Monitor

See every production call, the moment it lands.

Stream live calls from any source into one feed, scored in real time on the same metrics as your tests — so quality problems surface in minutes, not months.

Start monitoring freeBook a demo
app.rubrichq.io/observability/conversations
N
Nooruz zaman's Orga…
Owner
Platform
Dashboard
Test Agents
Co-Pilot
Metrics
Analytics
Prompt Optimizer
Evaluation
Live Simulations
Batch
Observability
Conversations
Notifications
Settings
N
Nooruz zaman
nooruzzonline@gmail.com
ObservabilityConversations

View and analyze all your conversation calls.

FiltersFetch CallsColumnsAdd Call
Call IdentifierAssistantCall SuccessCall End ReasonPhone NumberAvg LatencyUser Int.AI Int.Timestamp
call_9f3ax82kBilling BotPasscustomer ended+1 (415) 555-0148847 ms10Jun 19, 2026, 11:18 AMNEW
call_7b1mz04pSupport BotPassassistant ended+1 (628) 555-0192512 ms01Jun 19, 2026, 11:09 AM
call_2k9qd57tCollections BotFailsilence timed out+1 (917) 555-01731203 ms32Jun 19, 2026, 10:54 AM
call_5w8pe13rSales BotPasscustomer ended+1 (212) 555-0136634 ms00Jun 19, 2026, 10:41 AM
call_3h6tn29vRetention BotEvaluatingcustomer ended+1 (305) 555-0119———Jun 19, 2026, 10:33 AM
call_6p2vc48mSupport BotPassvoicemail+1 (646) 555-0157402 ms00Jun 19, 2026, 10:18 AM
call_8x4rt61wBilling BotFailcustomer ended+1 (213) 555-01641588 ms24Jun 19, 2026, 10:02 AM
call_1d7yk39sOnboarding BotPassassistant ended+1 (312) 555-0188721 ms10Jun 19, 2026, 9:47 AM
call_4j8wq21xSupport BotPassassistant ended+1 (469) 555-0127689 ms01Jun 19, 2026, 9:31 AM
call_9m5tb73cSales BotFailcustomer ended+1 (786) 555-01441342 ms23Jun 19, 2026, 9:14 AM
Page 1 of 18Prev12345Next
Slack · #alerts7:42 PM
5 calls found with latency > 3s
Email · on-call7:42 PM
Latency spike · Billing Bot
Webhook200 OK7:43 PM
Alert pushed to your endpoint
POST api.acme.com · latency.breached

Ingest from anywhere

REST API, CSV, webhooks, or native voice-vendor integrations — every call lands in one feed.

Scored in real time

The same 100s of metrics from your tests run automatically on live production calls.

Alerts & triage

Auto-tag and alert on the calls that need you — escalations, regressions, and violations.

Ingestion

Connect any call source

However your calls are produced, RubricHQ pulls them in. Push from your stack in real time, bulk-import history, or connect a voice vendor in a couple of clicks.

Real-time REST API — POST a call the moment it ends
Bulk CSV import for historical call archives
Native vendor integrations — Vapi, Retell, LiveKit, Pipecat
Webhooks & SDKs to auto-stream from your own stack
REST APIReal-timePush calls as they finish
POST /observability/addaudio_urlagent_id
CSV UploadBulk importBackfill historical calls
calls_june.csv1,204 callsmapped columns
Voice vendorsNativeOne-click connections
VapiRetellLiveKitPipecat
Webhooks & SDKAuto-streamWire into your pipeline
Webhook eventsNode SDKPython SDK

The dashboard

Production health at a glance

Every live call rolls up into the metrics that matter — refreshed in real time across your whole fleet of agents.

Calls today1,284▲ 12% vs yesterday
Resolution rate91.2%▲ 2.4 pts
Avg latency0.46s▼ 0.08s faster
Escalation rate6.4%▼ 1.1 pts
Avg sentiment4.3 / 5▲ 0.2
Containment87%self-serve resolved
Dead air > 3s1.2%▲ 0.3 pts
Compliance pass98.6%3 flagged today
Avg latency · last 24h
Billing Bot · production
Threshold 1.5s2 breaches
00:0004:0008:0012:0016:0020:00now
Set acceptable bounds and out-of-range periods surface at a glance — not buried in logs

More in Monitor

From a live feed to the one call you need to hear.

