Voice and Text Intelligence - Deepgram
API capability pack for bounded Deepgram REST API operations against https://api.deepgram.com only. Enroll a Deepgram API connection named deepgram that injects the Authorization header, commonly Authorization: Token <API_KEY>; no credential is embedded in this manifest. The pack transcribes and analyzes one explicit pre-recorded media URL, analyzes explicit text, reads public and project model catalogs, reads projects, usage fields, usage breakdowns, and project balances for meeting notes, call analytics, media intelligence, product research, compliance review, usage monitoring, and model inventory workflows. Transcription and text analysis are synchronous JSON read-tier inference calls that consume Deepgram quota but do not create durable customer resources through this pack. The pack intentionally excludes WebSocket streaming, turn-based audio, voice-agent configuration and variables, token generation, binary file upload, binary text-to-speech output, callback and callback_method parameters, arbitrary extra metadata, project/key/member/invite/request mutations or deletes, self-hosted distribution credentials, billing purchase exports beyond balances, regional/custom host switching, arbitrary Deepgram API passthrough, dynamic outbound callbacks, and any operation outside api.deepgram.com.
What this pack installs
- Deepgram audio intelligence brief Recipe
- Deepgram text intelligence brief Recipe
- Deepgram model catalog brief Recipe
- Deepgram project usage digest Recipe
- deepgram Ingredient
Trust & control
What installing this whole pack would let it do. Recued grants these permissions at install — review them there before approving.