Skip to content

Release Notes — v1.4.1

Released: 2026-07-18
Type: Patch Release — 7 bug fixes, zero breaking changes

Summary

v1.4.1 is a patch release addressing multiple edge-case bugs, linter/static analysis warnings, and dynamic versioning issues identified in v1.4.0. It improves exception safety in code blocks, normalizes GPU reports, and resolves unresolved import warnings in environments without optional web dependencies installed.

Resolved Issues & Bug Fixes

1. Exception Safety in track_block()

Restructured the track_block() context manager using a try/finally block. This ensures that session duration and carbon metrics are cleanly calculated and logged even if the wrapped code blocks raise user exceptions, aligning it with the robust exception-handling design of measure().

2. GPU Chart Data Normalization

Normalized the _gpu_samples structure inside the core engine's PDF generator (generate_pdf_report). The core continuous hardware monitor tracks metrics using 3-tuples (timestamp, utilization, power). The PDF reporting module now filters and maps these to 2-tuples (timestamp, utilization) before rendering to prevent visual plotting and parsing errors.

3. Django Middleware Renaming

Renamed the middleware class defined in middleware/django.py from EcoTraceMiddleware to EcoTraceDjangoMiddleware to perfectly reflect the migration guidelines documented in v1.1.2. A backward-compatible alias EcoTraceMiddleware is preserved at module scope to prevent integration breakage.

4. Dynamic Dashboard Versioning

Replaced the hardcoded version in the live dashboard HTML footer. It now dynamically injects the package's current version (e.g. v1.4.1) retrieved directly from metadata at request resolution.

5. Format-Aware Export Filenames

The CLI ecotrace export command now determines the default output filename according to the selected format. Specifying --csv automatically defaults to ecotrace_export.csv rather than confusingly writing CSV data to ecotrace_report.json.

6. Dynamic Web Imports (No Static Warnings)

Refactored middleware/django.py to lazily and dynamically load optional dependencies (django and asgiref) using importlib. This removes critical 'Unresolved Import' red alerts inside IDEs and static analyzers for developers using EcoTrace without standard web framework libraries in their environment.

7. Corrected GPU Energy Accumulation

Initialized the total_energy_kwh property inside the main engine constructor. This permits cumulative GPU/ML tracking contexts to correctly increment the total session energy metrics across consecutive training run boundaries.

Upgrade Instructions

This is a non-breaking patch release. Upgrade via pip:

pip install --upgrade ecotrace