Feature/on demand plugin filtering (#166)

* fix(web): Resolve font display and config API error handling issues

- Fix font catalog display error where path.startsWith fails
  (path is object, not string)
- Update save_main_config to use error_response() helper
- Improve save_raw_main_config error handling consistency
- Add proper error codes and traceback details to API responses

* fix(web): Prevent fontCatalog redeclaration error on HTMX reload

- Use window object to store global font variables
- Check if script has already loaded before declaring variables
- Update both window properties and local references on assignment
- Fixes 'Identifier fontCatalog has already been declared' error

* fix(web): Wrap fonts script in IIFE to prevent all redeclaration errors

- Wrap entire script in IIFE that only runs once
- Check if script already loaded before declaring variables/functions
- Expose initializeFontsTab to window for re-initialization
- Prevents 'Identifier has already been declared' errors on HTMX reload

* fix(web): Exempt config save API endpoints from CSRF protection

- Exempt save_raw_main_config, save_raw_secrets_config, and save_main_config from CSRF
- These endpoints are called via fetch from JavaScript and don't include CSRF tokens
- Fixes 500 error when saving config via raw JSON editor

* fix(web): Exempt system action endpoint from CSRF protection

- Exempt execute_system_action from CSRF
- Fixes 500 error when using system action buttons (restart display, restart Pi, etc.)
- These endpoints are called via HTMX and don't include CSRF tokens

* fix(web): Exempt all API v3 endpoints from CSRF protection

- Add before_request handler to exempt all api_v3.* endpoints
- All API endpoints are programmatic (HTMX/fetch) and don't include CSRF tokens
- Prevents future CSRF errors on any API endpoint
- Cleaner than exempting individual endpoints

* refactor(web): Remove CSRF protection for local-only application

- CSRF is designed for internet-facing apps to prevent cross-site attacks
- For local-only Raspberry Pi app, threat model is different
- All endpoints were exempted anyway, so it wasn't protecting anything
- Forms use HTMX without CSRF tokens
- If exposing to internet later, can re-enable with proper token implementation

* fix(web): Fix font path double-prefixing in font catalog display

- Only prefix with 'assets/fonts/' if path is a bare filename
- If path starts with '/' (absolute) or 'assets/' (already prefixed), use as-is
- Fixes double-prefixing when get_fonts_catalog returns relative paths like 'assets/fonts/press_start.ttf'

* fix(web): Remove fontsTabInitialized guard to allow re-initialization on HTMX reload

- Remove fontsTabInitialized check that prevented re-initialization on HTMX content swap
- The window._fontsScriptLoaded guard is sufficient to prevent function redeclaration
- Allow initializeFontsTab() to run on each HTMX swap to attach listeners to new DOM elements
- Fixes fonts UI breaking after HTMX reload (buttons, upload dropzone, etc. not working)

* fix(api): Preserve empty strings for optional string fields in plugin config

- Add _is_field_required() helper to check if fields are required in schema
- Update _parse_form_value_with_schema() to preserve empty strings for optional string fields
- Fixes 400 error when saving MQTT plugin config with empty username/password
- Resolves validation error: 'Expected type string, got NoneType'

* fix(config): Add defaults to schemas and fix None value handling

- Updated merge_with_defaults to replace None values with defaults
- Fixed form processing to skip empty optional fields without defaults
- Added script to automatically add defaults to all plugin config schemas
- Added defaults to 89 fields across 10 plugin schemas
- Prevents validation errors from None values in configs

Changes:
- schema_manager.py: Enhanced merge_with_defaults to replace None with defaults
- api_v3.py: Added _SKIP_FIELD sentinel to skip optional fields without defaults
- add_defaults_to_schemas.py: Script to add sensible defaults to schemas
- Plugin schemas: Added defaults for number, boolean, and array fields

