> ## Documentation Index
> Fetch the complete documentation index at: https://docs.callem.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Sustainability

> Track the carbon footprint and energy consumption of your voice AI operations.

# Sustainability

The Sustainability tab in Analytics provides a real-time view of the environmental impact of your voice AI operations. Monitor CO2 emissions, energy consumption, and the benefits of hosting on low-carbon infrastructure.

## Accessing Sustainability Metrics

Navigate to **Monitor > Analytics** and click the **Sustainability** tab (next to Overview).

## KPIs

The dashboard displays 6 key performance indicators:

| KPI                       | Unit     | Description                                                    |
| ------------------------- | -------- | -------------------------------------------------------------- |
| **Total CO2**             | g / kg   | Total greenhouse gas emissions for the selected period         |
| **CO2 / conversation**    | g        | Average emissions per conversation                             |
| **Energy consumed**       | Wh / kWh | Total energy consumption across all components                 |
| **Energy / conversation** | Wh       | Average energy per conversation                                |
| **Low-carbon energy**     | %        | Share of low-carbon energy in the datacenter's electricity mix |
| **LLM tokens processed**  | count    | Total tokens processed by language models                      |

## How Emissions Are Calculated

Emissions are computed from three components of the voice AI pipeline:

| Component | What it measures                                                                                   |
| --------- | -------------------------------------------------------------------------------------------------- |
| **LLM**   | Token processing (inference) — proportional to the number of tokens generated                      |
| **TTS**   | Text-to-speech synthesis — only uncached characters count; cached responses produce zero emissions |
| **STT**   | Speech-to-text transcription — proportional to call duration                                       |

Energy consumption is calculated per component and then converted to CO2 using the carbon intensity of the datacenter's electricity mix. Callem's infrastructure runs on **low-carbon energy** (92% from nuclear, hydro, wind, and solar).

<Info>
  TTS cache hits produce zero emissions. When a TTS response is served from cache, no GPU compute is needed. The dashboard only counts `uncachedChars`.
</Info>

## Charts

### CO2 Emissions Over Time

A stacked area chart showing daily emissions broken down by component (LLM, TTS, STT). Hover to see exact values.

### Breakdown by Component

A donut chart showing the percentage split between LLM, TTS, and STT emissions.

### CO2 by Agent

A horizontal bar chart ranking agents by total emissions. Useful for identifying which agents consume the most resources.

## CSRD Compliance

The Sustainability dashboard supports **CSRD** (Corporate Sustainability Reporting Directive) reporting requirements:

| ESRS Standard                 | What we provide                                      |
| ----------------------------- | ---------------------------------------------------- |
| **E1-5** (Energy consumption) | Total kWh, kWh per conversation, renewable share     |
| **E1-6** (GHG emissions)      | Total CO2, CO2 per conversation, breakdown by source |

<Tip>
  Per-model and per-provider coefficients will be available in future updates for more granular reporting.
</Tip>