Skip to content

Release Notes — v1.1.0

Released: 2026-05-19 Type: Minor — Backwards-Compatible Feature Release

Summary

v1.1.0 brings EcoTrace into the enterprise observability ecosystem. This release focuses on two production-grade integrations that allow organizations to route carbon metrics directly into their existing monitoring infrastructure and track emissions from their full Django and Celery application stack.

New Features

1. OpenTelemetry (OTel) Exporter

EcoTrace can now push carbon metrics directly into any OpenTelemetry-compatible observability platform — Grafana, Datadog, New Relic, Prometheus.

from ecotrace import EcoTrace
from ecotrace.exporters import OTelExporter

eco = EcoTrace(region_code="DE")
OTelExporter(ecotrace_instance=eco)  # auto-registers

A counter metric named ecotrace.carbon.emitted (unit: g) is created with two attributes per measurement:

  • ecotrace.function — the name of the tracked function or block
  • ecotrace.region — the grid intensity region code

Exporter dispatch is fully non-blocking: measurements are submitted to a 2-worker ThreadPoolExecutor so no slow telemetry backend can delay user code.

Install: pip install ecotrace[otel]

2. Django Middleware

Full WSGI and ASGI support for tracking carbon per HTTP request in Django applications.

# settings.py
MIDDLEWARE = [
    "ecotrace.middleware.django.EcoTraceMiddleware",
    ...
]

# Optional settings
ECOTRACE_LOG_CSV = True   # log each request to ecotrace_log.csv

Every response receives two injected headers: - X-Eco-Carbon-Emitted — carbon footprint in gCO2 - X-Eco-Duration — request duration in seconds

The middleware never interferes with 4xx/5xx responses — exceptions propagate naturally.

Install: pip install ecotrace[web]

3. Celery Plugin

Track carbon emissions per background task with automatic signal integration.

from ecotrace import EcoTrace
from ecotrace.plugins.celery import EcoTraceCelery

eco = EcoTrace(region_code="US")
EcoTraceCelery(ecotrace_instance=eco, log_to_csv=True)

Connects to four Celery worker signals:

Signal Purpose
task_prerun Start CPU monitor, stamp start time
task_postrun Finalize measurement, log emissions
task_retry Finalize current attempt independently (prevents double-counting)
task_revoked Clean up state on kill/revoke (prevents memory leak)

Install: pip install ecotrace[celery]

Core Engine Changes

  • EcoTrace.add_exporter(exporter) — New public API to register any object implementing .export(carbon_emitted, func_name, duration, region).
  • Thread-safe exporter dispatch — Exporters run on a dedicated 2-worker ThreadPoolExecutor to prevent blocking.
  • Graceful shutdown — _exporter_pool.shutdown(wait=True) is called on process exit to flush all in-flight export tasks before the process terminates.

Installation

# Core only
pip install ecotrace==1.1.0

# With OTel support
pip install "ecotrace[otel]==1.1.0"

# With Django + Celery
pip install "ecotrace[web,celery]==1.1.0"

# Everything
pip install "ecotrace[all]==1.1.0"

Bug Fixes

  • Fixed GPU TDP resolution to use nvmlDeviceGetPowerManagementLimit (runtime limit) instead of nvmlDeviceGetPowerManagementDefaultLimit (factory cap), which caused over-estimation on power-capped GPUs (e.g. throttled data centre A100s).

What's Next — v1.2.0

  • Sampling Mode — @eco.track(sample_rate=0.1) for statistical measurement at high call frequencies, removing the overhead barrier for adoption in high-throughput production services.