How albinjacob/munitas works
Munitas is a self-hosted governance platform for datasets, AI agents, and pipelines. Its main user flow is to register and seal an artifact, request or grant purpose-bound access where needed, run a pipeline, then record an independent gate decision on its output. The console calls the control-plane API; the API coordinates persistence, policy checks, storage, and durable worker execution. The README documents the common governance model and OPA checks; worker and several feature modules are indexed but not excerpted, so their internal links are omitted where not evidenced.
Main entry points: app/main.py Control-plane API · worker/main.py Worker service · pipeline.py Run coordination · App.tsx Web console
Components 25
User console
- Web console
web/src/App.tsx - Identity session
web/src/identity/IdentityContext.tsx - Run monitor
web/src/features/pipeline/PipelineRun.tsx - Pipeline API client
web/src/api/pipeline.ts - DAG pipeline client
web/src/api/dag_pipelines.ts
Governance control plane
- Control-plane API
platform/api/app/main.py - Authentication
platform/api/app/auth.py - Policy evaluation
platform/api/app/opa.py - Grants and leases
platform/api/app/grants.py - Dataset registration
platform/api/app/ingest.py - Sealed versions
platform/api/app/versions.py - Agent registry
platform/api/app/agents.py - Pipeline registry
platform/api/app/dag_pipelines.py - Run coordination
platform/api/app/pipeline.py - Human gate decision
platform/api/app/pipeline.py
Pipeline execution
- Worker service
worker/main.py - Workflow orchestration
worker/workflows.py - DAG execution
worker/dag_workflow.py - Agent run execution
worker/agent_run_workflow.py
Platform services
- Platform database
platform/api/app/db.py - Object storage
platform/api/app/storage.py - SeaweedFS grants
platform/api/app/seaweed.py - Workflow service
platform/api/app/temporal_client.py
Other components
- Operator
- OPA service
Connections 26
- Operator → Web console: uses
- Web console → Identity session: establishes identity
web/src/App.tsx - Web console → Run monitor: monitors runs
web/src/App.tsx - Run monitor → Pipeline API client: requests run data
web/src/features/pipeline/PipelineRun.tsx - Pipeline API client → Control-plane API: calls API
- DAG pipeline client → Control-plane API: registers pipelines
web/src/api/dag_pipelines.ts - Operator → Dataset registration: registers data
- Operator → Agent registry: registers agent
- Operator → Pipeline registry: registers pipeline
- Control-plane API → Authentication: authenticates requests
platform/api/app/auth.py - Control-plane API → Policy evaluation: checks policy
- Control-plane API → Platform database: reads and writes
platform/api/app/db.py - Dataset registration → Object storage: stores content
- Object storage → SeaweedFS grants: selects backend
platform/api/app/storage.py - Grants and leases → SeaweedFS grants: applies prefix grants
platform/api/app/seaweed.py - Run coordination → Workflow service: starts workflows
platform/api/app/pipeline.py - Run coordination → Platform database: records run
platform/api/app/pipeline.py - Workflow service → Worker service: dispatches work
- Worker service → Workflow orchestration: registers workflows
worker/main.py - Worker service → DAG execution: registers workflows
worker/main.py - Worker service → Agent run execution: registers workflows
worker/main.py - Run coordination → Pipeline registry: uses sealed definition
platform/api/app/dag_pipelines.py - Workflow orchestration → Human gate decision: ends at gate
- DAG execution → Human gate decision: ends at gate
- Agent run execution → Human gate decision: requires readiness decision
- Policy evaluation → OPA service: evaluates requests