Files
headquarter/apps/api/src/api/system/terminal.py
T
alex 37ccaa4fdc refactor: organize API routers and services into subpackages
Service organization (19 files moved into 6 subpackages):
- services/instance/ — event_bus, health_monitor, lifecycle_hooks
- services/config/ — config_profile_resolver
- services/git/ — clone, git_operations, git_service
- services/build/ — docker_build, manifest_compiler
- services/terminal/ — terminal_manager, terminal_session
- services/shared/ — correlation, file_service, notification_service,
  permission_fixer, readiness_probe, ssh_keys, tunnel, workspace_manager

API router organization (16 files moved into 6 subpackages):
- api/tool/ — tool_instances, tool_types, tool_definitions,
  tool_types_validation, sessions (extracted from tool_instances)
- api/config/ — config_profiles, user_config
- api/workspace/ — workspaces, workspace_files, workspace_git,
  workspace_instances
- api/user/ — users, auth, ssh_keys
- api/project/ — projects, git_repositories
- api/system/ — health, events, notifications, dashboard, terminal,
  instance_proxy

Updated main.py imports and all __init__.py re-exports.
Sessions router extracted from tool_instances.py into api/tool/sessions.py.

Quality gates: py_compile passed, ruff passed.
2026-06-04 12:24:14 +02:00

750 lines
25 KiB
Python

"""WebSocket terminal endpoint for tool instances."""
import asyncio
import json
import logging
import uuid
from contextlib import suppress
from fastapi import APIRouter, Depends, HTTPException, WebSocket, status
from sqlalchemy import select
from sqlalchemy.ext.asyncio import AsyncSession
from starlette.websockets import WebSocketDisconnect
from src.auth.dependencies import get_current_user_id, get_db_session
from src.models import TerminalSessionModel
from src.models import ToolInstance
from src.models import ToolType
from src.services.terminal.terminal_manager import MaxSessionsExceededError, terminal_manager
router = APIRouter()
logger = logging.getLogger(__name__)
class SessionRef:
"""Mutable reference to a terminal session, allowing updates during reset."""
def __init__(self, session, slot_session_id: str | None = None):
self.session = session
self.slot_session_id = slot_session_id or session.session_id
@router.websocket(
"/ws/tool-instances/{instance_id}/terminal",
)
async def terminal_websocket_default(
websocket: WebSocket,
instance_id: str,
db_session: AsyncSession = Depends(get_db_session),
) -> None:
"""WebSocket endpoint for terminal access (default session alias).
Backward-compatible route that maps to the default session.
"""
await _handle_terminal_websocket(websocket, instance_id, None, db_session)
@router.websocket(
"/ws/tool-instances/{instance_id}/terminal/{session_id}",
)
async def terminal_websocket_specific(
websocket: WebSocket,
instance_id: str,
session_id: str,
db_session: AsyncSession = Depends(get_db_session),
) -> None:
"""WebSocket endpoint for a specific terminal session."""
await _handle_terminal_websocket(websocket, instance_id, session_id, db_session)
async def _handle_terminal_websocket(
websocket: WebSocket,
instance_id: str,
target_session_id: str | None,
db_session: AsyncSession,
) -> None:
"""Shared WebSocket handler for terminal sessions.
Args:
websocket: The WebSocket connection.
instance_id: UUID string of the tool instance.
target_session_id: Specific session ID (slot key). None means default session.
db_session: Database session.
"""
logger.debug(
"Terminal WebSocket connection attempt for instance %s (session=%s)",
instance_id,
target_session_id or "default",
)
await websocket.accept()
logger.debug("Terminal WebSocket accepted for instance %s", instance_id)
try:
# Parse instance_id
instance_uuid = uuid.UUID(instance_id)
except ValueError:
logger.error("Invalid instance ID: %s", instance_id)
await websocket.close(code=4001, reason="Invalid instance ID")
return
# Authenticate user from session cookie
user_id = await _get_user_from_websocket(websocket, db_session)
if user_id is None:
logger.warning(
"Unauthorized terminal access attempt for instance %s", instance_id
)
await websocket.close(code=4003, reason="Unauthorized")
return
# Get instance and verify ownership
instance = await db_session.get(ToolInstance, instance_uuid)
if instance is None:
logger.warning("Instance %s not found", instance_id)
await websocket.close(code=4004, reason="Instance not found")
return
if instance.owner_id != user_id:
logger.warning(
"Forbidden terminal access for instance %s by user %s",
instance_id,
user_id,
)
await websocket.close(code=4003, reason="Forbidden")
return
if instance.status != "running" or not instance.container_id:
logger.warning(
"Instance %s not running (status=%s, container_id=%s)",
instance_id,
instance.status,
instance.container_id,
)
await websocket.close(code=4004, reason="Instance not running")
return
logger.debug("Terminal auth passed for instance %s, user %s", instance_id, user_id)
# Verify the container actually exists (may have been removed/recreated)
from src.services.docker import get_container_status
container_status = get_container_status(instance.container_id)
if container_status["status"] == "not_found":
logger.error(
"Container %s for instance %s not found (may have been removed)",
instance.container_id,
instance_id,
)
await websocket.close(
code=4004, reason="Container not found — restart the tool instance"
)
return
# Fetch tool type to get startup_command
tool_type = await db_session.get(ToolType, instance.tool_type_id)
startup_command = tool_type.startup_command if tool_type else None
if startup_command:
logger.debug(
"Using startup command for instance %s: %s",
instance_id,
startup_command,
)
session = None
# Get or create terminal session
try:
if target_session_id is None:
# Default session alias
session = await terminal_manager.get_or_create_session(
instance_uuid,
instance.container_id,
startup_command=startup_command,
)
slot_session_id = "default"
else:
# Specific session
session = terminal_manager.get_session(
instance_id,
target_session_id,
)
if session is None:
