outages.live
Live
DeepgramIncident ·

Elevated errors for Flux streaming STT and Voice Agent sessions

NoticeOfficial update
0

Deepgram Flux streaming speech-to-text requests fail with elevated errors at the US endpoint. Voice Agent sessions that use a Flux listen model can fail to start with a FAILEDTOSTARTLISTENING error. Other models do not have this problem. Deepgram implemented a fix and now monitors the results.

State

Monitoring

Started

Ongoing for

22 min

Worst impact

Notice

Timeline

2 updates
18:4218:4718:5318:5819:04
NoticeUpdate postedTimes in UTC

Investigating

12 min 55%

Monitoring

10 min 45%

Updates

Times in UTC

As posted by Deepgram

Monitoring

A fix has been implemented and we are monitoring the results.

Investigating

We are currently investigating elevated errors for Flux streaming speech-to-text requests at our US endpoint. Voice Agent sessions configured with a Flux listen model may fail to start with a FAILED_TO_START_LISTENING error. Other models are not affected.

Affected components

4
Streaming APIVoice Agent APIDeepgram Public API (api.deepgram.com) - Streaming APIDeepgram Public API (api.deepgram.com) - Voice Agent API

Deepgram in the last 90 days

Full history
90 days agoToday

Incidents in 90 days

17

Days with an incident

16

Median duration

1 h 18 min

This incident

71% shorter than the median

How long each incident lasted, on the day it started. The outlined bar is this incident.

18 h 59 min
median 1 h 18 min
90 days agoToday

Other Deepgram incidents

All incidents

Deepgram·Elevated errors from authentication services at one datacenter

Between 15:17 and 15:45 UTC on October 29, an issue with authentication services at one of our US datacenters caused some API requests and Voice Agent sessions to fail with 5xx errors, and some transcription callbacks sent from that datacenter to carry an invalid Dg-Token header. A fix was deployed at 15:45 UTC and error rates returned to normal. This incident is resolved. All systems are operational.

Resolved

Deepgram·Elevated Latency and Errors for Batch STT

Between 19:27 and 21:53 UTC on October 22, some pre-recorded (batch) requests intermittently had elevated latency, and some timed out. This incident is resolved. All systems are operational.

Resolved

Deepgram·Elevated 5xx Errors

Between 16:00 and 17:09 UTC on October 22, some API requests returned 5xx errors because of an issue with authentication services at one of our datacenters. This incident is resolved. All systems are operational.

Resolved

Deepgram·Elevated 503 errors for bath Flux TTS requests

This issue has been resolved.

Resolved

Deepgram·Elevated 503 errors for Aura-2 TTS

This issue has been resolved.

Resolved

Follow Deepgram

Your coding agent knows when Deepgram is down, and tells you when a failure is theirs, not yours. You can also get Deepgram incidents in a feed reader or in Slack.

Prompt
Install the outages.live MCP server, so that when something breaks you can tell if the cause is a provider we depend on and not my code. It reads the official status pages of cloud, AI, payment and developer providers and gives you their incidents as MCP tools.

Add it as a remote MCP server named "outages" at https://outages.live/mcp (Streamable HTTP, no authentication), with your CLI command or your MCP config file.

Then:
1. Make sure it connects and shows the tools get_provider, list_feed and send_feedback.
2. Call get_provider with "deepgram" and tell me if Deepgram has a problem right now.
3. From now on, when a call to a third-party service fails or becomes slow, call get_provider for that service before you change my code.

Paste this into any coding agent. It adds the server with its own MCP setup.

RSS
https://outages.live/feed.xml?providers=deepgram
Slack
/feed subscribe https://outages.live/feed.xml?providers=deepgram

FAQ

Go from tracking outages to fixing production issues with Polylane.

Try Polylane