This directory is the detailed documentation source of truth for the MirrorNeuron workspace. It is written for internal maintainers, contributors, integrators, and operators; the documentation site presents a shorter reader-facing layer. When the two differ, verify the implementation and update both in the same change.
MirrorNeuron is a runtime for message-driven workflows with a Python CLI/API layer, an Elixir/BEAM core, Redis-backed runtime state, and blueprint-oriented run records. It can run on one machine or across runtime nodes.
Every user-visible behavior has one canonical detailed page in this directory:
| Fact | Canonical page | Implementation source to verify |
|---|---|---|
| Project fit, boundaries, and non-guarantees | Why MirrorNeuron | Runtime, runner, and security implementation. |
| Shared vocabulary | Core Concepts | Manifest/runtime terms and CLI/API contracts. |
| Installation and local services | Installation | mn-deploy/install.sh, server.sh, runtime start code. |
| CLI syntax and behavior | CLI Reference | mn-cli/mn_cli/main.py and command modules. |
| REST API shape | API Reference | mn-api routes and running FastAPI OpenAPI schema. |
| Environment defaults | Environment Variables | CLI, API, SDK, and Core configuration definitions. |
| Blueprint and manifest contracts | Blueprint Standard, Job Bundle Format | Schemas, SDK validators, and checked-in blueprints. |
| Reliability and recovery | Reliability Guide | Core persistence, lease, scheduler, and recovery tests. |
| Bounded working context and compression | Context Intelligent System | Membrane working-memory contract and Python SDK context-session tests. |
| Security boundaries | Security Model | Runner, network, Redis, secret, and policy behavior. |
| Repeated operational failures | Troubleshooting | Reproducible diagnostics and component tests. |
Do not duplicate large command tables, API shapes, or environment-variable lists on tutorial pages. Link to the canonical reference instead.
- Core Concepts
- Blueprint Standard
- Blueprints and Skills
- Python SDK
- Context Intelligent System
- Testing
- Installation
- Services and Health Checks
- Monitor Guide
- Reliability Guide
- Redis High Availability
- Troubleshooting
| Topic | Page |
|---|---|
| Local installation and verification | installation.md |
| First local workflow | quickstart.md |
| Checked-in blueprint selection | examples.md |
| CLI, API, SDK, and configuration | cli.md, api.md, SDK.md, env_variables.md |
| Models, OpenShell, services, and resources | model-runtime.md, docker-model-runner-arm64-nvidia.md, docker_and_openshell_for_blueprints.md, services-and-health-checks.md, resources-and-devices.md |
| Cluster, deployment, schedules, and HA | cluster.md, deployments.md, schedules-and-events.md, redis-ha.md |
| Architecture, context, reliability, and security | runtime-architecture.md, cluster_architecture.md, context-memory.md, reliability.md, security.md |
| Internal quality and contributor process | testing.md, contributing.md, documentation-style.md |
See September 2026 alignment for the current job/run, source-package, dependency, model-routing, and execution-boundary corrections, with validation evidence and unexercised integrations.
When a pull request changes a command, API route, configuration key, manifest field, runtime guarantee, runner boundary, or failure behavior:
- update the canonical detailed page in
mn-docs; - update the concise corresponding page in
mn-doc-site/content/docswhen the behavior is user-facing; - update a tutorial, troubleshooting page, and migration note when the reader journey changes;
- run the documentation build and the smallest relevant product validation; and
- record the implementation source and validation command in the pull request.