDOW_MS = 2
FLOW_CONTROL_THRESHOLD = 64 * 1024
FLOW_CONTROL_RESUME = 32 * 1024
def __init__(self, session_id, instance_id, container_id, ...):
self._master_fd: int | None = None
self._process: asyncio.subprocess.Process | None = None
self._output_buffer: deque[bytes] = deque(maxlen=self.BUFFER_SIZE)
self._websockets: set[WebSocket] = set()
self._batch_buffer = bytearray()
self._batch_timer: asyncio.TimerHandle | None = None
self._unacknowledged_bytes = 0
self._paused = False
self._read_handler_set = False
async def start(self):
self._master_fd, slave_fd = pty.openpty()
self._set_terminal_size(80, 24)
self._process = await asyncio.create_subprocess_exec(
"docker", "exec", "-it", "-e", "TERM=xterm-256color",
self.container_id, "bash", "-il",
stdin=slave_fd, stdout=slave_fd, stderr=slave_fd,
)
os.close(slave_fd)
def _start_reading(self):
"""Register fd with asyncio event loop for event-driven reading."""
if self._read_handler_set or self._master_fd is None:
return
loop = asyncio.get_event_loop()
loop.add_reader(self._master_fd, self._on_fd_readable)
self._read_handler_set = True
def _on_fd_readable(self):
"""Callback when PTY fd has data available."""
if self._master_fd is None or self._paused:
return
try:
data = os.read(self._master_fd, 4096)
if data:
self._add_to_buffer(data)
self._queue_output(data)
self.last_activity = time.time()
except (OSError, IOError):
pass
def _queue_output(self, data: bytes):
"""Add to batch buffer, schedule flush."""
self._batch_buffer.extend(data)
self._unacknowledged_bytes += len(data)
if self._unacknowledged_bytes > self.FLOW_CONTROL_THRESHOLD:
self._pause_output()
if self._batch_timer is None:
loop = asyncio.get_event_loop()
self._batch_timer = loop.call_later(
self.BATCH_WINDOW_MS / 1000, self._flush_batch
)
def _flush_batch(self):
"""Flush batched output to all WebSockets."""
self._batch_timer = None
if not self._batch_buffer:
return
payload = bytes(self._batch_buffer)
self._batch_buffer.clear()
dead = set()
for ws in self._websockets:
try:
asyncio.create_task(ws.send_bytes(payload))
except Exception:
dead.add(ws)
self._websockets -= dead
def acknowledge_data(self, char_count: int):
"""Client acknowledges processed bytes."""
self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
if self._paused and self._unacknowledged_bytes < self.FLOW_CONTROL_RESUME:
self._resume_output()
def _pause_output(self):
"""Pause reading from PTY."""
if self._read_handler_set and self._master_fd is not None:
loop = asyncio.get_event_loop()
loop.remove_reader(self._master_fd)
self._read_handler_set = False
self._paused = True
def _resume_output(self):
"""Resume reading from PTY."""
