# Omnia vs Langfuse

> Omnia acts on what it measures: simulated users before release, rollout on live scores and policy on each tool call, with traces still going to Langfuse. Langfuse records and scores, and acting on it is left to you.

Category: Tracing and evaluation. Checked against Langfuse, now part of ClickHouse’s published material on 2026-10-22. Canonical page: https://www.altairalabs.ai/vs/langfuse/

## We already trace everything in Langfuse.

Keep it. Omnia’s traces go to Langfuse as standard OpenTelemetry. What Langfuse leaves to you is acting on what it sees: holding a version back, rolling one back, refusing a tool call.

## When it matters: An online evaluator flags a jump in wrong answers.

**On Omnia:** That version was on part of the traffic, as every new version is. The same metrics stopped its promotion and rolled it back, and its failing sessions became regression scenarios. (Omnia docs: https://omnia.altairalabs.ai/tutorials/progressive-rollouts/)

**On Langfuse:** Online evaluation scores live traces as they arrive. Langfuse describes its role as ex-post evaluation, and points to other libraries for run-time measures. (source: https://langfuse.com/docs/security-and-guardrails)

## Scores that move traffic

Evaluations score live sessions, and the rollout reads those scores. Promotion and rollback follow from them automatically. *(enterprise)* (Omnia docs: https://omnia.altairalabs.ai/tutorials/progressive-rollouts/)

Langfuse: Online evaluators score traces as they arrive; acting on the score is up to you. (source: https://langfuse.com/docs/evaluation/get-started/online)

## Simulated users, built in

Simulated users ship with the platform and run against each version through its real tools, hostile and duplex voice callers included. The release waits on the suite. (Omnia docs: https://promptarena.altairalabs.ai/arena/explanation/scenario-design/)

Langfuse: Simulated multi-turn conversations are a cookbook built on the OpenEvals library. (source: https://langfuse.com/guides/cookbook/example_simulated_multi_turn_conversations)

## Policy before the tool runs

Each tool call’s arguments and caller are checked against a CEL rule first. Named fields in the response are redacted before the model reads it. *(enterprise)* (Omnia docs: https://omnia.altairalabs.ai/how-to/security/configure-tool-policies/)

Langfuse: Run-time security is left to separate libraries. (source: https://langfuse.com/docs/security-and-guardrails)

## At a glance

| Topic | Omnia | Omnia docs | Langfuse | Langfuse docs | Source |
|---|---|---|---|---|---|
| Open source | PromptArena and PromptKit, Apache 2.0. | https://promptarena.altairalabs.ai/arena/explanation/scenario-design/ | MIT outside the enterprise folders. | documented | https://langfuse.com/docs/open-source |
| OpenTelemetry | Emits it, and ingests it from agents elsewhere. | https://omnia.altairalabs.ai/reference/platform/telemetry/ | Ingests OTLP over HTTP. | documented | https://langfuse.com/integrations/native/opentelemetry |
| CI gate | The suite gates the release. | https://promptarena.altairalabs.ai/arena/how-to/interfaces/run-in-ci/ | A GitHub Action fails on a score threshold. | documented | https://langfuse.com/docs/evaluation/experiments/experiments-ci-cd |
| Simulated users | Built in, against real tools. | https://promptarena.altairalabs.ai/arena/explanation/scenario-design/ | Cookbook. | partly documented | https://langfuse.com/guides/cookbook/example_simulated_multi_turn_conversations |
| Voice testing | Barge-in, overlap, dead air. | https://promptarena.altairalabs.ai/arena/how-to/voice/setup-voice-testing/ | Not documented. | not documented | https://langfuse.com/docs/evaluation/overview |
| Runs and rolls out agents | Yes, with automatic rollback. | https://omnia.altairalabs.ai/tutorials/progressive-rollouts/ | Observes only. | not documented | https://langfuse.com/docs/evaluation/overview |
| Runtime enforcement | Per-call policy and redaction. *(enterprise)* | https://omnia.altairalabs.ai/how-to/security/configure-tool-policies/ | Not in scope. | not documented | https://langfuse.com/docs/security-and-guardrails |

## Questions

**Can I keep Langfuse and use Omnia?**

Yes. Omnia emits standard OpenTelemetry, which Langfuse ingests.

**Who owns Langfuse?**

ClickHouse acquired Langfuse in January 2026; the core stays MIT licensed.