* fix(config): Fix save button spinner by checking HTTP status code

- Fixed handleConfigSave to check xhr.status instead of event.detail.successful
- With hx-swap="none", HTMX doesn't set event.detail.successful
- Now properly detects successful saves (status 200-299) and stops spinner
- Improved error message extraction from API responses
- Also fixed handleToggleResponse for consistency

* fix(web-ui): Resolve GitHub token warning persistence after save

- Made checkGitHubAuthStatus() return Promise for proper async handling
- Clear sessionStorage dismissal flag when token is saved
- Add delay before status check to ensure backend token reload
- Wait for status check completion before hiding settings panel

Fixes issue where GitHub token warnings and pop-ups would not
disappear after successfully saving a token in the web UI.

* fix(web-ui): Add token validation and improve GitHub token warning behavior

- Add token validation to backend API endpoint to check if token is valid/expired
- Implement _validate_github_token() method in PluginStoreManager with caching
- Update frontend to show warning only when token is missing or invalid
- Keep settings panel accessible (collapsible) when token is configured
- Collapse settings panel content after successful token save instead of hiding
- Display specific error messages for invalid/expired tokens
- Clear sessionStorage dismissal flag when token becomes valid

Fixes issue where GitHub token warnings and settings panel would not
properly hide/show based on token status. Now validates token validity
and provides better UX with collapsible settings panel.

* fix(web-ui): Fix CSS/display issue for GitHub token warning and settings

- Update all hide/show operations to use both classList and style.display
- Fix checkGitHubAuthStatus() to properly hide/show warning and settings
- Fix dismissGithubWarning() to use both methods
- Fix toggleGithubTokenSettings() with improved state checking
- Fix collapse button handler with improved state checking
- Fix saveGithubToken() to properly show/collapse settings panel

This ensures elements actually hide/show when status changes, matching
the pattern used elsewhere in the codebase (like toggleSection). All
buttons (dismiss, close, collapse) should now work correctly.

* fix(web-ui): Fix GitHub token expand button functionality

- Convert collapse button handler to named function (toggleGithubTokenContent)
- Improve state checking using class, inline style, and computed style
- Re-attach event listener after saving token to ensure it works
- Add console logging for debugging
- Make function globally accessible for better reliability

Fixes issue where expand button didn't work after saving token.

* fix(web-ui): Remove X button and improve GitHub token panel behavior

- Remove X (close) button from GitHub token configuration panel
- Replace toggleGithubTokenSettings() with openGithubTokenSettings() that only opens
- Auto-collapse panel when token is valid (user must click expand to edit)
- Auto-detect token status on page load (no need to click save)
- Simplify saveGithubToken() to rely on checkGitHubAuthStatus() for UI updates
- Ensure expand button works correctly with proper event listener attachment

The panel now remains visible but collapsed when a token is configured,
allowing users to expand it when needed without the ability to completely hide it.

* refactor(web-ui): Improve GitHub token collapse button code quality

- Update comment to reflect actual behavior (prevent parent click handlers)
- Use empty string for display to defer to CSS instead of hard-coding block/none
- Extract duplicate clone-and-attach logic into attachGithubTokenCollapseHandler() helper
- Make helper function globally accessible for reuse in checkGitHubAuthStatus()

Improves maintainability and makes code more future-proof for layout changes.

* fix(web-ui): Fix collapse/expand button by using removeProperty for display

- Use style.removeProperty('display') instead of style.display = ''
- This properly removes inline styles and defers to CSS classes
- Fixes issue where collapse/expand button stopped working after refactor

* fix(web-ui): Make display handling consistent for token collapse

- Use removeProperty('display') consistently in all places
- Fix checkGitHubAuthStatus() to use removeProperty instead of inline style
- Simplify state checking to rely on hidden class with computed style fallback
- Ensures collapse/expand button works correctly by deferring to CSS classes

* fix(web-ui): Fix token collapse button and simplify state detection

- Simplify state checking to rely on hidden class only (element has class='block')
- Only remove inline display style if it exists (check before removing)
- Add console logging to debug handler attachment
- Ensure collapse/expand works by relying on CSS classes

Fixes issues where:
- Collapse button did nothing
- Auto-detection of token status wasn't working

* debug(web-ui): Add extensive debugging for token collapse button

- Add console logs to track function calls and element detection
- Improve state detection to use computed style as fallback
- Add wrapper function for click handler to ensure it's called
- Better error messages to identify why handler might not attach

This will help identify why the collapse button isn't working.