# Session not in memory — may have been lost on server restart.
# Try to restore from the DB row.
db_row = await db_session.get(
TerminalSessionModel, uuid.UUID(target_session_id)
)
if (
db_row is not None
and db_row.instance_id == instance_uuid
and db_row.status != "closed"
):
logger.info(
"Restoring terminal session %s for instance %s from DB",
target_session_id,
instance_id,
)
session = await terminal_manager.create_session(
instance_uuid,
instance.container_id,
startup_command=startup_command,
name=db_row.name,
session_id=target_session_id,
)
else:
logger.warning(
"Session %s not found for instance %s",
target_session_id,
instance_id,
)
await websocket.close(code=4004, reason="Session not found")
return
# Determine slot key for reset scoping
key = terminal_manager._find_key_by_internal_id(
instance_id, session.session_id
)
slot_session_id = key[1] if key else target_session_id
logger.debug(
"Terminal session ready for instance %s (session_id=%s, slot=%s)",
instance_id,
session.session_id,
slot_session_id,
)
# Attach WebSocket to session
await terminal_manager.attach_websocket(session, websocket)
logger.debug("WebSocket attached to session for instance %s", instance_id)
# Send connected status
await websocket.send_json({"type": "status", "status": "connected"})
logger.debug("Sent connected status for instance %s", instance_id)
# Use mutable session reference so loops can survive reset
session_ref = SessionRef(session, slot_session_id)
# Start write loop and heartbeat (read is now event-driven in TerminalSession)
write_task = asyncio.create_task(
_write_loop(session_ref, websocket, instance_id)
)
heartbeat_task = asyncio.create_task(_heartbeat_loop(websocket))
logger.debug("Started terminal loops for instance %s", instance_id)
# Wait for either task to complete (indicating disconnect or error)
done, pending = await asyncio.wait(
[write_task, heartbeat_task],
return_when=asyncio.FIRST_COMPLETED,
)
logger.debug(
"Terminal loop completed for instance %s, done=%s",
instance_id,
len(done),
)
# Cancel remaining tasks
for task in pending:
task.cancel()
except WebSocketDisconnect:
logger.debug("WebSocket disconnected for instance %s", instance_id)
except Exception as exc:
logger.error(
"Terminal session error for instance %s: %s",
instance_id,
str(exc),
exc_info=True,
)
with suppress(Exception):
await websocket.close(code=4000, reason=f"Error: {exc}")
finally:
# Detach WebSocket, don't kill session
with suppress(Exception):
if session is not None:
await terminal_manager.detach_websocket(session, websocket)
logger.debug(
"WebSocket detached from session for instance %s", instance_id
)
async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> None:
"""Read input from WebSocket and send to container."""
try:
while True:
session = session_ref.session
if not session.is_alive() or session._closed:
await asyncio.sleep(0.1)
continue
message = await websocket.receive()
if message["type"] == "websocket.receive":
if "bytes" in message:
await session.write_input(message["bytes"])
elif "text" in message:
text = message["text"]
if text.startswith("{"):
# Control message (JSON)
try:
ctrl = json.loads(text)
msg_type = ctrl.get("type")
if msg_type == "resize":
cols = ctrl.get("cols", 80)
rows = ctrl.get("rows", 24)
logger.debug(
"Received resize message for instance %s: %sx%s",
instance_id,
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":
# Reset terminal session (scoped to current slot)
logger.debug(
"Resetting terminal session for instance %s (slot=%s)",
session.instance_id,
session_ref.slot_session_id,
)
await websocket.send_json(
{"type": "status", "status": "resetting"}
)
# Reset the session scoped to its slot
new_session = await terminal_manager.reset_session(
session.instance_id,
session.container_id,
startup_command=session.startup_command,
session_id=session_ref.slot_session_id,
name=session.name,
)
# Update the mutable session reference
session_ref.session = new_session
# Attach to new session
await terminal_manager.attach_websocket(
new_session, websocket
)
await websocket.send_json(
{"type": "status", "status": "connected"}
)
# Continue the loop with the new session
continue
except json.JSONDecodeError:
# Not a valid JSON control message, treat as regular input
await session.write_input(text.encode("utf-8"))
else:
await session.write_input(text.encode("utf-8"))
elif message["type"] == "websocket.disconnect":
break
except Exception:
pass
async def _heartbeat_loop(websocket: WebSocket) -> None:
"""Send periodic ping messages to detect disconnections."""
