Remove FakeMessage wrapper class and implement consistent message editing
for web app code login flow. User-submitted code is now deleted immediately
and instructions message transforms in-place to show menu, matching the
email login UX pattern.
Changes:
- Remove FakeMessage class from start_command handler
- Use direct edit_message_text for all message updates
- Delete user's code message for cleaner chat interface
- Preserve web_login_message_id in context throughout flow
- Apply same pattern to handle_text_message (8-char code handler)
- Ensure single editable message from instructions to menu
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Consolidate login flow to use single message that edits in place
- Auto-delete all user messages (email, code, password, /start command)
- Remove all emojis from bot messages for cleaner interface
- Fix "Retrimite Cod" bug - buttons now persist after resending code
- Replace "Oracle" with "ROA" in all user-facing messages
- Add clear instructions for each step (email input, code input, password)
- Implement message tracking with context.user_data['login_message_id']
- Clean chat history - only menu message remains visible
Files modified:
- app/bot/email_handlers.py: Complete refactor of email login flow
- app/bot/handlers.py: Update /start command to delete user message and edit existing message
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Cache System (Backend):
- Implemented two-tier hybrid cache: L1 (in-memory) + L2 (SQLite)
- L1 cache: Fast dictionary-based with 5-minute TTL for hot data
- L2 cache: Persistent SQLite with 1-hour TTL for warm data
- Cache decorator with automatic tier management and fallback
- Cache key generation with per-user isolation
- Event monitoring system for cache statistics
- Cache benchmarking utilities for performance testing
- Added cache management endpoints: /api/cache/stats, /api/cache/clear, /api/cache/benchmark
- Cache configuration via environment variables (CACHE_ENABLED, CACHE_L1_TTL, etc.)
Backend Services:
- Updated dashboard_service to use @cached decorator with request context
- Added cache support to invoice_service and treasury_service
- Integrated cache manager into main.py with lifespan events
- Added Request parameter to service methods for cache metadata
Frontend Enhancements:
- New CacheStatsView.vue for real-time cache monitoring dashboard
- Cache store (cacheStore.js) for state management
- Updated router to include /cache-stats route
- Navigation updates in DashboardHeader and HamburgerMenu
- Cache stats accessible from main navigation
Telegram Bot Improvements:
- Enhanced formatters with YTD comparison data
- Improved menu navigation and button layout
- Better error handling and user feedback
- Bot startup improvements with graceful shutdown
Auth & Middleware:
- Enhanced middleware with cache metadata injection
- Improved request state handling for cache source tracking
Development:
- Updated start-dev.sh with better error handling
- Added TELEGRAM_EMAIL_AUTH_PLAN.md documentation
- Updated requirements.txt with aiosqlite for async SQLite
Performance:
- L1 cache provides <1ms response for hot data
- L2 cache provides ~5ms response for warm data
- Database queries only for cold data or cache misses
- Cache hit rates tracked and displayed in real-time
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
Implements cache tier identification in Telegram bot to display data source:
- "db" for database queries
- "cached L1" for in-memory cache hits
- "cached L2" for SQLite cache hits
Backend changes:
- Added cache metadata fields to TrendsResponse and DashboardSummary models
(cache_hit, response_time_ms, cache_source)
- Updated /api/dashboard/summary and /api/dashboard/trends endpoints to
include cache metadata when X-Include-Cache-Metadata header is present
- Cache metadata is extracted from request.state (set by @cached decorator)
Telegram bot changes:
- Updated API client to send X-Include-Cache-Metadata header
- Modified helpers to extract cache_source from backend responses
- Updated handlers to pass cache metadata to formatters
- Performance footer now displays specific cache tier (L1 vs L2)
Fixed Pydantic serialization issue:
- Changed field names from _cache_hit to cache_hit (without underscore)
- Pydantic excludes underscore-prefixed fields from JSON by default
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Add YTD year-over-year comparison table for cash flow evolution
- Extend monthly evolution from 6 to 12 months with dynamic year extraction
- Simplify monthly view to show only Net values aligned with YTD table
- Upgrade client/supplier display from Top 5 to Top 10 with alphabetical sorting
- Remove Refresh and Export buttons from dashboard and evolution views
- Add get_trends() API method for 12-month historical data from backend
- Fix default years to 2025/2024 for accurate YTD calculations
Changes:
- client.py: New get_trends() method calls /api/dashboard/trends endpoint
- helpers.py: Rewrite get_cashflow_evolution_data() to use trends and calculate YTD
- formatters.py: Complete redesign with YTD table and simplified 12-month Net view
- menus.py: Alphabetical sorting for clients/suppliers, removed refresh buttons
- handlers.py: Disabled refresh/export buttons on dashboard and evolution views
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
**Interface improvements:**
- Add persistent Login/Logout buttons to main menu
- Help button now updates text above menu (not separate message)
- Expired token automatically transforms menu to Login state
- Consolidate linking success messages (single welcome + menu)
**Fix callback_data length limits (Telegram 64-byte limit):**
- Truncate client/supplier names to 40 chars in callback_data
- Use full names for API calls but truncated for buttons
- Fix pagination buttons for long partner names (30 chars limit)
- Search entities by prefix match to handle truncated names
**Treasury improvements:**
- Show ALL bank/cash accounts (removed 5-item limit)
- Remove unnecessary Refresh/Export buttons from treasury views
- Handle "Message is not modified" error gracefully
**Bug fixes:**
- Fix Markdown parsing errors (replace <cod> with `CODUL_TAU`)
- Fix silent errors when token expires (show user-friendly message)
- Fix Button_data_invalid errors on pagination and details
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move load_dotenv() call before all app.* imports in main.py to fix the
import-time environment variable loading issue.