self._paused = False
self._start_reading()
```
## Frontend Changes
### WebSocket Binary Mode
```typescript
const ws = new WebSocket(wsUrl);
ws.binaryType = "arraybuffer"; // Receive ArrayBuffer directly
ws.onmessage = (event) => {
if (event.data instanceof ArrayBuffer) {
const data = new Uint8Array(event.data);
termRef.current?.write(data);
// Flow control: acknowledge processed bytes
ackAccumulator += data.length;
if (ackAccumulator >= 4096) {
ws.send(JSON.stringify({ type: "ack", chars: ackAccumulator }));
ackAccumulator = 0;
}
}
};
```
### WebGL Renderer
```typescript
import { WebglAddon } from "xterm-addon-webgl";
const webglAddon = new WebglAddon();
try {
term.loadAddon(webglAddon);
} catch (e) {
console.warn("WebGL failed, using DOM renderer", e);
}
```
## WebSocket Protocol
### Message Types
**Client → Server:**
- `{"type": "input", "data": "base64_encoded"}` — Keystrokes
- `{"type": "resize", "cols": 80, "rows": 24}` — Resize
- `{"type": "ack", "chars": 4096}` — Flow control acknowledgment
- `{"type": "reset"}` — Reset session
**Server → Client:**
- Binary frame: raw terminal output bytes
- `{"type": "status", "status": "connected"}` — Connection status
- `{"type": "ping"}` — Heartbeat (server → client)
## Testing
### Unit Tests
- `test_terminal_session_v2.py` — Test batching, flow control, resize, reset
- `test_terminal_manager.py` — Test session lifecycle with new session class
### Integration Tests
- `test_terminal_websocket.py` — Full WebSocket round-trip
### Performance Tests
- `benchmark_terminal_latency.py` — Measure input/output latency
- `benchmark_terminal_throughput.py` — Measure max throughput
## Rollback Plan
Since this is a full rewrite with no legacy support:
- Keep a backup branch of the old terminal code
- Feature flag in frontend: `?terminal=v2` to test before full rollout
- Monitor error rates after deployment
@@ -0,0 +1,160 @@
# SDD Exploration: Responsive Web Terminal
## Status
**Phase:** explore
**Date:** 2026-06-02
**Owner:** el Gentleman (parent session)
**Scope:** Terminal I/O latency, rendering performance, connection stability
## Goal
Achieve VS Code Server-level terminal responsiveness: near-local latency on keystrokes, smooth scrolling, no jank on output bursts, and instant resize reactions.
## Current Architecture
### Data Flow
```
Container shell → docker exec PTY → host PTY master fd → select.select(0.1s)
→ Python read loop (10ms sleep fallback) → WebSocket.send_bytes()
→ WebSocket (Blob mode) → frontend arrayBuffer decode → xterm.js.write()
```
### Key Files
| File | Responsibility |
|------|---------------|
| `apps/web/src/components/terminal.tsx` | xterm.js, WebSocket client, FitAddon |
| `apps/api/src/api/terminal.py` | WebSocket endpoint, auth, read/write/heartbeat loops |
| `apps/api/src/services/terminal_session.py` | PTY creation, docker exec subprocess, I/O |
| `apps/api/src/services/terminal_manager.py` | Session lifecycle, persistence, idle cleanup |
### Current Bottlenecks
#### 1. Blocking Read with 100ms Timeout
```python
# terminal_session.py:read_output()
readable, _, _ = select.select([self._master_fd], [], [], 0.1)
if readable:
data = os.read(self._master_fd, 4096)
```
**Problem:** `select.select` blocks up to 100ms when no data is available. With the read loop in `terminal.py` doing `asyncio.sleep(0.01)` between calls, worst-case latency from shell output to WebSocket is ~110ms.
**VS Code approach:** node-pty uses libuv's epoll/kqueue watchers — event-driven, no polling timeout.
#### 2. WebSocket Blob → arrayBuffer Conversion
```typescript
// terminal.tsx
ws.onmessage = (event) => {
if (event.data instanceof Blob) {
event.data.arrayBuffer().then((buffer) => {
const data = new Uint8Array(buffer);
termRef.current?.write(data);
});
}
}
```
**Problem:** Blob → arrayBuffer is async and adds microtask latency. Also forces GC pressure from transient Blobs.
**VS Code approach:** Uses `ws.binaryType = "arraybuffer"` — receives ArrayBuffer directly, zero-copy into Uint8Array.
#### 3. No asyncio-Native PTY Reading
The PTY master fd is read with synchronous `os.read()` inside an async coroutine. This blocks the event loop thread for the duration of the read.
