Add internal and public telemetry

[davidberenstein1957]

Is your feature request related to a problem? Please describe.
As OSS developers, it is not clear how people are using the library.

Describe the solution you'd like
🏷️ Three-Tier System

Tier	Env Variable	Data Visibility	First-Run Prompt
Disabled	CODECARBON_TELEMETRY=disabled	None	"Telemetry disabled"
Minimal	CODECARBON_TELEMETRY=minimal	Private (CodeCarbon team only)	"Basic environment info only"
Extensive	CODECARBON_TELEMETRY=extensive	Public (leaderboard)	"Full data + emissions shared publicly"

|

---

📊 Complete Metrics by Tier
Tier 0: Disabled
• No telemetry collected

---

Tier 1: Minimal (Environment & Hardware)
Location

Field	Description
os	Platform system (e.g., Linux-5.10.0-x86_64)
country_name	Detected country
country_iso_code	ISO code (e.g., USA, FRA)
region	Cloud region or geographic region
cloud_provider	AWS, GCP, Azure, or empty (on-prem)
cloud_region	e.g., us-east-1
longitude / latitude	Rounded to 1 decimal (~10km precision)

Device Information

Field	Description
cpu_count	Number of logical CPUs
cpu_physical_count	Number of physical cores
cpu_model	CPU model name
cpu_architecture	x86_64, arm64
gpu_count	Number of GPUs
gpu_model	GPU model name(s)
gpu_driver_version	NVIDIA driver version
gpu_memory_total_gb	Total GPU memory
ram_total_size_gb	Total RAM
cuda_version	CUDA version (if available)
cudnn_version	cuDNN version (if available)

Python Environment

Field	Description
python_version	e.g., 3.11.5
python_implementation	CPython, PyPy, Jython
python_executable_hash	SHA256 hash of sys.executable (privacy-safe)
python_env_type	conda, venv, virtualenv, uv, pip, system
codecarbon_version	Current version
codecarbon_install_method	pip, conda, uv, editable

---

Tier 2: Extensive (Full + Public)
All of Tier 1 + Emissions Data

Field	Description
total_emissions_kg	Total CO2 emissions in kg
emissions_rate_kg_per_sec	Emissions rate
energy_consumed_kwh	Total energy
cpu_energy_kwh	CPU energy
gpu_energy_kwh	GPU energy
ram_energy_kwh	RAM energy
duration_seconds	Total tracking duration
cpu_utilization_avg	Average CPU %
gpu_utilization_avg	Average GPU %
ram_utilization_avg	Average RAM %

Usage Patterns

Field	Description
tracking_mode	machine, process, or hybrid
api_mode	offline, online, or both
output_methods	file, api, logger, prometheus, logfire
hardware_tracked	cpu, gpu, ram (combinations)
task_tracking_used	bool
decorator_vs_context	decorator, context_manager, or direct
measure_power_interval_secs	Configured measurement interval

Error & Performance Metrics

Field	Description
hardware_detection_success	bool
rapl_available	bool
gpu_detection_method	nvidia-smi, pytorch, rocm, none
first_measurement_time_ms	Time to first measurement
tracking_overhead_percent	CPU overhead
errors_encountered	List of error types
warning_count	Number of warnings logged

Environment Context

Field	Description
ide_used	vscode, pycharm, jupyter, none
notebook_environment	jupyter, colab, kaggle, none
ci_environment	github-actions, gitlab, jenkins, none
python_package_manager	pip, poetry, uv, conda, pipenv
framework_detected	pytorch, tensorflow, JAX, none

ML Ecosystem

Field	Description
has_torch	bool (+ version if available)
has_transformers	bool (+ version)
has_diffusers	bool (+ version)
has_tensorflow	bool (+ version)
has_keras	bool (+ version)
has_pytorch_lightning	bool (+ version)
has_fastai	bool (+ version)
ml_framework_primary	Which ML framework is primary

Container & Deployment

Field	Description
container_runtime	docker, podman, kubernetes, none
in_container	bool
host_machine_hash	Anonymous machine fingerprint (opt-in only)
Describe alternatives you've considered
Not tracking anything

Additional context
Add any other context or screenshots about the feature request here.