ceven

Deepgram

Transcribes audio files in real time, generates concise summaries from voice recordings, and converts text into natural speech for automated voice workflows.

Try Deepgram in Ceven

Ask Ceven anything
Standard

Why use Ceven?

  1. AI native Deepgram integration

    • Describe the outcome and Ceven picks the right Deepgram calls, fills the parameters, and checks the result.
    • Structured, agent friendly tool schemas so each call runs reliably instead of by guesswork.
    • Rich coverage for reading, writing, and querying your Deepgram data, across all 8 of its actions.

  2. Managed auth

    • Built in OAuth with automatic token refresh and rotation.
    • One place to manage, scope, and revoke Deepgram access.
    • Per user and per environment credentials instead of shared keys.

  3. Agent optimized design

    • Actions are tuned from real success and error rates so reliability climbs over time.
    • Full execution logs so you always know what ran in Deepgram, when, and on whose behalf.
    • The agent pauses and asks when Deepgram is unclear instead of plowing ahead.

  4. Enterprise grade security

    • Fine grained access so you control which agents and people can reach Deepgram.
    • Least privilege by default, read scopes first and only the writes a workflow needs.
    • A full audit trail of every Deepgram action to support review and sign off.

Supported tools

Every action Ceven's agents can run on Deepgram, and when to use it.

Transcribe Audio
Use this when you need to convert a publicly accessible audio URL into a full text transcript.
Summarize Audio
Pull a concise summary from an audio file URL to get the main points without listening to the whole recording.
Detect Topics
Use this to identify the primary themes and subjects discussed within an audio recording.
Text to Speech
Convert a text string into a natural sounding audio file for voice responses.
Get Public Models
Pull a list of all available speech to text models to choose the best one for your language or use case.
Get TTS Models
List all available text to speech voice models to select the right tone and persona.
List Projects
Retrieve a list of all Deepgram projects associated with your API key.
Get Usage Summary
Pull high level metrics for a specific project, including processed duration and billable hours.
List Project Scopes
Retrieve all permission scopes for a project to verify what the agent can access.
Get Model Metadata
Pull specific details on a model version, including whether it is deprecated or current.
Check Audio Accessibility
Verify if a provided audio URL is reachable before sending it for transcription.
Validate API Key
Test the connection to Deepgram to ensure the project key is active and valid.
List Deepgram Projects
Tool to list all deepgram projects. use after authenticating with your api key.
Get Project Usage Summary
Tool to retrieve a summary of usage data for a specified deepgram project. use when you need high level metrics (submitted, processed, billable durations, etc.) optionally filtered by time window, model, accessor, or tag.
Get Public TTS Models
Tool to fetch metadata about all latest public tts voice models. use when you need to list available deepgram tts voices.
Transcribe Pre recorded Audio
Tool to transcribe pre recorded audio files into text. use when you need to convert an audio file at a url to text after confirming the file url is publicly accessible.
Text to Speech (REST)
Tool to convert text into natural sounding speech. use when you need tts audio from text inputs.
Topic Detection
Tool to transcribe audio and detect topics. use when you need to identify the main topics discussed in an audio recording after confirming the file url is publicly accessible.

18 actions · scroll to see them all

Frequently asked questions

Ceven requires that audio files be hosted at a publicly accessible URL. When a workflow triggers a transcription, the agent sends the link to Deepgram, which then streams the audio for processing. We do not host your audio files on our own servers, which reduces latency and improves security. If your files are behind a private firewall, you will need to generate a temporary signed URL from your storage provider like AWS S3 or Google Cloud Storage before passing the link to the Ceven agent. Once the transcription is complete, the text is returned to the workflow for further processing or storage in your CRM.
Yes. Ceven allows you to specify the model in the workflow settings or via a prompt. You can choose between the general model for standard English or the Nova models for higher accuracy and faster speeds. If you are unsure which one to use, you can use the Get Public Models action to see the latest available versions. The agent can be instructed to use different models based on the audio source, such as using a specific model for phone calls and another for high quality studio recordings to optimize for both cost and accuracy.
Deepgram is highly scalable, but there are specific limits based on your account tier. For most users, pre recorded files can be several hours long. However, you should be aware that very large files may experience timeouts if the network connection is unstable. A common quirk with the Deepgram API is that extremely short files, such as those under one second, may occasionally return an empty transcript or an error depending on the model used. We recommend ensuring your audio clips are trimmed to remove long silences at the start and end for the best results.
By default, Deepgram provides options to opt out of data training. When you connect your account through Ceven, the data handling follows the settings defined in your Deepgram console. If you have a dedicated or enterprise agreement with Deepgram that prohibits data training, those rules are respected at the API level. Ceven does not store your raw audio files, only the resulting text transcripts and summaries if you choose to save them to your knowledge base. You can manage your data retention policies directly within the Deepgram dashboard.
Summarization provides a concise narrative version of the entire conversation, focusing on the overall meaning and outcome. Topic detection, on the other hand, acts like an automated tagging system. It identifies specific keywords and themes throughout the audio, such as pricing, technical support, or onboarding. In a Ceven workflow, you would use summarization to give a human a quick overview of a call, but you would use topic detection to route the transcript to a specific department. For example, if the topic pricing is detected, the agent can automatically notify the sales team.
Ceven primarily interacts with Deepgram via REST APIs for pre recorded audio and text to speech. While Deepgram offers a powerful WebSocket API for real time streaming, Ceven currently optimizes for asynchronous workflows where audio is processed after it is recorded. This means your workflow will trigger once a file is uploaded to your cloud storage. If you need real time interaction, you can set up a system that saves small audio chunks to a URL and has Ceven process those chunks in rapid succession to simulate a real time experience for the end user.
If Deepgram returns an error, such as a 400 bad request due to an unsupported audio format, the Ceven agent will capture the error message and report it in the workflow logs. You can build error handling into your workflow, such as having the agent notify an admin if a file fails to transcribe three times. Common failure points include invalid URLs or files that are too small to process. We recommend using a supported format like WAV or MP3 to ensure the highest success rate across all Deepgram models.
Ceven uses the API key associated with your Deepgram project. The permissions of the agent are identical to the permissions of that key. If your key is restricted to only read usage data and cannot perform transcriptions, the agent will receive a permission denied error when attempting to transcribe. You can use the List Project Scopes action to see exactly what the current key is allowed to do. To grant the agent more power, you must update the key permissions in the Deepgram console and the changes will take effect immediately without needing to reconnect.

Alternatives to Deepgram

Other tools that solve a similar problem. Ceven supports these too, so you can switch or run more than one at once.

AssemblyAI logoAssemblyAIRev AI logoRev AIOpenAI Whisper logoOpenAI Whisper

Try Ceven on your stack

Plug Ceven on top of the tools you already run. Connect Deepgram and the rest of your stack, describe the outcome, and its agents handle the work end to end, days of it in minutes.

Get started for free