TazLab System Map

Scope

This page is the high-level map of the main TazLab systems and how they interact operationally.

Current Synthesis

TazLab is organized as a layered operator-driven system:

  • tazpod/ is the operator environment and secret-handling entrypoint
  • ephemeral-castle/ provisions provider-specific infrastructure and bootstraps cluster foundations
  • tazlab-k8s/ holds the provider-agnostic GitOps desired state managed by Flux
  • mnemosyne-mcp-server/ provides semantic memory retrieval on top of Postgres and Gemini embeddings
  • AGENTS.ctx/ governs contexts, active memory, and cross-project operating rules
  • wiki.tazlab.net/ is the structured documentation and synthesis layer being built to help agents and humans navigate the ecosystem coherently

Main Components

Operator Layer

  • primary repository hub: tazpod
  • tazpod/ manages the local/containerized operator environment
  • secrets are unlocked into RAM and the durable recovery anchor is the encrypted vault stored in S3
  • GitHub access commonly depends on ~/secrets/github-token

Infrastructure Provisioning Layer

Related subtracks:

GitOps Layer

  • primary repository hub: tazlab-k8s
  • tazlab-k8s/ is the Flux-watched desired-state repository
  • it declares infrastructure operators, cluster instances, and applications
  • secrets are consumed via ExternalSecret resources rather than stored inline
  • the concrete layer-by-layer reconciliation graph is documented in TazLab Flux DAG

Technology Reference Layer

Related cross-cutting page:

Memory And Knowledge Layers

  • repository hubs: AGENTS.ctx, mnemosyne-mcp-server, wiki.tazlab.net
  • AGENTS.ctx/memory/ carries the active present-state baseline and chronology
  • mnemosyne provides semantic retrieval over persisted memories
  • wiki.tazlab.net is intended to become the organized documentation surface for durable, interlinked knowledge pages

Related cross-cutting page:

Operating Pattern

The operator works through tazpod, infrastructure is provisioned and bootstrapped through ephemeral-castle, runtime desired state is reconciled through tazlab-k8s, and knowledge is increasingly split across active memory, semantic retrieval, and the wiki depending on the task.

Supporting Pages

Open Questions

  • Which system diagrams or tabular maps should be added next for faster onboarding?
  • Which recurring operational workflows should be promoted from context files into dedicated wiki runbooks?