Deepgram Api

汇总 Deepgram 语音转写、文字转语音、语音智能体与音频智能相关 API 参考材料，包含 api、docs、recipes、starters、setup-mcp 五个子技能，提供 REST 与 WebSocket 接口、模型、鉴权、自托管说明、多语言示例片段、起步应用以及 MCP 服务器配置步骤，用于构建或调试 Deepgram 集成。

sophia-amadeus

热度 15.61

最近更新 2026-07-22 13:45:58

分类：开发依赖：第三方 API / MCP 服务应用场景：API 集成 / 调试排错 / 模板 / 研究

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add deepgram-api

Metadata

name: deepgram-api
description: Deepgram speech-to-text, text-to-speech, voice agents, and audio intelligence — API reference, docs navigator, recipes, starters, and MCP setup. Triggered when working with Deepgram APIs, transcription, TTS, or voice agent integrations.

Deepgram API Skills

This workspace skill provides structured Deepgram API reference material pulled directly from the official deepgram/skills repository.

Included sub-skills

Sub-skill	Purpose
`api/`	Full REST + WebSocket API reference — Nova STT, Flux STT, TTS, Voice Agent, Read intelligence, models, projects, auth, self-hosted
`docs/`	Topic-by-topic documentation navigator with pointers to official docs
`recipes/`	Minimal runnable feature snippets per language
`starters/`	Runnable starter apps (framework × feature matrix)
`setup-mcp/`	How to install and configure the Deepgram MCP server

Quick reference

Base servers:

REST & STT/TTS WebSocket: https://api.deepgram.com
Voice Agent WebSocket: https://agent.deepgram.com

Primary models:

Nova (/v1/listen) — general transcription, rich feature set
Flux (/v2/listen) — conversational audio, built-in turn detection

STT WebSocket common mistakes (from the API skill):

Send KeepAlive as text frame, not binary
Never send empty byte payloads
encoding must match actual audio format

When to use this

Building or debugging Deepgram integrations
Looking up endpoint params, error codes, message formats
Designing voice agent or transcription pipelines
Setting up the Deepgram MCP server for AI tool access