try:
while True:
await asyncio.sleep(30) # Ping every 30 seconds
try:
await websocket.send_json({"type": "ping"})
except Exception:
# WebSocket is closed or broken
break
except Exception:
pass
async def _get_terminal_instance(
instance_id: uuid.UUID,
user_id: uuid.UUID,
db_session: AsyncSession,
) -> ToolInstance:
"""Fetch instance and validate auth, ownership, and running status.
Args:
instance_id: UUID of the tool instance.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
The validated ToolInstance.
Raises:
HTTPException: If instance not found, not owned, or not running.
"""
instance = await db_session.get(ToolInstance, instance_id)
if instance is None:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND, detail="Instance not found"
)
if instance.owner_id != user_id:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail="Not authorized to access this instance",
)
if instance.status != "running" or not instance.container_id:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST, detail="Instance is not running"
)
return instance
@router.get(
"/instances/{instance_id}/terminal/sessions",
summary="List terminal sessions",
description="List terminal sessions for a tool instance with live WebSocket state.",
)
async def list_terminal_sessions(
instance_id: uuid.UUID,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""List terminal sessions for an instance.
Args:
instance_id: UUID of the tool instance.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with sessions list.
"""
await _get_terminal_instance(instance_id, user_id, db_session)
# Query active DB rows for this instance
result = await db_session.execute(
select(TerminalSessionModel)
.where(TerminalSessionModel.instance_id == instance_id)
.where(TerminalSessionModel.status != "closed")
.order_by(TerminalSessionModel.created_at.asc())
)
db_rows = result.scalars().all()
# Build response with live has_websockets flag.
# Include DB rows even without in-memory counterparts (e.g. after
# server restart) so the frontend can display tabs and reconnect.
sessions = []
for row in db_rows:
live_session = terminal_manager.get_session(str(instance_id), str(row.id))
sessions.append(
{
"id": str(row.id),
"name": row.name,
"status": row.status,
"has_websockets": live_session.has_websockets()
if live_session
else False,
"created_at": row.created_at.isoformat() if row.created_at else None,
"last_activity_at": row.last_activity_at.isoformat()
if row.last_activity_at
else None,
}
)
return {"sessions": sessions}
@router.post(
"/instances/{instance_id}/terminal/sessions",
summary="Create terminal session",
description="Create a new terminal session for a running tool instance.",
status_code=status.HTTP_201_CREATED,
)
async def create_terminal_session(
instance_id: uuid.UUID,
data: dict,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""Create a new terminal session.
Args:
instance_id: UUID of the tool instance.
data: Request body with optional name.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with new session details.
Raises:
HTTPException: 409 if max sessions reached.
"""
instance = await _get_terminal_instance(instance_id, user_id, db_session)
assert instance.container_id is not None
# Fetch tool type to get startup_command
tool_type = await db_session.get(ToolType, instance.tool_type_id)
startup_command = tool_type.startup_command if tool_type else None
name = data.get("name")
try:
session = await terminal_manager.create_session(
instance_id,
instance.container_id,
startup_command=startup_command,
name=name,
)
except MaxSessionsExceededError:
raise HTTPException(
status_code=status.HTTP_409_CONFLICT,
detail="Maximum of 5 terminal sessions reached for this instance",
) from None
return {
"id": session.session_id,
"name": session.name,
"status": session.status,
"created_at": session.last_activity,
}
@router.delete(
"/instances/{instance_id}/terminal/sessions/{session_id}",
summary="Close terminal session",
description="Close a specific terminal session.",
)
async def close_terminal_session(
instance_id: uuid.UUID,
session_id: str,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""Close a terminal session.
Args:
instance_id: UUID of the tool instance.
session_id: ID of the session to close.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with closure status.