* debug(web-ui): Add comprehensive debugging for GitHub token features

- Add console logs to checkGitHubAuthStatus() to track execution
- Re-attach collapse handler after plugin store is rendered
- Add error stack traces for better debugging
- Ensure handler is attached when content is dynamically loaded

This will help identify why:
- Auto-detection of token status isn't working
- Collapse button isn't functioning

* fix(web-ui): Move checkGitHubAuthStatus before IIFE to fix scope issue

- Move checkGitHubAuthStatus function definition before IIFE starts
- Function was defined after IIFE but called inside it, causing it to be undefined
- Now function is available when called during initialization
- This should fix auto-detection of token status on page load

* debug(web-ui): Add extensive logging to GitHub token functions

- Add logging when checkGitHubAuthStatus is defined
- Add logging when function is called during initialization
- Add logging in attachGithubTokenCollapseHandler
- Add logging in store render callback
- This will help identify why functions aren't executing

* fix(web-ui): Move GitHub token functions outside IIFE for availability

- Move attachGithubTokenCollapseHandler and toggleGithubTokenContent outside IIFE
- These functions need to be available when store renders, before IIFE completes
- Add logging to initializePlugins to track when it's called
- This should fix the 'undefined' error when store tries to attach handlers

* fix(web-ui): Fix GitHub token content collapse/expand functionality

- Element has 'block' class in HTML which conflicts with 'hidden' class
- When hiding: add 'hidden', remove 'block', set display:none inline
- When showing: remove 'hidden', add 'block', remove inline display
- This ensures proper visibility toggle for the GitHub API Configuration section

* feat(display): Implement on-demand plugin filtering with restart

- Add on-demand plugin filtering to DisplayController initialization
  - Filters available_modes to only include on-demand plugin's modes
  - Allows plugin internal rotation (e.g., NFL upcoming, NCAA FB Recent)
  - Prevents rotation to other plugins
- Implement restart mechanism for on-demand activation/clear
  - _restart_with_on_demand_filter() saves state and restarts with filter
  - _restart_without_on_demand_filter() restores normal operation
  - Supports both systemd service and direct process execution
- Add state preservation across restarts
  - Saves/restores rotation position from cache
  - Restores on-demand config from cache after restart
- Add service detection method
  - Detects if running as systemd service
  - Uses file-based approach for environment variable passing
- Update API endpoints with restart flow comments
- Update systemd service file with on-demand support notes
- Add comprehensive error handling for edge cases

* perf(web-ui): Optimize GitHub token detection speed

- Call checkGitHubAuthStatus immediately when script loads (if elements exist)
- Call it early in initPluginsPage (before full initialization completes)
- Use requestAnimationFrame instead of setTimeout(100ms) for store render callback
- Reduce save token delay from 300ms to 100ms
- Token detection now happens in parallel with other initialization tasks
- This makes token status visible much faster on page load

* fix(ui): Move on-demand modal to base.html for always-available access

- Move on-demand modal from plugins.html to base.html
- Ensures modal is always in DOM when Run On-Demand button is clicked
- Fixes issue where button in plugin_config.html couldn't find modal
- Modal is now available regardless of which tab is active

* fix(ui): Initialize on-demand modal unconditionally on page load

- Create initializeOnDemandModal() function that runs regardless of plugins tab
- Modal is in base.html so it should always be available
- Call initialization on DOMContentLoaded and with timeout
- Fixes 'On-demand modal elements not found' error when clicking button
- Modal setup now happens even if plugins tab hasn't been loaded yet

