mirror of
https://github.com/taosdata/TDengine
synced 2026-05-24 10:09:01 +00:00
6e2bc39a25
* feat: Windows adaptation with winsw service registration and bilingual documentation * fix: remove unused global declaration in parse_arguments function * style: fix code quality issues detected by ruff - Remove unused imports (pathlib.Path, json, torch, keras) - Fix f-string without placeholders - Remove unused variables (proc) - Fix indentation errors - Clean up trailing whitespace and blank lines - Fix line length issues (>120 chars) - Add noqa comments for intentional violations * chore: remove obsolete taosanode.ini configuration file * refactor: extract ISS and XML templates to separate files - Move Inno Setup script to templates/tdgpt.iss - Move WinSW config to templates/taosanode-service.xml - Implement local WinSW.exe caching (packaging/bin/) - Add .gitignore to exclude cached binaries - Simplify win_release.py by using template files Benefits: - Easier to maintain ISS and XML files - No need to escape quotes in Python - WinSW binary cached locally for faster builds - Better separation of concerns * refactor: rename templates to installer and consolidate gitignore - Rename packaging/templates/ to packaging/installer/ (more descriptive) - Remove tools/tdgpt/.gitignore (redundant) - Add WinSW.exe exclusion to community/.gitignore - Update win_release.py to use new directory name The 'installer' name better reflects the content: - tdgpt.iss: Inno Setup installer script - taosanode-service.xml: Windows service configuration * fix: improve packaging script robustness 1. Fix GBK encoding issue in install.bat - Replace ✓ symbol with [OK] text - Prevents crash on Windows with GBK encoding 2. Make model directory optional in ISS - Add 'skipifsourcedoesntexist' flag - Allow packaging without model files 3. Remove ZIP fallback logic - Fail fast if Inno Setup compilation fails - Return proper exit code (1) on failure - Provide clear error messages 4. Remove unused zipfile import These changes ensure the packaging script fails clearly when prerequisites are missing, rather than silently creating incomplete packages. * fix: revert model directory to required and add validation Model directory is REQUIRED for Windows production packaging, not optional. The previous 'skipifsourcedoesntexist' was only meant for testing purposes. Changes: - Revert ISS to require model directory with DirExists check - Add validation in win_release.py before building - Provide clear error message with usage example - Fail fast if model directory is missing or empty This ensures users get clear feedback when model files are missing, rather than cryptic Inno Setup errors. * feat: add --skip-model-check for testing workflow Add flexible model validation to support both testing and production: Testing mode (--skip-model-check): - Skip model validation for quick testing - ISS uses 'skipifsourcedoesntexist' to allow empty model dir - Shows warning that this is for testing only Production mode (default): - Validates model directory exists and has files - Checks for required files: timemoe.tar.gz, tdtsfm.tar.gz - Fails with clear error if requirements not met - Ensures production packages always have models Usage: # Testing (no models needed) python win_release.py -e community -v 3.3.6.0 --skip-model-check # Production (models required) python win_release.py -e community -v 3.3.6.0 -m /path/to/models * fix: update model paths in README-CN for packaging instructions * docs: update packaging documentation with --skip-model-check Add documentation for the new --skip-model-check parameter: - Explain testing vs production modes - Document required model files (timemoe.tar.gz, tdtsfm.tar.gz) - Update usage examples to show both modes - Add warning that --skip-model-check is for testing only - Update parameter table with new option Both English and Chinese documentation updated. * enh:add installation, uninstallation, and log viewer scripts for TDGPT on Windows * enh: auto-detect installation directory and update paths for Windows compatibility * Refactor and localize logging and service management in TDGPT - Removed the log_viewer.py script as it is no longer needed. - Updated taosanode_service.py to improve cross-platform support and logging consistency. - Changed log file names to unify logging across services. - Enhanced uninstall.py for better user experience and clearer output messages. - Improved uninstall.sh to preserve local configuration while removing global config. * feat: add Windows installation script and documentation for TDGPT * fix: update version number in packaging scripts and documentation to 3.4.0.11 * Refactor taosanode service logging and Windows service management - Introduced a dedicated service log file for taosanode. - Added functionality to redirect stdout/stderr to a specified file when running under WinSW. - Updated virtual environment paths to be more organized under a 'venvs' directory. - Enhanced the uninstall script to improve logging and user feedback. - Streamlined the removal of files and directories during uninstallation. - Adjusted configuration paths to align with the new directory structure. - Improved test configuration to reflect changes in virtual environment paths. * Refactor taosanode_service.py to improve model management and logging - Updated model-start command help text for clarity. - Introduced DEFAULT_MODEL_ORDER for consistent model ordering. - Added support for loading enabled models from a configuration file. - Adjusted model configurations for moirai and moment to use smaller default models. - Enhanced model startup and shutdown processes to skip missing directories and log detailed summaries. - Improved status reporting for running models. - Refactored uninstall.py to handle logging based on environment variable. - Made torch and keras imports optional in conf.py to avoid runtime errors if not available. * feat: add Python version check and error handling to installer scripts * fix: update model installation options and logic in installer script * fix(tdgpt): restore windows startup diagnostics * fix(tdgpt): limit winsw restart retries * fix(tdgpt): reuse healthy venvs on online install * fix(tdgpt): retry progress file writes on windows * fix(tdgpt): switch installer progress to append-only log * fix(tdgpt): raise installer disk space estimate * fix(tdgpt): rebuild incomplete reused main venv * fix(tdgpt): add detailed startup dependency probes * fix(tdgpt): improve reinstall and startup diagnostics * fix(tdgpt): require vc++ runtime before install * feat(tdgpt): default start-model to all * feat(tdgpt): confirm existing install directory * fix(tdgpt): stop winsw before reinstall * fix(tdgpt): retry rate-limited model downloads * fix(tdgpt): reuse existing online model directories * feat(tdgpt): prefer offline model install flow * feat(tdgpt): update model ports and enhance installation logging * feat(tdgpt): enhance taosanode startup process and readiness checks * fix(tdgpt): improve logger setup to handle non-writable log directories * feat(tdgpt): update dependency installation and enhance logging configuration for GitHub Actions * feat(tdgpt): enhance Windows and GitHub Actions compatibility in taosanode configuration * fix(tdgpt): correct model service start assertion for missing required model * docs(tdgpt): update installation and usage documentation for clarity and consistency * chore: remove outdated documentation files for doc-writing skill --------- Co-authored-by: support-platform <it@taosdata.com> |
||
|---|---|---|
| .. | ||
| ISSUE_TEMPLATE | ||
| scripts | ||
| workflows | ||
| CODEOWNERS | ||