Production callEscalated
call_9f3ax82k+1 (415) 555-0148· Billing Bot
I was charged twice for my March invoice — I want a refund.
I'm not able to process refunds — let me transfer you to a human agent.
Replay as scenario
Generated scenarioFROM CALL
PersonaMaria C. · frustrated
GoalRefund duplicate charge
InvoiceMarch · $49.99 ×2
ExpectedOffer refund, no transfer
Re-run against your latest promptRun scenario
Replay

Turn any production call into a test

Found a real call that went wrong? Replay it as a scenario — same caller persona, goal, and context — then re-run it against your fixes until it passes. Every production failure becomes a regression test.

Alerts

Get pinged when quality slips

Set thresholds on any metric and RubricHQ alerts you the moment production drifts — a latency spike, an escalation surge, a compliance drop. Straight to Slack, email, or webhook.

New monitorActive
Aggregation
Latency p95Job successEscalation rateCustom metric
Condition
Alert whenLatency p95>1.5soverlast 30 min
Notify
Slack
Email
Webhook
Human review
Delivered to 3 channelsCreate monitor
Needs review31 of 1,284 today
Call IDPhoneAgentTags
call_9f3a+1 (415) 555-0148Billing Bot
hallucination
call_7b1m+1 (628) 555-0192Support Bot
escalatedunresolved
call_5w8p+1 (212) 555-0136Billing Bot
latency 2.8s
call_3h6t+1 (305) 555-0119Support Bot
pii leak
Triage

Surface the calls that need you

Every production call is auto-tagged by the issues it hit — hallucination, escalation, latency, PII — so your team works a clean queue instead of scrubbing thousands of recordings.

How it works

From any source to scored in real time

Production calls stream in from every source, land in one live feed, and each call is scored on your metrics the moment it arrives.

1Ingest from anywhere
2Live call feed
3Scored in real time
REST APILive
POST /api/v1/observability/add
{
"call_id": "call_9f3a…",
"audio_url": "https://…",
"agent_id": "billing-bot",
"duration": 165
}
CSV Upload
CSV
calls_june.csv
1,204 calls · bulk import
External Vendor
VapiRetellLiveKitPipecat
Production calls142 today
call_9f3ax82kNEWResolved
+1 (415) 555-0148·Billing Bot·02:45
call_7b1mz04pEscalated
+1 (628) 555-0192·Support Bot·01:58
call_2k9qd57tResolved
+1 (917) 555-0173·Sales Bot·03:21
call_5w8pe13rVoicemail
+1 (212) 555-0136·Billing Bot·00:52
call_3h6tn29vResolved
+1 (305) 555-0119·Support Bot·02:07
LatencyAudio
0.42savg responsefast
P95 0.88s · 0 dead air > 3s
LLM MetricsLLM-as-judge
Authentication verifiedPass
Empathy & tone
Code MetricsCode
PII redactionPass
Compliance phrasePass

Why it matters

What happens without production monitoring?

Bad calls slip by unnoticed

The 6% of production calls that go wrong never reach a human — until a customer complains, leaves a one-star review, or quietly churns. Manual spot-checks catch almost none of them.

1 bad call shared online = 100s of lost prospects

Regressions ship silently

A prompt tweak or model upgrade degrades quality in production. With no live monitoring, you find out weeks later — from falling CSAT, not from your own dashboards.

Avg regression caught 3+ weeks after deploy

Compliance drift at scale

Disclosures get skipped, PII gets mishandled, scripts drift — quietly, across thousands of live calls. Every untracked call is a potential violation.

Single HIPAA / PCI violation: up to $50K

Put your production calls on watch.

Stream live calls in, score them automatically, and never miss a bad call again.

Start monitoring freeBook a demo