* fix(ui): Add safety check for updatePluginTabStates function

- Check if updatePluginTabStates exists before calling
- Prevents TypeError when function is not available
- Fixes error when clicking plugin tabs

* fix(ui): Add safety checks for all updatePluginTabStates calls

- Add safety check in Alpine component tab button handler
- Add safety check in Alpine  callback
- Prevents TypeError when function is not available in all contexts

* fix(ui): Add safety check in Alpine  callback for updatePluginTabStates

* debug(ui): Add console logging to trace on-demand modal opening

- Add logging to runPluginOnDemand function
- Add logging to __openOnDemandModalImpl function
- Log plugin lookup, modal element checks, and display changes
- Helps diagnose why modal doesn't open when button is clicked

* debug(ui): Add logging for modal display change

* debug(ui): Add more explicit modal visibility settings and computed style logging

- Set visibility and opacity explicitly when showing modal
- Force reflow to ensure styles are applied
- Log computed styles to diagnose CSS issues
- Helps identify if modal is hidden by CSS rules

* debug(ui): Increase modal z-index and add bounding rect check

- Set z-index to 9999 to ensure modal is above all other elements
- Add bounding rect check to verify modal is in viewport
- Helps diagnose if modal is positioned off-screen or behind other elements

* debug(display): Add detailed logging for on-demand restart flow

- Log when polling finds requests
- Log service detection result
- Log file writing and systemctl commands
- Log restart command execution and results
- Helps diagnose why on-demand restart isn't working

* debug(display): Add logging for on-demand request polling

- Log request_id comparison to diagnose why requests aren't being processed
- Helps identify if request_id matching is preventing processing

* fix(ui): Force modal positioning with !important to override any conflicting styles

- Use cssText with !important flags to ensure modal is always visible
- Remove all inline styles first to start fresh
- Ensure modal is positioned at top:0, left:0 with fixed positioning
- Fixes issue where modal was still positioned off-screen (top: 2422px)

* debug(ui): Add logging to on-demand form submission

- Log form submission events
- Log payload being sent
- Log response status and data
- Helps diagnose why on-demand requests aren't being processed

* fix(display): Remove restart-based on-demand activation

- Replace restart-based activation with immediate mode switch
- On-demand now activates without restarting the service
- Saves rotation state for restoration when on-demand ends
- Fixes infinite restart loop issue
- On-demand now works when display is already running

* docs: Add comprehensive guide for on-demand cache management

- Document all on-demand cache keys and their purposes
- Explain when manual clearing is needed
- Clarify what clearing from cache management tab does/doesn't do
- Provide troubleshooting steps and best practices

* fix(display): Ensure on-demand takes priority over live priority

- Move on-demand check BEFORE live priority check
- Add explicit logging when on-demand overrides live priority
- Improve request_id checking with both instance and persisted checks
- Add debug logging to trace why requests aren't being processed
- Fixes issue where on-demand didn't interrupt live NHL game

* fix(display): Ensure on-demand takes priority over live priority

- Move on-demand check BEFORE live priority check in main loop
- Add explicit logging when on-demand overrides live priority
- Fixes issue where on-demand didn't interrupt live NHL game

* fix(display): Improve on-demand request processing and priority

- Add persistent processed_id check to prevent duplicate processing
- Mark request as processed BEFORE processing to prevent race conditions
- Improve logging to trace request processing
- Ensure on-demand takes priority over live priority (already fixed in previous commit)

* fix(display): Remove duplicate action line

* fix(display): Fix live priority and ensure on-demand overrides it

- Fix live priority to properly set active_mode when live content is detected
- Ensure on-demand check happens before live priority check
- Add debug logging to trace on-demand vs live priority
- Fix live priority to stay on live mode instead of rotating

* fix(display): Add debug logging for on-demand priority check

* fix(display): Add better logging for on-demand request processing

- Add logging to show when requests are blocked by processed_id check
- Add logging to show on-demand state after activation
- Helps debug why on-demand requests aren't being processed

