AI observability & LLMOps on Google Cloud
Model and LLM traces, experiments, evaluation and operational visibility.
3 documented options on Google Cloud. Provider coverage reflects reviewed evidence, not an exhaustive compatibility claim.
Read this page as JSON or Markdown.
| Listing | Documented clouds | Deployment and infrastructure | Licensing | Installation | Operations and limitations |
|---|---|---|---|---|---|
| Arize Phoenix Self-Hosted | AWS, Azure, Google Cloud, Railway, Render | Customer-operated Servers / containers | Source Available Elastic License 2.0; includes restrictions on providing the software as a hosted/managed service and on license-key circumvention. This is source-available. | Use the official self-hosted Docker/Compose or Kubernetes deployment instructions and configure persistent storage and instrumentation endpoints. | Customer manages trace capture, storage, authentication, retention, evaluation providers, backups and upgrades.Scope and evidenceThe self-hosted product is distinct from commercial Arize offerings. Traces may contain prompts and outputs; evaluation jobs can call separately configured external model services even when trace storage is local.
Cloud evidence reviewed 2026-11-12 |
| Langfuse Self-Hosted | AWS, Azure, Google Cloud, Railway, Render | Customer-operated Servers / containers | Open Source / Open Core MIT core; designated enterprise directories and add-on capabilities use separate license terms or keys. | Docker Compose for a single VM; Helm/Kubernetes and official AWS, Azure and Google Cloud Terraform modules. | Customer operates the deployment, persistence, access policies, retention, backups and scaling.Scope and evidenceTraces can contain prompts, outputs and other sensitive application data, so configure capture and retention deliberately. Self-hosted trace storage does not move the underlying application model into the same account. Optional model-based evaluators or the gateway use configured providers, which may be external. The single-VM Compose example does not provide production high availability or backups. Langfuse v4.82.0 fixes an SSRF path in GitHub Enterprise SSO and moves audit logs to ClickHouse. Review outbound SSO connectivity and audit-log storage during upgrade.
Cloud evidence reviewed 2026-11-12 |
| MLflow Tracking Server Self-Hosted | AWS, Azure, Google Cloud | Customer-operated Servers / containers | Open Source / Open Core Apache 2.0 for upstream MLflow; connected model providers and managed distributions have separate terms. | Python MLflow server command, official Docker Compose bundle or Kubernetes Helm deployment. | Customer operates server access, metadata and artifact stores, retention, backups, scaling and version upgrades.Scope and evidenceTraces can capture model inputs, outputs and intermediate metadata. Point clients at the customer-hosted server and review recording/redaction settings. Applications, evaluators and optional model integrations may still call external providers; hosting the tracking server does not host all training or inference runtimes. Production database, storage and security configuration differ from a local quickstart.
Cloud evidence reviewed 2026-11-12 |
Deployment evidence and category fit
Arize Phoenix Self-Hosted
AI trace collection, evaluation, datasets, experiments and prompt analysis for instrumented model applications.
The self-hosted product is distinct from commercial Arize offerings. Traces may contain prompts and outputs; evaluation jobs can call separately configured external model services even when trace storage is local.
- Phoenix self-hosting deployment options ↗
Vendor / upstream documentation · Reviewed 2026-11-12. Vendor support commitment: not assessed; confirm the selected release and contract.
The current Phoenix self-hosting index includes a Railway one-click template as well as AWS CloudFormation, Azure ARM and Google Cloud Run. These deploy the Phoenix application, not the separately licensed Arize AX product. A Railway project is customer-controlled while Railway manages its infrastructure.
Cloud evidence reviewed 2026-11-12. Full listing and qualifications
Langfuse Self-Hosted
LLM/agent tracing, evaluation and prompt management deployed with customer-controlled storage. Evaluation and optional gateway providers must be scoped independently.
Traces can contain prompts, outputs and other sensitive application data, so configure capture and retention deliberately. Self-hosted trace storage does not move the underlying application model into the same account. Optional model-based evaluators or the gateway use configured providers, which may be external. The single-VM Compose example does not provide production high availability or backups. Langfuse v4.82.0 fixes an SSRF path in GitHub Enterprise SSO and moves audit logs to ClickHouse. Review outbound SSO connectivity and audit-log storage during upgrade.
- Langfuse customer cloud deployment on Google Cloud ↗
Vendor / upstream documentation · Reviewed 2026-11-12. Vendor support commitment: not assessed; confirm the selected release and contract.
The vendor documents deployment of the self-hosted Langfuse application and its backing services in the customer account or project, including provider-specific infrastructure. This is distinct from Langfuse Cloud; database, object storage, model calls and enterprise licensing must be configured for the chosen topology.
Cloud evidence reviewed 2026-11-12. Full listing and qualifications
MLflow Tracking Server Self-Hosted
Experiment tracking and LLM/agent tracing server. This entry describes customer-hosted tracking/observability services, not every model runtime, managed MLflow product or inference backend.
Traces can capture model inputs, outputs and intermediate metadata. Point clients at the customer-hosted server and review recording/redaction settings. Applications, evaluators and optional model integrations may still call external providers; hosting the tracking server does not host all training or inference runtimes. Production database, storage and security configuration differ from a local quickstart.
- Self-hosted MLflow deployment on Google Cloud ↗
Vendor / upstream documentation · Reviewed 2026-11-12. Vendor support commitment: not assessed; confirm the selected release and contract.
The project documents deployment of the MLflow server, metadata database and artifact storage on Google Cloud. This does not refer to a managed MLflow edition or to model serving alone.
Cloud evidence reviewed 2026-11-12. Full listing and qualifications
Deployment and procurement checks
- Private offers
- Cloud Marketplace vendors can negotiate a private offer for a specified Cloud Billing account. Check duration, renewal terms and commitment treatment in the actual agreement. Provider source ↗ · Reviewed 2026-11-14
- Private connectivity
- Private Service Connect connects VPC consumers to supported managed or producer services. Confirm the actual product endpoint and network/DNS architecture. Provider source ↗ · Reviewed 2026-11-14
- External workload access
- Workload Identity Federation can grant external workloads scoped Google Cloud access without long-lived service-account keys. Validate the trust configuration and exact roles. Provider source ↗ · Reviewed 2026-11-14
- Regulated environments
- Assured Workloads applies control packages for residency and access boundaries. Confirm supported services, regions and the product deployment; it is not automatic product certification. Provider source ↗ · Reviewed 2026-11-14
For every provider, confirm marketplace/private-offer terms, commitment eligibility, regional service availability, private connectivity and any vendor access against your selected deployment. Provider capabilities do not automatically establish application support.
Plan this deployment
Account boundary: Google Cloud project. Access: IAM policies and service accounts. Network: VPC network.
Provider documentation ↗ · Guidance reviewed 2026-11-10. Confirm the application edition, region, support access, backup locations and operating responsibilities before choosing.
Compare this category across clouds · Search these listings · Compare 2–4 listings