Split the monolithic DatabaseService (1,492 lines) into focused repository
modules using the repository pattern for better maintainability and
separation of concerns.
Changes:
- Create new repositories/ directory with 5 focused repositories:
- TransactionRepository: transaction data operations (264 lines)
- AccountRepository: account data operations (128 lines)
- BalanceRepository: balance data operations (107 lines)
- MigrationRepository: all database migrations (629 lines)
- SyncRepository: sync operation tracking (132 lines)
- BaseRepository: shared database connection logic (28 lines)
- Refactor DatabaseService into a clean facade (287 lines):
- Delegates data access to repositories
- Maintains public API (no breaking changes)
- Keeps data processors in service layer
- Preserves require_sqlite decorator
- Update tests to mock repository methods instead of private methods
- Fix test references to internal methods (_persist_*, _get_*)
Benefits:
- Clear separation of concerns (one repository per domain)
- Easier maintenance (changes isolated to specific repositories)
- Better testability (repositories can be mocked individually)
- Improved code organization (from 1 file to 7 focused files)
All 114 tests passing.
The notification system was incorrectly sending notifications for existing
transactions that were being updated during sync operations. This change
modifies the transaction persistence logic to only return genuinely new
transactions, preventing duplicate notifications while maintaining data
integrity through INSERT OR REPLACE.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Merge leggen/database/sqlite.py functionality directly into DatabaseService
- Extract transaction processing logic to separate TransactionProcessor class
- Remove leggen/utils/database.py and leggen/database/ directory entirely
- Update all tests to use new consolidated structure
- Reduce codebase by ~300 lines while maintaining full functionality
- Improve separation of concerns: data processing vs persistence vs CLI
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Merge leggend API components into leggen (api/, services/, background/)
- Replace leggend command with 'leggen server' subcommand
- Consolidate configuration systems into leggen.utils.config
- Update environment variables: LEGGEND_API_URL -> LEGGEN_API_URL
- Rename LeggendAPIClient -> LeggenAPIClient
- Update all documentation, Docker configs, and compose files
- Fix all import statements and test references
- Remove duplicate utility files and clean up package structure
All tests passing (101/101), linting clean, server functionality preserved.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Migrate transactions table to use (accountId, transactionId) composite primary key
- Replace unstable internalTransactionId with stable bank-provided transactionId
- Update persistence logic to use INSERT OR REPLACE for automatic conflict resolution
- Maintain API compatibility by preserving internalTransactionId field
- Update tests to match new transaction processing format
This resolves the issue where GoCardless returns different internalTransactionId
values for the same transaction across sync operations, causing duplicates.
- Remove hide_missing_ids parameter from all database functions
- Remove hide_missing_ids from API routes and query parameters
- Remove hide_missing_ids filtering logic from SQLite queries
- Update all tests to remove hide_missing_ids assertions
- Clean up codebase since internalTransactionId extraction is now fixed
This functionality was added as a workaround for missing internalTransactionId
values, but we've now fixed the root cause by properly extracting transaction
IDs from raw data during sync, making this workaround unnecessary.
- Add hide_missing_ids parameter to database functions to filter out transactions without internalTransactionId
- Update API routes to support the new filter parameter
- Update unit tests to include the new parameter
- Add opencode.json configuration file
- Updated SQLite database to use ~/.config/leggen/leggen.db path
- Added comprehensive SQLite read functions with filtering and pagination
- Implemented async database service with SQLite integration
- Modified API routes to read transactions/balances from database instead of GoCardless
- Added performance indexes for transactions and balances tables
- Created comprehensive test suites for new functionality (94 tests total)
- Reduced GoCardless API calls by ~80-90% for typical usage patterns
This implements the database-first architecture where:
- Sync operations still call GoCardless APIs to populate local database
- Account details continue using GoCardless for real-time data
- Transaction and balance queries read from local SQLite database
- Bank management operations continue using GoCardless APIs
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