* fix(display): Add detailed logging for on-demand activation and checking

- Log on-demand state after activation to verify it's set correctly
- Add debug logging in main loop to trace on-demand check
- Helps identify why on-demand isn't overriding live priority

* fix(display): Add debug logging for on-demand check in main loop

* fix(display): Remove restart logic from _clear_on_demand and fix cache delete

- Replace cache_manager.delete() with cache_manager.clear_cache()
- Remove restart logic from _clear_on_demand - now clears immediately
- Restore rotation state immediately without restarting
- Fixes AttributeError: 'CacheManager' object has no attribute 'delete'

* fix(display): Remove restart logic from _clear_on_demand

- Remove restart logic - now clears on-demand state immediately
- Restore rotation state immediately without restarting
- Use clear_cache instead of delete (already fixed in previous commit)
- Fixes error when stopping on-demand mode

* feat(display): Clear display before activating on-demand mode

- Clear display and reset state before activating on-demand
- Reset dynamic mode state to ensure clean transition
- Mimics the behavior of manually stopping display first
- Should fix issue where on-demand only works after manual stop

* feat(display): Stop display service before starting on-demand mode

- Stop the display service first if it's running
- Wait 1.5 seconds for clean shutdown
- Then start the service with on-demand request in cache
- Mimics the manual workflow of stopping display first
- Should fix issue where on-demand only works after manual stop

* feat(display): Filter plugins during initialization for on-demand mode

- Check cache for on-demand requests during initialization
- Only load the on-demand plugin if on-demand request is found
- Prevents loading background services for other plugins
- Fixes issue where Hockey/Football data loads even when only Clock is requested

* fix(display): Use filtered enabled_plugins list instead of discovered_plugins

- Use enabled_plugins list which is already filtered for on-demand mode
- Prevents loading all plugins when on-demand mode is active
- Fixes issue where all plugins were loaded even in on-demand mode

* fix(display): Fix on-demand stop request processing and expiration check

- Always process stop requests, even if request_id was seen before
- Fix expiration check to handle cases where on-demand is not active
- Add better logging for stop requests and expiration
- Fixes issue where stop button does nothing and timer doesn't expire

* fix(display): Fix on-demand stop processing, expiration, and plugin filtering

- Fix stop request processing to always process stop requests, bypassing request_id checks
- Fix expiration check logic to properly check on_demand_active and expires_at separately
- Store display_on_demand_config cache key in _activate_on_demand for plugin filtering
- Clear display before switching to on-demand mode to prevent visual artifacts
- Clear display_on_demand_config cache key in _clear_on_demand to prevent stale data
- Implement plugin filtering during initialization based on display_on_demand_config

Fixes issues where:
- Stop button did nothing (stop requests were blocked by request_id check)
- Expiration timer didn't work (logic issue with or condition)
- Plugin filtering didn't work on restart (config cache key never set)
- Display showed artifacts when switching to on-demand (display not cleared)
- All plugins loaded even in on-demand mode (filtering not implemented)

* fix(web): Allow on-demand to work with disabled plugins

- Remove frontend checks that blocked disabled plugins from on-demand
- Backend already supports temporarily enabling disabled plugins during on-demand
- Update UI messages to indicate plugin will be temporarily enabled
- Remove disabled attribute from Run On-Demand button

Fixes issue where disabled plugins couldn't use on-demand feature even
though the backend implementation supports it.

