Skip to content
Pydantic Docs

Architecture

Self-hosted Logfire is made up of independent Kubernetes workloads plus production infrastructure that you provide. The exact rendered workloads depend on the Helm chart version and feature flags, so use the Logfire Helm chart README and values reference for chart-version-specific details.

Runtime Flow

graph
    Client[SDKs, browsers, and API clients]
    Edge[logfire-service]
    Frontend[logfire-frontend-service]
    Backend[logfire-backend]
    Ingest[logfire-ff-ingest]
    Processor[logfire-ff-ingest-processor]
    Query[logfire-ff-query-api]
    Crud[logfire-ff-crud-api]
    Cache[logfire-ff-cache-byte]
    Workers[maintenance and compaction workers]
    Dex[logfire-dex]
    Redis[logfire-redis]
    Postgres[(PostgreSQL)]
    ObjectStore[(Object Storage)]
    IdP[Identity Provider]

    Client --> Edge
    Edge --> Frontend
    Edge --> Backend
    Edge --> Ingest
    Backend --> Dex
    Dex --> IdP
    Backend --> Crud
    Backend --> Redis
    Ingest --> Processor
    Processor --> ObjectStore
    Processor --> Postgres
    Query --> Cache
    Query --> Postgres
    Query --> ObjectStore
    Crud --> Postgres
    Workers --> Postgres
    Workers --> ObjectStore
    Cache --> ObjectStore

Workload Groups

  • Edge and frontend: logfire-service routes public traffic; logfire-frontend-service serves frontend assets.
  • Application backend: logfire-backend handles application APIs, authentication integration, and UI backend behavior.
  • Ingest path: logfire-ff-ingest receives telemetry; logfire-ff-ingest-processor processes and writes it.
  • Query path: logfire-ff-query-api reads telemetry from object storage and metadata from PostgreSQL. Some deployments can also render logfire-ff-query-worker.
  • Metadata APIs: logfire-ff-crud-api handles project, organization, dashboard, and related metadata operations.
  • Cache and coordination: logfire-ff-cache-byte backs query caching; logfire-redis supports live query streaming and autocomplete cache.
  • Background work: logfire-ff-maintenance-worker and logfire-ff-compaction-worker handle maintenance and compaction.
  • Identity: logfire-dex integrates Logfire with your identity provider.
  • Internal telemetry: logfire-otel-collector sends Logfire’s own telemetry to the meta project.
  • Optional features: the chart can render feature-specific workloads such as logfire-remote-mcp and logfire-ai-gateway.

External Dependencies

Self-hosted Logfire depends on production-grade infrastructure outside the chart:

  • PostgreSQL stores application metadata, identity data, and FusionFire metadata for files in object storage.
  • Object storage stores telemetry data.
  • An identity provider supplies user authentication through Dex.
  • Ingress, Gateway API, or your routing layer exposes logfire-service on a stable hostname.

Telemetry payloads are stored in object storage, not PostgreSQL.