What CloudGrid Is

CloudGrid is a focused observability application for OpenTelemetry data from services and AI-agent workloads. It receives OTLP traces, logs, and metrics, stores them behind private Go services, and exposes investigation workflows through a TypeScript GraphQL BFF and a React UI.

Who It Is For

CloudGrid is for engineers who already emit OpenTelemetry and want a local or small-team workspace for debugging service and agent behavior without running a full production observability stack.

Primary users:

User	Main job
Local developer	Run CloudGrid locally, create a project, send OTLP data, and inspect traces, logs, and metrics.
Team engineer	Select a project, investigate telemetry, pivot between evidence, and share URLs.
Platform admin	Manage projects, members, ingest credentials, retention policies, and alert rules.
AI-agent engineer	Create datasets, run evaluations, inspect metric results, compare candidates, and review optimization evidence when AI Eval is enabled.

What CloudGrid Does Today

• Accepts OTLP/HTTP JSON and protobuf for traces, logs, and metrics on 4318.
• Accepts OTLP/gRPC protobuf for traces, logs, and metrics on 4317.
• Routes ingestion through NATS JetStream to storage-write.
• Persists telemetry in SurrealDB through private Go services.
• Reads telemetry through GraphQL queries served by the TypeScript BFF.
• Streams live trace updates through GraphQL subscriptions backed by storage-read.
• Manages companies, projects, memberships, ingest credentials, dashboards, retention policies, and alert records through control-plane.
• Executes project alert rules through the alert evaluator, with in-app/email delivery and bridge-backed adapter paths for provider-specific notifications.
• Supports local no-login mode and deployed SSO mode.
• Supports optional AI evaluation and optimization workflows behind a feature flag.

What Is Still Production-Readiness Work

The specs define the production target, and the repository now includes release workflow and Helm chart definitions. Published signed service images, SBOM/provenance output, and release manifests exist only after the release workflow runs. Retention policy CRUD and alert rule/silence/history CRUD are implemented; production retention deletion still depends on enabling the storage-maintenance scheduler and SurrealDB retention adapter for the target deployment. Alert execution, project discovery, dashboard alert widgets, and in-app/email/webhook delivery are implemented; provider-specific delivery such as Slack, WhatsApp, SMS, or incident tools uses the bridge-backed adapter path.

Do not configure CloudGrid local mode on an untrusted network. Local mode intentionally skips login.

Core Data Flow

diagram

Next Step

Choose a runtime mode in Runtime modes, then run the local stack with Local quickstart.