Home / Team workspace

A control plane that recommends. You decide what ships.

Your whole team reads the same conversation-QA state from a workspace that runs on your infrastructure end to end.

binds 127.0.0.1 · default --port 8321
$ hotato serve
workspace · conversation inspector
Conversation inspector: one call's origin, per-dimension evaluations with deterministic and advisory results in separate lanes, and reviewer notes.
Conversation inspector — one call, zoomed in: origin, per-dimension results, and the evidence each one links to.

Five views, one instrument.

Release readiness, the scenario matrix, failure clusters, production health, and this inspector all read the same registry — each with a machine mirror at ?format=json, so an agent drives the workspace without scraping HTML.

workspace · scenario matrix
Scenario matrix: three scenarios against the current and previous release, per-dimension status and pass^k reliability.
Scenario matrix — scenarios × release, per-dimension status and pass^k.
workspace · failure clusters
Failure clusters: failed evaluations grouped by observable signature — dimension, assertion kind, reason class — with salience bars.
Failure clusters — grouped by observable signature; the cause is yours to name.
workspace · production health
Production health: ingest volume, evaluated coverage, and per-dimension failure rate over time, recorded and simulated kept apart.
Production health — coverage and per-dimension failure rate over time.

Release readiness sits on the front page. Self-host the workspace →

In your VPC, on loopback, kept apart.

The server binds 127.0.0.1 unless you tell it otherwise, checks one bearer token in constant time, and opens only a listening socket.

Stdlib-only — http.server plus sqlite3 — so a zero-egress test can whitelist loopback and hold every view to it. Recorded and simulated calls compute apart everywhere, and a thin sample stays flagged as low sample, N=3. Redacted transcript segments render as [redacted] in the HTML and the JSON mirror alike, and every request appends one line to an audit log, recording only a token prefix.

Read the self-host guide →

hotato · self-host
$ docker compose up -d
Creating network hotato_default   ... done
Creating hotato-workspace         ... done
Serving http://127.0.0.1:8321 · bearer token required
One container: the SQLite registry, a content-addressed evidence store, all inside your network boundary.

Your whole team, one registry.

Bundled recordings, a per-dimension scorecard, and the same self-hosted workspace, in one command.

$ uvx hotato start --demo

One recording is enough to start. Get started →