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

# Deepgram

> Speech-to-text, text-to-speech, and audio intelligence — transcription, TTS, and text analysis

# Deepgram

Shanone's Deepgram integration gives your agent 12 tools for voice AI: audio transcription, speaker diarization, summarization, text-to-speech, and text intelligence (sentiment, topics, intents, entities).

## Getting Started

<Steps>
  <Step title="Get a Deepgram API Key">
    Sign in to the [Deepgram Console](https://console.deepgram.com/project/api-keys) and create a new API Key for your project.
  </Step>

  <Step title="Add it to Shanone">
    Open **Integrations** in the Shanone dashboard, select Deepgram, and paste in the API Key.
  </Step>

  <Step title="Retry your request">
    Once saved, `shanone_execute_tool` calls for `deepgram_*` tools will succeed.
  </Step>
</Steps>

## Available Tools

Shanone provides **12 tools** for Deepgram, organized into these categories:

<AccordionGroup>
  <Accordion title="Transcription">
    `deepgram_transcribe`, `deepgram_analyze_audio`, `deepgram_detect_entities`, `deepgram_summarize_audio`
  </Accordion>

  <Accordion title="Speech Synthesis">
    `deepgram_text_to_speech`, `deepgram_list_voices`
  </Accordion>

  <Accordion title="Text Intelligence">
    `deepgram_analyze_text`, `deepgram_summarize_text`
  </Accordion>

  <Accordion title="Management">
    `deepgram_list_models`, `deepgram_get_usage`, `deepgram_get_balance`, `deepgram_list_project_models`
  </Accordion>
</AccordionGroup>

## Common Use Cases

<CardGroup cols={2}>
  <Card title="Meeting transcription" icon="mic">
    Transcribe a recorded call with speaker diarization and summarize the key points
  </Card>

  <Card title="Voice interfaces" icon="volume-2">
    Convert generated text responses to speech for a voice-based agent experience
  </Card>

  <Card title="Support ticket triage" icon="inbox">
    Run sentiment and entity extraction on transcribed support calls to prioritize follow-up
  </Card>

  <Card title="Usage monitoring" icon="gauge">
    Track API usage and remaining balance to avoid unexpected billing surprises
  </Card>
</CardGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="deepgram_transcribe fails with a fetch error">
    `audio_url` must be a publicly accessible URL — Deepgram fetches the file server-side and can't reach URLs behind authentication or on a local network.
  </Accordion>

  <Accordion title="deepgram_analyze_audio / deepgram_summarize_audio return unexpectedly generic results">
    These intelligence features rely on `smart_format` and clean transcription first — for noisy audio, try `nova-3` (the most accurate model) before running summarization or entity detection on top.
  </Accordion>

  <Accordion title="deepgram_get_usage or deepgram_get_balance shows unexpected numbers">
    Usage and balance are tracked at the Deepgram project level tied to the API Key, not per-user within Shanone — if multiple integrations share the same Deepgram project, usage will reflect all of them combined.
  </Accordion>
</AccordionGroup>