Production monitoring · Autessa Evaluations

Version performance

Version performance splits evaluation results by agent version, every published version and the draft, so you can see whether the latest release is better than the last. A regression after a publish shows up right away, with a shortcut to open that version.

Goes beyond a typical eval tool

Start free

Version Performance BreakdownSAMPLE

71%

78%

84%

74%

89%

v1v2v3v4Draft

Success rate by version. Version 4 dipped after publish; the draft fixes it.

The problem

Agents change; quality should not drift

Instructions, tools and models change every week. Without results split by version, a drop in quality blends into the monthly average and nobody can say which change caused it.

What you get

A Version Performance Breakdown under the monthly stats in Agent History.
Total conversations, evaluated, successful and success rate per version.
A chart and a table view of the same numbers.
A Browse shortcut opens that agent version.
The statistics API returns the same per-version breakdown.
Pairs with eval history, where test runs show the version they tested.

How it works

How does version performance work in Autessa Evaluations?

Publish a version
Customers start talking to it; conversations are tagged with the version.
Open Agent History
Expand Version Performance Breakdown.
Compare
Check success rate per version, then Browse the version to see what changed.

Every option

Per-version numbers

Option	What it does
Total conversations	Conversations with that version in the date range.
Evaluated	How many were evaluated, by people or automatically.
Successful	Including a “successful with caveats” bucket.
Success rate	Share of evaluated conversations that succeeded.

From code

Do it from the CLI or API

Run `autessa <command> help` for the full list of arguments.

CLI documentation →

Get statistics by version from the CLI

autessa agents statistics \
  --agentId <agentId> \
  --startDate 2026-09-01T00:00:00 \
  --endDate 2026-09-30T23:59:59

Permissions

Who can do what?

Who	Can
Read	View the agent and its versions, test-run it, and read its history and results.
Write	Everything Read can, plus edit the draft and record human evaluations.
Publish	Publish and roll back versions, which is what customers get.
Admin	Share and delete the agent. The owner is always an Admin.

Connected

How it connects to the rest of Evaluations

EVALUATIONS FEATURE

Eval history

Test the next version before it reaches customers.

EVALUATIONS FEATURE

AutoEval

Keep live conversations scored so every version has numbers.

AUTESSA PLATFORM

Agent versions

Publish and roll back versions when the numbers say so.

FAQ

Questions about version performance

How do I know if a new version is worse?

Compare its success rate to the previous version in the Version Performance Breakdown. A drop after a publish is visible as soon as conversations are evaluated.

Is the draft included?

Yes. The breakdown includes each published version and the draft.

Can I get it over the API?

Yes. The statistics endpoint returns a `versionBreakdown` with per-version numbers.

Also in Production monitoring:Conversation logs Human evaluation and labeling Quality dashboards Multi-agent evaluation Export and sync

Try version performance in Autessa Evaluations

Version performance splits evaluation results by published version and draft, so a regression after a publish is visible the same day.

Start free

Test sets · simulated users · four-dimension scoring · response checks · grading API

100% customizable, code and all

Make Autessa Evaluations fully your own.

Use it out of the box, then change anything. Ask Autobot, Autessa’s AI builder, for a new field, screen, report, workflow or agent and it builds it into your copy of the app. Or open the code and edit it yourself. Nothing is locked: the app, its data and its code are yours.

Start free and make it yours