"""
await _get_terminal_instance(instance_id, user_id, db_session)
# Find the session by internal ID to determine its slot key
key = terminal_manager._find_key_by_internal_id(str(instance_id), session_id)
if (
key is None
and terminal_manager.get_session(str(instance_id), session_id) is not None
):
key = (str(instance_id), session_id)
if key is None:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND, detail="Session not found"
)
await terminal_manager.close_session(key[0], key[1])
return {"status": "closed", "session_id": session_id}
@router.post(
"/instances/{instance_id}/terminal/sessions/{session_id}/reset",
summary="Reset terminal session",
description="Reset a specific terminal session, killing the current shell and starting fresh.",
)
async def reset_specific_terminal_session(
instance_id: uuid.UUID,
session_id: str,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""Reset a specific terminal session.
Args:
instance_id: UUID of the tool instance.
session_id: ID of the session to reset.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with reset session details.
"""
instance = await _get_terminal_instance(instance_id, user_id, db_session)
assert instance.container_id is not None
# Determine slot key for reset
key = terminal_manager._find_key_by_internal_id(str(instance_id), session_id)
if (
key is None
and terminal_manager.get_session(str(instance_id), session_id) is not None
):
key = (str(instance_id), session_id)
if key is None:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND, detail="Session not found"
)
# Fetch tool type to get startup_command
tool_type = await db_session.get(ToolType, instance.tool_type_id)
startup_command = tool_type.startup_command if tool_type else None
# Preserve name if possible
live_session = terminal_manager.get_session(str(instance_id), session_id)
name = live_session.name if live_session else None
new_session = await terminal_manager.reset_session(
instance_id,
instance.container_id,
startup_command=startup_command,
session_id=key[1],
name=name,
)
return {
"id": new_session.session_id,
"name": new_session.name,
"status": new_session.status,
}
@router.post(
"/instances/{instance_id}/terminal/sessions/{session_id}/rename",
summary="Rename terminal session",
description="Rename a specific terminal session.",
)
async def rename_terminal_session(
instance_id: uuid.UUID,
session_id: str,
data: dict,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""Rename a terminal session.
Args:
instance_id: UUID of the tool instance.
session_id: ID of the session to rename.
data: Request body with new name.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with updated session details.
"""
await _get_terminal_instance(instance_id, user_id, db_session)
new_name = data.get("name")
if not new_name or not isinstance(new_name, str):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST, detail="Name is required"
)
# Update in-memory session name if live
live_session = terminal_manager.get_session(str(instance_id), session_id)
if live_session:
live_session.name = new_name
# Update DB row
db_row = await db_session.get(TerminalSessionModel, uuid.UUID(session_id))
if db_row is None:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND, detail="Session not found"
)
db_row.name = new_name
await db_session.commit()
return {"id": str(db_row.id), "name": new_name}
@router.post(
"/instances/{instance_id}/terminal/reset",
summary="Reset terminal session (legacy alias)",
description="Reset the default terminal session for a tool instance. Preserved for backward compatibility.",
)
async def reset_terminal_session(
instance_id: uuid.UUID,
user_id: uuid.UUID = Depends(get_current_user_id),
db_session: AsyncSession = Depends(get_db_session),
) -> dict:
"""Reset the default terminal session for an instance (legacy alias).
Args:
instance_id: UUID of the tool instance.
user_id: ID of the authenticated user.
db_session: Database session.
Returns:
Dictionary with status message.
"""
instance = await _get_terminal_instance(instance_id, user_id, db_session)
assert instance.container_id is not None
# Fetch tool type to get startup_command
tool_type = await db_session.get(ToolType, instance.tool_type_id)
startup_command = tool_type.startup_command if tool_type else None
try:
# Reset the default session
new_session = await terminal_manager.reset_session(
instance_id,
instance.container_id,
startup_command=startup_command,
)
logger.info(
"Terminal session reset for instance %s (new session_id=%s)",
instance_id,
new_session.session_id,
)
return {
"status": "success",
"message": "Terminal session reset successfully",
"instance_id": str(instance_id),
"session_id": new_session.session_id,
}
except Exception as exc:
logger.error(
"Failed to reset terminal session for instance %s: %s",
instance_id,
str(exc),
exc_info=True,
)
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Failed to reset terminal session: {exc}",
) from exc
async def _get_user_from_websocket(
websocket: WebSocket,
db_session: AsyncSession,
) -> uuid.UUID | None:
"""Extract and validate user ID from session cookie in WebSocket.
Args:
websocket: The WebSocket connection.
db_session: Database session.
Returns:
The user's UUID if authenticated, None otherwise.
"""
from src.auth.session import decode_session_cookie
from src.config import Settings
session_cookie = websocket.cookies.get("session")
if not session_cookie:
return None
settings = Settings()
try:
payload = decode_session_cookie(settings=settings, cookie_value=session_cookie)
return uuid.UUID(str(payload["user_id"]))
except (ValueError, KeyError):
return None