**VS Code approach:** node-pty's C++ binding hooks into libuv's event loop natively — true async I/O.
#### 4. Docker Exec Subprocess Overhead
```python
# terminal_session.py:start()
self.process = await asyncio.create_subprocess_exec(
"docker", "exec", "-it", "-e", "TERM=xterm",
self.container_id, "bash", "-c", shell_cmd,
stdin=self._slave_fd, stdout=self._slave_fd, stderr=self._slave_fd,
)
```
**Problem:** Spawns a new `docker exec` process on the host. Adds process startup latency and an extra process hop.
**Alternative:** Docker Engine API's `attach` endpoint with `logs=0&stream=1&stdin=1&stdout=1&stderr=1` — streams directly to the API container via Unix socket. No host subprocess.
#### 5. No Flow Control / Backpressure
If a command dumps output faster than the WebSocket can send (e.g., `cat /dev/urandom | base64`), data piles up in:
- The PTY kernel buffer (limited, ~4KB)
- Python's deque circular buffer (10KB)
- WebSocket's internal buffer (unbounded in some implementations)
- xterm.js parser queue
**VS Code approach:** Implements explicit flow control — pauses the PTY when the client buffer exceeds a threshold, resumes when drained.
#### 6. xterm.js Renderer
Current: DOM renderer (default).
**VS Code approach:** Canvas renderer with WebGL addon for GPU-accelerated rendering.
## Measurement Baseline
Before optimization, we need metrics:
| Metric | How to Measure | Target |
|--------|---------------|--------|
| Input latency | Time from keypress to character appearing | < 16ms (1 frame) |
| Output throughput | Bytes/sec for `cat /dev/zero` | > 1 MB/s |
| Resize latency | Time from resize message to shell reacting | < 50ms |
| Reconnection time | Time from disconnect to full replay | < 200ms |
| Frame drops | Dropped frames during `yes` command | 0 |
## Improvement Directions
### Direction A: Low-Latency Read Loop (Quick Win)
Replace `select.select` + `os.read` with `asyncio` native approach:
- Use `loop.add_reader()` to register a callback when fd is readable
- Or use `asyncio.to_thread()` with blocking `os.read` and immediate wake
- Eliminate the 100ms timeout and 10ms sleep
### Direction B: WebSocket Binary Mode (Quick Win)
Set `ws.binaryType = "arraybuffer"` on frontend, send binary frames directly.
Eliminates Blob → arrayBuffer conversion.
### Direction C: Docker Engine API Attach (Medium)
Replace `docker exec` subprocess with direct container attach via Docker SDK or HTTP API:
```python
from docker import DockerClient
client = DockerClient()
container = client.containers.get(container_id)
socket = container.attach_socket(params={...})
# socket is a raw TCP/Unix socket — read with asyncio
```
**Pros:** No subprocess overhead, direct stream to container
**Cons:** Requires Docker SDK or raw HTTP over Unix socket; needs `docker` group permissions
### Direction D: Flow Control (Medium)
Add backpressure mechanism:
1. Measure WebSocket send buffer depth on backend
2. Pause reading from PTY when buffer exceeds threshold (e.g., 64KB)
3. Resume when buffer drains below threshold
4. Frontend: measure xterm.js parser queue depth, pause via control message
### Direction E: WebGL Renderer (Quick Win)
Add xterm-addon-webgl:
```typescript
import { WebglAddon } from 'xterm-addon-webgl';
term.loadAddon(new WebglAddon());
```
**Pros:** GPU-accelerated, much faster for large output bursts
**Cons:** Falls back to canvas/DOM if WebGL unavailable; slightly higher init time
### Direction F: Output Batching (Quick Win)
Batch small writes before sending over WebSocket:
- Collect output for 1-2ms
- Send as single binary frame
- Reduces WebSocket frame overhead for high-frequency small writes (e.g., progress bars)
## Recommended Next Steps
1. **Measure baseline** with synthetic benchmarks
2. **Implement Directions A + B + F** (low-risk, high-impact)
3. **Evaluate Direction C** (Docker API attach) vs keeping docker exec
4. **Add Direction D** (flow control) if throughput tests show issues
5. **Add Direction E** (WebGL) as frontend enhancement
## Risks
- Docker API attach may not support PTY mode as cleanly as `docker exec -it`
- WebGL addon may have compatibility issues on older GPUs
- Flow control adds complexity; premature optimization risk
- Changes to core I/O loop could introduce stability regressions
@@ -0,0 +1,71 @@
# Proposal: High-Performance Web Terminal
## Status
**Phase:** proposal → spec → design → tasks → apply
**Date:** 2026-06-02
**Owner:** el Gentleman
**Scope:** Terminal I/O latency, rendering performance, connection stability
## Problem
The current web terminal has noticeable latency on keystrokes, choppy scrolling, and poor performance during output bursts. Users report it feels "slow" compared to VS Code Server's terminal, which feels almost local.