* fix(display): Resolve plugin_id when sent as mode in on-demand requests

- Detect when mode parameter is actually a plugin_id and resolve to first display mode
- Handle case where frontend sends plugin_id as mode (e.g., 'football-scoreboard')
- Add fallback to use first available display mode if provided mode is invalid
- Add logging for mode resolution debugging

Fixes issue where on-demand requests with mode=plugin_id failed with 'invalid-mode' error

* feat(display): Rotate through all plugin modes in on-demand mode

- Store all modes for on-demand plugin instead of locking to single mode
- Rotate through available modes (live, recent, upcoming) when on-demand active
- Skip modes that return False (no content) and move to next mode
- Prioritize live modes if they have content, otherwise skip them
- Add on_demand_modes list and on_demand_mode_index for rotation tracking

Fixes issue where on-demand mode stayed on one mode (e.g., football_recent)
and didn't rotate through other available modes (football_live, football_upcoming).
Now properly rotates through all modes, skipping empty ones.

* fix(display): Improve on-demand stop request handling

- Always process stop requests if on-demand is active, even if same request_id
- Add better logging when stop is requested but on-demand is not active
- Improve logging in _clear_on_demand to show which mode rotation resumes to
- Ensure stop requests are properly acknowledged

Fixes issue where stop button shows as completed but display doesn't resume
normal rotation. Stop requests now properly clear on-demand state and resume.

* security(web): Fix XSS vulnerability in GitHub auth error display

Replace innerHTML usage with safe DOM manipulation:
- Use textContent to clear element and create text nodes
- Create <strong> element via createElement instead of string HTML
- Add safe fallback ('Unknown error') for error messages
- Ensure authData.error/authData.message are treated as plain text
- Avoid trusting backend-provided data as HTML

Fixes XSS vulnerability where malicious HTML in error messages could
be injected into the DOM.

* style(api): Remove unnecessary str() in f-string for error message

Remove explicit str(e) call in error_response f-string since f-strings
automatically convert exceptions to strings. This matches the style used
elsewhere in the file.

Changed: f"Error saving configuration: {str(e)}"
To:      f"Error saving configuration: {e}"

* fix(store): Skip caching for rate-limited 403 responses

When a 403 response indicates a rate limit (detected by checking if
'rate limit' is in response.text.lower()), return the error result but
do NOT cache it in _token_validation_cache. Rate limits are temporary
and should be retried, so caching would incorrectly mark the token as
invalid.

Continue to cache 403 responses that indicate missing token permissions,
as these are persistent issues that should be cached.

This prevents rate-limited responses from being incorrectly cached as
invalid tokens, allowing the system to retry after the rate limit
resets.

* fix(display): Prevent ZeroDivisionError when on_demand_modes is empty

Add guards to check if on_demand_modes is non-empty before performing
any rotation/index math operations. When on_demand_active is True but
on_demand_modes is empty, clear on-demand mode instead of attempting
division by zero.

Fixed in three locations:
1. Mode selection logic (line ~1081): Check before accessing modes
2. Skip to next mode when no content (line ~1190): Guard before modulo
3. Rotate to next mode (line ~1561): Guard before modulo

This prevents ZeroDivisionError when a plugin has no available display
modes or when on_demand_modes becomes empty unexpectedly.

* fix(display): Improve guard for empty on_demand_modes in rotation skip

Refine the guard around lines 1195-1209 to:
- Check if on_demand_modes is empty before any modulo/index operations
- Log warning and debug trace when no modes are configured
- Skip rotation (continue) instead of clearing on-demand mode
- Only perform modulo and index operations when modes are available
- Only log rotation message when next_mode is valid

This prevents ZeroDivisionError and ensures all logging only occurs
when next_mode is valid, providing better traceability.

* fix(display): Populate on_demand_modes when restoring on-demand state from cache

When restoring on-demand state from cache during initialization (around
lines 163-197), the code sets on_demand_active, on_demand_plugin_id and
related fields but does not populate self.on_demand_modes, causing the
run loop to see an empty modes list after restart.

Fix by:
1. Adding _populate_on_demand_modes_from_plugin() method that retrieves
   the plugin's display modes from plugin_display_modes and builds the
   ordered modes list (prioritizing live modes with content, same logic
   as _activate_on_demand)
2. Calling this method after plugin loading completes (around line 296)
   when on_demand_active and on_demand_plugin_id are set
3. Setting on_demand_mode_index to match the restored mode if available,
   otherwise starting at index 0

This ensures on_demand_modes is populated after restart, preventing
empty modes list errors in the run loop.

