Changelog¶

All notable changes to GigQ are documented here. See the full CHANGELOG.md on GitHub for links to diffs.

0.5.1 (2026-03-21)¶

Added¶

• retry_delay (seconds) on Job, @task, and CLI: failed jobs with retries remaining can wait before becoming eligible again (MCP server included).

Changed¶

• Documentation: scheduling guide, llms.txt / llms-full.txt, PyPI metadata and keywords, README and docs cleanup.

Fixed¶

• Broken link to the hosted documentation site.

0.5.0 (2026-03-21)¶

Added¶

• parent_results: dependent jobs automatically receive parent job results via a parent_results parameter. Auto-detected from function signature, or controlled via Job(pass_parent_results=True/False/None) and Workflow.add_task(..., pass_parent_results=...).
• examples/data_pipeline.py: sequential pipeline example demonstrating parent_results chaining (generate → transform → format).
• MCP server (mcp_server/): Model Context Protocol server for AI agent integration — submit jobs and workflows, monitor queues, retrieve results. Separate package (gigq-mcp).
• Docs integration page for the MCP server.

Changed¶

• Library no longer configures logging on import. NullHandler is used by default (standard Python library convention). Call setup_logging() explicitly for verbose output. The CLI configures its own logging.
• parallel_tasks.py example updated: summarise now uses parent_results to receive and process parent job outputs.

Fixed¶

• Code formatting (Black) in job_queue.py and tests/job_functions.py.

0.3.0 (2026-03-16)¶

Added¶

• --concurrency N flag for the gigq worker command — one worker process, N threads, each independently claiming and executing jobs from the queue.
• SQLite WAL (Write-Ahead Logging) mode enabled by default on all connections, reducing lock contention under concurrent access.
• Per-thread worker IDs (e.g., worker-abc-0, worker-abc-1) for monitoring concurrent threads.

Changed¶

• Worker class accepts a new concurrency parameter (default: 1, no breaking change).
• Worker.current_job_id is now a thread-safe property backed by threading.local().
• Signal handlers in Worker.start() are only registered when running on the main thread.

0.2.1 (2026-03-13)¶

Added¶

• JobQueue.stats() method to retrieve aggregate job counts by status in a single query.
• gigq stats CLI command to display queue statistics in a table.
• JobQueue.get_result() method to fetch only the deserialized job result for completed jobs.

Changed¶

• Minimum Python version bumped from 3.9 to 3.10.

0.2.0 (2025-03-24)¶

Added¶

• Thread-local SQLite connection management for improved performance.
• New db_utils module for database connection handling.
• close_connections() function to clean up thread resources.

Changed¶

• JobQueue and Worker now reuse connections within each thread.

0.1.1 (2025-03-22)¶

Fixed¶

• Fixed project shields in README.
• Implemented comprehensive test coverage.

0.1.0 (2025-03-15)¶

Initial release of GigQ.

Added¶

• Core job queue functionality with SQLite backend.
• Job definition with parameters, priorities, and dependencies.
• Worker implementation for job processing.
• Workflow support for dependent jobs.
• Command-line interface for job and worker management.
• Automatic retry and timeout handling.
• Documentation with MkDocs.