## Goals
| Metric | Current | Target | How Measured |
|--------|---------|--------|--------------|
| Input latency (keypress → char visible) | ~110ms | < 16ms (1 frame) | `term.write()` timestamp diff |
| Output throughput (`cat /dev/zero`) | ~200KB/s | > 1 MB/s | Bytes/sec over 5s |
| Resize latency | ~200ms | < 50ms | Time from resize msg to shell SIGWINCH |
| Reconnection + replay | ~2s | < 300ms | Time from WS open to first rendered char |
| Frame drops during `yes` | Many | 0 | `requestAnimationFrame` counter |
## Non-Goals
- Changing the terminal UI/UX (chrome, controls, tabs)
- Adding new terminal features (search, multi-cursor, etc.)
- Changing authentication or session persistence model
- Supporting non-Docker container runtimes
## Constraints
- Must work with existing tool instance lifecycle (docker containers)
- Must preserve WebSocket-based architecture
- Must preserve session persistence across reconnections
- Must work in both development and production compose setups
## Solution Overview
Replace the blocking `select.select()` PTY read loop with asyncio-native event-driven I/O. Replace `docker exec` subprocess with Docker Engine API attach. Switch WebSocket to binary mode. Add output batching. Add WebGL renderer.
## Key Decisions
1. **Keep `docker exec` for now** — Docker SDK attach doesn't support PTY mode as cleanly. We can optimize the subprocess approach with proper fd handling.
2. **Use `asyncio.add_reader()`** — Native asyncio event-driven fd reading eliminates polling latency.
3. **Binary WebSocket frames**`ws.binaryType = "arraybuffer"` eliminates Blob conversion overhead.
4. **WebGL renderer with DOM fallback** — GPU acceleration where available, graceful fallback.
5. **Output batching with 2ms window** — Collect small writes before sending to reduce frame overhead.
6. **Flow control v2** — Client acknowledges processed bytes; server pauses reads when buffer is full.
## Risks
- **Event loop blocking**: `asyncio.add_reader()` on a PTY fd may not work on all platforms (should work on Linux)
- **WebGL compatibility**: Some GPUs/drivers may fail WebGL context creation
- **Docker exec subprocess**: Still adds overhead; may revisit Docker API attach in future
- **Full rewrite**: Large change surface; thorough testing required
## Acceptance Criteria
- [ ] Input latency < 16ms measured with synthetic benchmark
- [ ] Output throughput > 1 MB/s measured with `cat /dev/zero`
- [ ] Resize latency < 50ms
- [ ] Reconnection + replay < 300ms
- [ ] No frame drops during `yes` command
- [ ] All existing terminal tests pass
- [ ] WebGL renderer loads successfully on modern browsers
- [ ] Graceful fallback to DOM renderer if WebGL fails
- [ ] Flow control prevents memory bloat on `cat /dev/urandom | base64`
## Related
- `openspec/explorations/terminal-responsiveness.md` — Detailed bottleneck analysis
+176
View File
@@ -0,0 +1,176 @@
# Spec: High-Performance Web Terminal
## Overview
Complete rewrite of the terminal I/O pipeline for sub-frame latency and smooth rendering.