* docs: Update on-demand documentation to reflect current implementation

Replace obsolete log message reference with current log messages:
- Old: 'Activating on-demand mode... restarting display controller'
- New: 'Processing on-demand start request for plugin' and 'Activated on-demand for plugin'

Update Scenario 2 to reflect immediate mode switching:
- Changed title from 'Infinite Restart Loop' to 'On-Demand Mode Switching Issues'
- Updated symptoms to describe mode switching issues instead of restart loops
- Added note that on-demand now switches modes immediately without restarting
- Updated solution to include display_on_demand_state key

This reflects the current implementation where on-demand activates
immediately without restarting the service.

* fix(api): Fix undefined logger and service stop logic in start_on_demand_display

- Add module-level logger to avoid NameError when logging disabled plugin
- Only stop display service when start_service is True (prevents stopping
  service without restarting when start_service is False)
- Remove unused stop_result variable
- Clean up f-strings that don't need formatting
- Improve code formatting for logger.info call

Fixes issue where logger.info() would raise NameError and where the
service would be stopped even when start_service=False, leaving the
service stopped without restarting it.

---------

Signed-off-by: Chuck <33324927+ChuckBuilds@users.noreply.github.com>
Co-authored-by: Chuck <chuck@example.com>
This commit is contained in:
Chuck
2026-01-01 18:27:58 -05:00
committed by GitHub
parent a5c10d6f78
commit 67197635c9
9 changed files with 820 additions and 149 deletions

