-
Notifications
You must be signed in to change notification settings - Fork 0
Run Without A Vault
Note
Status: pending
Plan: .harness/PLAN.md tasks 2 (vault-less write path + repo-local marker) + 3 (--local-state entry point).
Goal: Opt a single repo into vault-less state so every phase write lands in <repo>/.harness/ — no Obsidian / GDrive / mounted vault required.
Prereqs: Filled by human. (agentm version that ships --local-state, python3 on PATH, a .git dir.)
Single-repo mode lets you drive the full phase workflow on a machine with no memory vault. You opt in explicitly — the installer writes a repo-local .project-mode=local marker at <repo>/.harness/.project-mode, and from then on harness_memory.py reads and writes state under <repo>/.harness/ instead of routing through a vault. For the model behind it, see Single-repo state mode.
Steps are pending — filled from the implementation diff at /work (plan tasks 2–3). The shape will be:
- Opt the repo into local mode at install time —
install.sh --local-state <target>(or the equivalent/setupprompt). Filled by human. - Confirm the repo-local marker landed at
<repo>/.harness/.project-modereadinglocal. Filled by human. - Run a phase write (e.g.
/plan) and confirm state lands repo-local under<repo>/.harness/with noValueError. Filled by human.
- Single-repo state mode — why this mode exists and the vault-vs-local resolution model.
-
Installer CLI reference — the
--local-stateflag. -
Project config reference — how
register()degrades when no vault is present. -
Repo layout reference — where the
.project-modemarker sits on disk.
🔧 How-to
- Your first install
- Install into a project
- Configure a new project
- Update an installed harness
- Cut a release
- Use auto-context in phases
- Use per-project install
- Audit the vault
- Find missing note links
- Use AgentMemory in any agent
- Tune auto-orchestration
- Run without a vault
- Choose a storage backend
- Stand up the memory MCP server