## Architecture
### Data Flow (New)
```
Container shell → docker exec PTY → host PTY master fd
→ asyncio.add_reader() callback (event-driven, zero polling)
→ output batcher (2ms window) → WebSocket.send_bytes()
→ WebSocket binary frame → frontend ArrayBuffer
→ xterm.js WebGL renderer → screen
```
### Components
#### 1. TerminalSession (backend)
**Responsibilities:**
- Create PTY via `pty.openpty()`
- Spawn `docker exec -it` with slave fd attached
- Read from PTY master fd using `asyncio.add_reader()`
- Batch output (2ms window) before sending to WebSocket
- Handle flow control (pause/resume reads based on client ack)
- Resize via `TIOCSWINSZ` + `SIGWINCH`
**Interface:**
```python
class TerminalSession:
async def start(self) -> None
async def read_loop(self, websocket) -> None # event-driven
async def write_input(self, data: bytes) -> None
async def resize(self, cols: int, rows: int) -> None
async def reset(self) -> None
async def close(self) -> None
# Flow control
def acknowledge_data(self, char_count: int) -> None
def pause_output(self) -> None
def resume_output(self) -> None
```
#### 2. TerminalManager (backend)
Unchanged responsibilities (session lifecycle, persistence, idle cleanup).
#### 3. WebSocket Handler (backend)
**Messages:**
| Direction | Type | Payload | Description |
|-----------|------|---------|-------------|
| C → S | `input` | `{"data": "base64"}` | Keystrokes / input |
| C → S | `resize` | `{"cols": 80, "rows": 24}` | Terminal resize |
| C → S | `ack` | `{"chars": 1024}` | Flow control ack |
| C → S | `reset` | `{}` | Reset session |
| S → C | `binary` | raw bytes | Terminal output |
| S → C | `status` | `{"status": "connected"}` | Connection status |
| S → C | `ping` | `{}` | Heartbeat |
**Key changes:**
- Output is sent as **binary WebSocket frames**, not Blob
- Flow control: server tracks unacknowledged bytes, pauses PTY reads at 64KB threshold
#### 4. TerminalComponent (frontend)
**Key changes:**
- `ws.binaryType = "arraybuffer"` before connection
- Binary frames written directly to xterm.js as `Uint8Array`
- Flow control: send `ack` messages every 4096 processed bytes
- WebGL renderer with DOM fallback
- Batch resize messages (debounce 50ms)
**xterm.js config:**
```typescript
const term = new Terminal({
cursorBlink: true,
fontSize: currentFontSize,
fontFamily: 'Menlo, Monaco, "Courier New", monospace',
lineHeight: 1.2,
letterSpacing: 0,
allowTransparency: false,
scrollback: 10000,
// Performance options
ignoreBracketedPasteMode: false,
fastScrollSensitivity: 5,
scrollSensitivity: 1,
});
```
### Flow Control Protocol
**Server-side buffer tracking:**
```python
self._unacknowledged_bytes = 0
self._flow_control_threshold = 64 * 1024 # 64KB
self._paused = False
def on_output(self, data: bytes) -> None:
self._unacknowledged_bytes += len(data)
if self._unacknowledged_bytes > self._flow_control_threshold:
self.pause_output()
def acknowledge_data(self, char_count: int) -> None:
self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
if self._paused and self._unacknowledged_bytes < self._flow_control_threshold / 2:
self.resume_output()
```
**Client-side ack strategy:**
- After every `term.write(data)`, accumulate processed bytes
- Send `ack` message every 4096 bytes or 100ms
### Output Batching
**Server-side batcher:**
```python
self._batch_buffer = bytearray()
self._batch_timer: asyncio.TimerHandle | None = None
self._batch_window_ms = 2
def queue_output(self, data: bytes) -> None:
self._batch_buffer.extend(data)
if self._batch_timer is None:
self._batch_timer = asyncio.get_event_loop().call_later(
self._batch_window_ms / 1000, self._flush_batch
)
async def _flush_batch(self) -> None:
self._batch_timer = None
if self._batch_buffer and websocket.open:
await websocket.send_bytes(bytes(self._batch_buffer))
self._batch_buffer.clear()