View File

@@ -6,9 +6,12 @@ import subprocess
import time
import hashlib
import uuid
import logging
from datetime import datetime
from pathlib import Path
logger = logging.getLogger(__name__)
# Import new infrastructure
from src.web_interface.api_helpers import success_response, error_response, validate_request_json
from src.web_interface.errors import ErrorCode
@@ -634,7 +637,7 @@ def save_main_config():
logging.error(error_msg)
return error_response(
ErrorCode.CONFIG_SAVE_FAILED,
f"Error saving configuration: {str(e)}",
f"Error saving configuration: {e}",
details=traceback.format_exc(),
status_code=500
)
@@ -1228,24 +1231,20 @@ def start_on_demand_display():
if not resolved_plugin:
return jsonify({'status': 'error', 'message': f'Mode {resolved_mode} not found'}), 404
# Note: On-demand can work with disabled plugins - the display controller
# will temporarily enable them during initialization if needed
# We don't block the request here, but log it for debugging
if api_v3.config_manager and resolved_plugin:
config = api_v3.config_manager.load_config()
plugin_config = config.get(resolved_plugin, {})
if 'enabled' in plugin_config and not plugin_config.get('enabled', False):
return jsonify({
'status': 'error',
'message': f'Plugin {resolved_plugin} is disabled in configuration'
}), 400
# Check if display service is running (or will be started)
service_status = _get_display_service_status()
if not service_status.get('active') and not start_service:
return jsonify({
'status': 'error',
'message': 'Display service is not running. Please start the display service or enable "Start Service" option.',
'service_status': service_status
}), 400
logger.info(
"On-demand request for disabled plugin '%s' - will be temporarily enabled",
resolved_plugin,
)
# Set the on-demand request in cache FIRST (before starting service)
# This ensures the request is available when the service starts/restarts
cache = _ensure_cache_manager()
request_id = data.get('request_id') or str(uuid.uuid4())
request_payload = {
@@ -1259,6 +1258,26 @@ def start_on_demand_display():
}
cache.set('display_on_demand_request', request_payload)
# Check if display service is running (or will be started)
service_status = _get_display_service_status()
service_was_running = service_status.get('active', False)
# Stop the display service first to ensure clean state when we will restart it
if service_was_running and start_service:
import time as time_module
print("Stopping display service before starting on-demand mode...")
_stop_display_service()
# Wait a brief moment for the service to fully stop
time_module.sleep(1.5)
print("Display service stopped, now starting with on-demand request...")
if not service_status.get('active') and not start_service:
return jsonify({
'status': 'error',
'message': 'Display service is not running. Please start the display service or enable "Start Service" option.',
'service_status': service_status
}), 400
service_result = None
if start_service:
service_result = _ensure_display_service_running()
@@ -1269,6 +1288,9 @@ def start_on_demand_display():
'message': 'Failed to start display service. Please check service logs or start it manually.',
'service_result': service_result
}), 500
# Service was restarted (or started fresh) with on-demand request in cache
# The display controller will read the request during initialization or when it polls
response_data = {
'request_id': request_id,
@@ -1293,6 +1315,8 @@ def stop_on_demand_display():
data = request.get_json(silent=True) or {}
stop_service = data.get('stop_service', False)
# Set the stop request in cache FIRST
# The display controller will poll this and restart without the on-demand filter
cache = _ensure_cache_manager()
request_id = data.get('request_id') or str(uuid.uuid4())
request_payload = {
@@ -1301,7 +1325,10 @@ def stop_on_demand_display():
'timestamp': time.time()
}
cache.set('display_on_demand_request', request_payload)
# Note: The display controller's _clear_on_demand() will handle the restart
# to restore normal operation with all plugins
service_result = None
if stop_service:
service_result = _stop_display_service()
@@ -2769,9 +2796,9 @@ def get_github_auth_status():
try:
if not api_v3.plugin_store_manager:
return jsonify({'status': 'error', 'message': 'Plugin store manager not initialized'}), 500
token = api_v3.plugin_store_manager.github_token
# Check if GitHub token is configured
if not token or len(token) == 0:
return jsonify({
@@ -2784,10 +2811,10 @@ def get_github_auth_status():
'error': None
}
})
# Validate the token
is_valid, error_message = api_v3.plugin_store_manager._validate_github_token(token)
if is_valid:
return jsonify({
'status': 'success',
@@ -2946,17 +2973,17 @@ def _get_schema_property(schema, key_path):
def _is_field_required(key_path, schema):
"""
Check if a field is required according to the schema.
Args:
key_path: Dot-separated path like "mqtt.username"
schema: The JSON schema dict
Returns:
True if field is required, False otherwise
"""
if not schema or 'properties' not in schema:
return False
parts = key_path.split('.')
if len(parts) == 1:
# Top-level field
@@ -2966,12 +2993,12 @@ def _is_field_required(key_path, schema):
# Nested field - navigate to parent object
parent_path = '.'.join(parts[:-1])
field_name = parts[-1]
# Get parent property
parent_prop = _get_schema_property(schema, parent_path)
if not parent_prop or 'properties' not in parent_prop:
return False
# Check if field is required in parent
required = parent_prop.get('required', [])
return field_name in required
@@ -3116,7 +3143,7 @@ def _set_nested_value(config, key_path, value):
# Skip setting if value is the sentinel
if value is _SKIP_FIELD:
return
parts = key_path.split('.')
current = config
@@ -3320,7 +3347,7 @@ def save_plugin_config():
# Only set if not skipped
if parsed_value is not _SKIP_FIELD:
_set_nested_value(plugin_config, base_path, parsed_value)
# Process remaining (non-indexed) fields
# Skip any base paths that were processed as indexed arrays
for key, value in form_data.items():
@@ -3340,7 +3367,7 @@ def save_plugin_config():
# Use helper to set nested values correctly (skips if _SKIP_FIELD)
if parsed_value is not _SKIP_FIELD:
_set_nested_value(plugin_config, key, parsed_value)
# Post-process: Fix array fields that might have been incorrectly structured
# This handles cases where array fields are stored as dicts (e.g., from indexed form fields)
def fix_array_structures(config_dict, schema_props, prefix=''):