Strong control-plane MVP, not a shrink-wrapped installer. The stack runs today: dashboard, approvals, CRM, orchestration, alerts, worker queueing, DAW-side scaffolding, and LAN/HTTPS access. The remaining productization work is making setup, settings, and recovery feel novice-safe without relying on terminal knowledge.
AI Audio Studio is a Mac-first studio operations platform. It gives a studio one operator-facing control room for intake, inbox handling, content drafting, session prep, QC, revision planning, approvals, and bounded workstation execution.
Think of the product as one control plane with optional worker capacity.
single_machineOne Mac runs the dashboard, database, orchestration, settings, alerts, and optionally the worker too.single_machine + local-workerOne Mac runs the control plane and also claims bounded worker tasks for filesystem and DAW-adjacent execution.control_plane_plus_workerOne always-on Mac runs the control plane and one or more additional workstations runstudio-worker.
The system is not “Mac mini only.” A Mac mini is a good control-plane host, but a Mac Studio or MacBook Pro can run the same stack. The right language is deployment posture, not one specific Apple model.
Core automation surfaces today:
- Lead intake normalization, scoring, draft reply generation, and approval
- Inbox triage, draft generation, and approval
- Content drafting and orchestration scaffolding
- Session prep, QC, render planning, and listening/review scaffolding
- Revision parsing, execution planning, and bounded DAW task queueing
Nothing sends, posts, or executes without an explicit approval boundary.
Use this first unless you already know you need a second workstation.
- best first bring-up path
- best for solo operators
- best for validating the full control room quickly
- remote worker not required
Use this when the same Mac should also handle bounded execution tasks.
- still one machine to manage
- useful for REAPER, local path work, and early DAW rehearsals
- no extra network/storage coordination beyond your local host setup
Use this when you want a dedicated orchestration host and a separate studio workstation.
control plane: dashboard, database, orchestration, approvals, CRM, alerts, n8nworker: mounted storage, plugin inventory, workstation validation, file operations, DAW execution- ideal split: always-on Mac mini or Mac Studio control plane + studio workstation worker
Important: the remote worker is optional capacity, not a separate product tier you must adopt on day one.
The intended access order is:
localhoston the host machineIP + HTTPacross the LANhostname + HTTPSfor normal operator use
Operator front door:
- preferred:
https://$STUDIO_DOMAIN - immediate fallback:
http://<control-plane-ip>:3000
Engineering and worker traffic:
- raw service ports remain valid
- workers can call the control plane by IP or resolvable hostname
- Caddy/HTTPS is for operator ergonomics, not a requirement for first success
Primary docs:
- docs/runbooks/startup.md: canonical bring-up order
- docs/runbooks/local-network.md: LAN, hostname, TLS, and front-door guidance
- docs/runbooks/studio-worker.md: optional worker deployment, shared storage,
launchd, and split-machine guidance - docs/ROADMAP.md: current next milestones
- docs/MASTER_PLAN.md: longer-range product direction
Longer-form guides:
- Docker Desktop installed and running
- Ollama installed on the Mac that will host local inference
infra/.envcreated frominfra/env.example- enough local RAM/disk for Docker, Postgres, n8n, and models
git clone <repo-url> ai-audio-studio
cd ai-audio-studio
cp infra/env.example infra/.env
# edit infra/.env for secrets, passwords, tokens, and machine-local paths
bash scripts/start-ollama.sh
docker compose --env-file infra/.env -f infra/docker-compose.yml up -d
# one-time, idempotent starter workflow import
bash scripts/bootstrap_n8n.sh infra/.envOpen:
- dashboard:
http://localhost:3000 - n8n:
http://localhost:5678
Optional profiles:
# add DAW-oriented services
docker compose --profile daw --env-file infra/.env -f infra/docker-compose.yml up -d
# let this same machine claim worker tasks too
docker compose --profile local-worker --env-file infra/.env -f infra/docker-compose.yml up -dOnly add a second worker machine after the control plane is already healthy:
docker compose --env-file infra/.env -f infra/docker-compose.worker.yml up -dResponsibilities:
- dashboard / control room
- project-state FSM, approvals, worker queueing, audit trail
- CRM, style profiles, workspace settings
- OpenClaw rules, playbooks, alert fan-out
- n8n bootstrap and webhook surface
- LAN/HTTPS operator front door
Optional responsibilities:
- mounted storage execution
- session prep and delivery packaging
- workstation validation and plugin inventory
- dry-run smoke rehearsal
- bounded DAW execution
If you add a remote worker, shared storage is part of the deployment contract.
- the repo does not provision SMB/NFS for you
- both machines must see the project/delivery paths
- if mount points differ, use
PATH_TRANSLATION_JSON
The system fails closed.
| Tier | Name | What it can do |
|---|---|---|
| 1 | Read | Observe and analyze |
| 2 | Draft | Prepare internal draft objects |
| 3 | Queue | Request approval for execution |
| 4 | Narrow Auto | Pre-approved bounded actions only |
If approval is unavailable, jobs wait. They do not auto-approve.
Strong today:
- control-plane stack runs end to end in Docker
- dashboard/control room is live on LAN and HTTPS front-door posture is documented
- single-machine and split-machine deployment stories are both supported
- workspace settings, service settings direction, and worker runtime are established
- worker registration, heartbeats, queueing, dry-run rehearsal, and plugin inventory exist
- operator docs now reflect the actual product posture instead of the legacy two-machine prototype
Still being finished:
- novice-safe onboarding and settings coverage across more services
- richer project review packets and listening/review surfaces
- broader operator-safe recovery and service actions
- deeper canned automation packs and outbound integrations
- live validation on more real DAW/workstation combinations
Core direction:
Bring-up and deployment:
- docs/runbooks/startup.md
- docs/runbooks/local-network.md
- docs/runbooks/studio-worker.md
- docs/runbooks/legacy-cutover.md
Reference and guides:
# stack validation
bash scripts/validate_stack.sh infra/env.example
# inspect logs
docker compose --env-file infra/.env -f infra/docker-compose.yml logs -f openclaw
# restart one service
docker compose --env-file infra/.env -f infra/docker-compose.yml restart project-state