What's new in EvaliQA, early October 2026

The first October release focuses on voice evaluation: timing metrics on every call, more natural test calls and a new speech-to-text option, plus cleaner project deletion and background report writing.

October 1, 2026 to October 8, 2026

Timing metrics on every voice call

Each call now measures your bot's response latency, interruptions, barge-ins, long silences, talk time and replies that do not suit voice, such as links or long digit strings. The run page sums them up in four tiles.

More natural test calls

EvaliQA's caller no longer talks over your bot, and the transcript records what was actually heard when a line was cut off.

ElevenLabs speech-to-text

One ElevenLabs key now covers both speech-to-text and text-to-speech in voice runs, and Deepgram connections work reliably.

Also shipped

  • Clearer voice runs and costs. Filter voice runs by call type (inbound or outbound), see each run's cost in dollars or credits, and export timing and cost details to CSV. Voice runs now count in projects, on the dashboard and in Reports, and voice plans keep the agent's phone number.
  • Reports write in the background. A new report appears in the list right away as Generating and opens when it is ready, so you no longer wait on a loading screen.
  • Deleting a project removes all its data. Deleting a project now removes its test plans, datasets, runs, reports, voice recordings and traces, and the confirmation shows exactly how much will go.
  • Hotfixes. We shipped several hotfixes during the week, including fixes for connector testing.
  • Dataset and workspace fixes. An edit that leaves a required column empty moves the dataset back to draft and says why, downloaded files get clean names, and your role is shown correctly right after switching workspaces.

All changelogs