Problem:
- Previously, load_dotenv() was called AFTER importing app.* modules
- When app.api.client was imported, it read BACKEND_URL at import time
- At that moment, .env was not yet loaded, so it used the hardcoded default
- This caused the bot to connect to wrong backend port even when .env was
correctly configured
Solution:
- Move load_dotenv() to line 20-21, immediately after standard library imports
- This ensures .env is loaded BEFORE any app.* modules are imported
- Now all modules see the correct environment variables from .env file
- Also updated BACKEND_URL default from 8001 to 8000 for consistency
Flow now:
1. Import standard libraries (os, Path, dotenv, etc.)
2. Load .env file (line 20-21) ✅
3. Import app.* modules (which can now read env vars correctly) ✅
4. Import telegram-python-bot and other dependencies
5. Start application
This follows Python best practices for environment variable loading and
ensures reliable configuration loading in Windows Service deployments.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Change hardcoded default backend URL from development port (8001) to
production port (8000) in Telegram bot API client.
This fixes the issue where Telegram bot would try to connect to wrong port
when BACKEND_URL environment variable is not properly loaded from .env file,
causing "Cannot connect to backend" errors during account linking.
Root cause: When .env file is not loaded correctly by Windows Service,
the code falls back to the hardcoded default value which was incorrectly
set to the development port 8001 instead of production port 8000.
Changes:
- reports-app/telegram-bot/app/api/client.py: Change default from 8001 to 8000
- Add comment explaining this is for production deployment
This ensures the bot connects to the correct backend port even if .env
configuration has issues during service startup on Windows Server.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Fix AttributeError crash when backend is unreachable during account linking.
Previously, when telegram bot couldn't connect to backend, the error handler
tried to access e.response.status_code on a ConnectError exception which
doesn't have a response attribute.
Changes to reports-app/telegram-bot/app/api/client.py:
- Import ConnectError from httpx
- Add separate exception handler for ConnectError before HTTPError handler
- Log clear error message indicating backend connectivity issue
- Return None gracefully instead of crashing with AttributeError
Changes to deployment/windows/docs/TELEGRAM_BOT_TROUBLESHOOTING.md:
- Add new section "Problem: Cannot connect to backend / Connection Errors"
- Add diagnostic steps for backend service verification
- Add checklist for BACKEND_URL configuration (http://localhost:8000)
- Add Issue 5: Backend Service Not Running
- Add Issue 6: Wrong Backend URL in Telegram Bot
- Include PowerShell commands for Windows Server troubleshooting
This fix ensures the Telegram bot provides clear error messages when backend
is unavailable instead of crashing, making debugging easier for production
deployments.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit fixes overly broad .gitignore patterns that were excluding
important source code files from version control. Previously, wildcard
patterns like *auth*, *token*, *secret*, *connection*, and *credential*
were excluding ALL files containing these words, including critical
application code.
Changes:
- Updated .gitignore with specific patterns for sensitive config files
(*.json, *.txt, *.yml, *.yaml extensions only)
- Removed broad wildcards that excluded source code files
Added missing source files:
- shared/auth/ (9 files): Complete authentication system
- JWT handler, middleware, auth service, models, routes
- reports-app/backend/app/routers/auth.py: Authentication API router
- reports-app/backend/app/auth_middleware_wrapper.py: Middleware wrapper
- reports-app/frontend/src/stores/auth.js: Vue.js auth store
- reports-app/frontend/tests/: E2E tests and fixtures for auth
- reports-app/telegram-bot/app/auth/: Telegram auth linking module
- deployment/windows/scripts/Setup-ClaudeAuth.ps1: Windows deployment script
- security/secrets_scanner.py: Security scanning utility
These files are essential for the application to function and should
have been included in the initial commit.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