```
## Docker Exec Subprocess
**Command:**
```bash
docker exec -it -e TERM=xterm-256color <container_id> bash -il
```
**Why keep docker exec:**
- Docker SDK `attach()` doesn't support PTY allocation cleanly
- `docker exec -it` is the standard way to get an interactive TTY
- Subprocess overhead is acceptable compared to PTY latency improvements
**Optimization:** Pre-warm the connection by reusing the same `docker exec` process for the session lifetime.
## Error Handling
| Scenario | Behavior |
|----------|----------|
| WebGL init fails | Fall back to DOM renderer, log warning |
| Flow control ack lost | Server resumes after timeout (5s) |
| PTY fd closed | Close WebSocket with code 4004 |
| Docker exec exits | Close WebSocket with code 4004, allow reconnect |
| Binary frame too large | Split into multiple frames (max 64KB) |
## Testing Strategy
1. **Unit tests:** Mock PTY fd, verify batching, flow control, resize
2. **Integration tests:** Full WebSocket round-trip with test container
3. **Performance tests:**
- `time cat /dev/zero | head -c 10M` — measure throughput
- Rapid keypress script — measure input latency
- Resize storm — measure resize latency
4. **Browser tests:** WebGL fallback on devices without GPU
## Migration
Full rewrite — no migration needed. Old terminal code can be deleted.
+60
View File
@@ -0,0 +1,60 @@
# Tasks: High-Performance Web Terminal
## Task 1: Rewrite TerminalSession with asyncio-native I/O
**Status:** pending
**Files:** `apps/api/src/services/terminal_session.py` (full rewrite)
**Description:**
- Replace `select.select()` with `asyncio.add_reader()` for event-driven PTY reading
- Add output batching (2ms window)
- Add flow control (pause/resume based on client ack)
- Keep docker exec subprocess (optimized)
- Remove circular buffer (not needed with event-driven architecture)
## Task 2: Update TerminalManager for new session class
**Status:** pending
**Files:** `apps/api/src/services/terminal_manager.py`
**Description:**
- Update imports to use rewritten TerminalSession
- Verify session lifecycle methods still work
- Update DB persistence calls
## Task 3: Update WebSocket endpoint for binary frames + flow control
**Status:** pending
**Files:** `apps/api/src/api/terminal.py`
**Description:**
- Accept binary output frames from TerminalSession
- Handle `ack` flow control messages from client
- Send `ping` heartbeat
- Maintain existing auth and session management
## Task 4: Update frontend for binary WebSocket + WebGL
**Status:** pending
**Files:** `apps/web/src/components/terminal.tsx`, `apps/web/package.json`
**Description:**
- Set `ws.binaryType = "arraybuffer"`
- Send flow control `ack` messages
- Add xterm-addon-webgl with DOM fallback
- Optimize resize handling
## Task 5: Add performance benchmarks
**Status:** pending
**Files:** `apps/api/tests/benchmark_terminal.py`
**Description:**
- Input latency benchmark
- Output throughput benchmark
- Resize latency benchmark
- Reconnection time benchmark
## Task 6: Update/fix unit tests
**Status:** pending
**Files:** `apps/api/tests/unit/test_tool_instances_legacy.py`, new tests
**Description:**
- Fix any tests broken by terminal changes
- Add tests for new TerminalSession features
## Task 7: Run full test suite
**Status:** pending
**Description:**
- Run all API tests
- Verify no regressions
- Report quality gate results