This commit introduces a new SqliteMemoryService that provides persistent
storage for agent memory using SQLite with async operations via aiosqlite.
Features:
- Persistent storage with SQLite database
- Async operations using aiosqlite for non-blocking database access
- Keyword-based search matching InMemoryMemoryService behavior
- User isolation by app_name and user_id
- Automatic database initialization with optimized indexes
- Additional utility methods: clear_memory() and get_memory_stats()
- Duplicate entry prevention with UNIQUE constraints
- Content filtering for empty or invalid events
Implementation follows existing ADK patterns:
- Extends BaseMemoryService with @OverRide decorators
- Similar interface to InMemoryMemoryService and VertexAi services
- Comprehensive test coverage with 24 test cases
- Code formatted with project standards (isort + pyink)
Dependencies:
- Added aiosqlite>=0.20.0 dependency to pyproject.toml
- Updated memory module __init__.py to export SqliteMemoryService
Testing:
- Complete test suite in tests/unittests/memory/test_sqlite_memory_service.py
- All 24 tests passing covering initialization, CRUD operations, search,
 user isolation, memory management, and edge cases

@gemini-code-assist gemini-code-assist bot left a comment

@gemini-code-assist gemini-code-assist bot left a comment

Address reviewer feedback with significant performance and robustness improvements:
**Performance Optimizations:**
- Implement SQLite FTS5 (Full-Text Search) for efficient database-level search
- Replace Python-based keyword matching with native FTS5 MATCH queries
- Add lazy one-time database initialization with asyncio.Lock
- Remove redundant _init_db() calls from every method
**Robustness Improvements:**
- Replace generic Exception handling with specific aiosqlite.Error
- Add proper error handling during database initialization
- Improve query preparation with FTS5-specific escaping
- Add database triggers to keep FTS5 table synchronized
**New Features:**
- FTS5 virtual table for scalable full-text search
- Query preparation function with special character handling
- Enhanced error recovery and graceful degradation
**Testing:**
- Add 8 new test cases covering FTS5 functionality
- Test lazy initialization behavior
- Test FTS5 query preparation and edge cases
- Test comprehensive error handling scenarios
- All 32 tests passing with improved coverage
This addresses the performance bottleneck in search_memory() and ensures
the service scales efficiently with large datasets while maintaining
reliability through proper error handling.

@gemini-code-assist gemini-code-assist bot left a comment

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

- Move GITHUB_TOKEN validation from module import time to function call time
- This prevents ValueError during module imports when GITHUB_TOKEN is not yet set
- Update utils.py to use lazy token loading via get_github_token() function
- Apply fix to both adk_pr_triaging_agent and adk_triaging_agent
- Fixes CI/CD issue where triaging bots fail during workflow initialization

@hangfei hangfei left a comment

This reverts commit 47eb886.

- Fix line length formatting for test function name
- Ensure compliance with project style guidelines
- All 34 tests still passing after formatting
Addresses reviewer feedback for PR google#2768 

- Add explicit Dict and Any imports from typing module
- Replace dict return type with Dict[str, Any] for Python 3.9 compatibility
- Ensures compatibility across Python 3.9-3.13 versions
- Fixes CI test failures on Python 3.9

- Add development-only Python 3.9 compatibility testing scripts
- Prevents accidental inclusion of temporary testing files
- Files: check_py39_compatibility.py, test_runtime_compatibility.py,
 test_py39_docker.py, test_simple_py39.py, test_core_py39.py

- Changed from creating asyncio.Lock() in __init__ to lazy initialization
- Prevents 'RuntimeError: There is no current event loop' in Python 3.9
- Lock is now created only when needed in _ensure_initialized()
- Maintains thread safety and proper initialization semantics
- Fixes all SQLite memory service unit test failures in CI
Tested with Docker across Python 3.9-3.12, all versions pass.

- Added test_sqlite_docker.py and test_real_sqlite_docker.py
- These are development-only testing files for validating the Python 3.9 fix
- Keep the repo clean by excluding temporary testing scripts

...API operation parser
- Updated OperationParser to use 'object' as default return type instead of 'Any'
- This provides better type consistency for OpenAPI operations
- Improves schema validation and type inference

fix: Change default return schema type from 'Any' to 'object' in OpenAPI operation parser

...API operation parser