Files
archie/deployments/INSTRUCTIONS.md
T

6.7 KiB

Running this distribution

This archive contains five linux/amd64 binaries:

Each service owns its data and reaches the others only through their gRPC contracts (or NATS, for task actions):

Service Owns Reaches
archie-state-store tasks, events, control-plane settings, identities and access policies, installed packages (PostgreSQL) nothing
archie-gateway conversations (PostgreSQL, its own tables) and every model call: chat turns, curators, MCP tools and sampling State Store; daemon over NATS for task actions
archied dispatch: forge polling, the task queue, the embedded NATS broker, agent containers State Store; Gateway
archie-messaging the Telegram, email and webhook channel connections Gateway; State Store for channel settings
archie-ui the dashboard and the webhook capture receiver Gateway; State Store

archied playbooks lint checks playbook binding YAML before it reaches the daemon, and archied playbooks serve is its language server for editors.

Without archie-messaging the channels are simply absent, so it needs a unit of its own (deployments/systemd-user-service.md).

archie-agent (the per-task sandboxed runtime) is not in this archive — it only ever runs inside the Docker container archied spawns per task, never as a host binary. Pull it from GHCR: ghcr.io/samcharles93/archie-agent.

Quickstart

  1. Pick a starting configuration. The full set of annotated templates lives in this repository's deployments/ directory on GitHub (https://github.com/samcharles93/archie-core/tree/main/deployments) — single-forge-github.toml, multi-forge-github-gitea.toml, local-ollama-standalone.toml, and docker-nats-stack.toml cover the common cases. Copy whichever fits to:

    mkdir -p ~/.config/archie
    cp <template>.toml ~/.config/archie/config.toml
    
  2. Edit config.toml (forge token, repos, [services.state] target, model providers) and place any secrets in ~/.config/archie/env or referenced secret:// locations.

  3. Start the five processes. Order does not matter: a service whose peer is not up yet waits for it and reports it as degraded on its health surface meanwhile. The Gateway waits for the daemon's NATS endpoint when the broker is embedded.

    ./archie-state-store -config ~/.config/archie/config.toml -listen 127.0.0.1:9090 &
    ./archie-gateway     -config ~/.config/archie/config.toml -listen 127.0.0.1:8585 &
    ./archie-messaging   -config ~/.config/archie/config.toml &
    ./archie-ui          -config ~/.config/archie/config.toml &
    ./archied            -config ~/.config/archie/config.toml
    

    Dashboard reachability comes from [web].listen (default 127.0.0.1:8484, which needs no dashboard token). For a non-loopback bind pass -listen with -token or -token-file to mint and persist one. archie-ui reads GATEWAY_TOKEN / STATE_STORE_TOKEN from its process environment only — it has no secret registry, so when those tokens live in a secret manager, whatever starts archie-ui must export them into its environment (the [services.*].target_token values in config.toml are the other supported source).

  4. For a persistent, 24/7 deployment instead of a foreground run, follow the systemd user-unit runbook: https://github.com/samcharles93/archie-core/blob/main/deployments/systemd-user-service.md.

Verifying an install

archied playbooks lint -dir <playbooks-dir> validates playbook binding YAML against the same schema archied loads at startup — run it against your playbooks directory before pointing archied at it.

Version

This archive's binaries are built from the archied release tag in its filename; archie-ui, archie-gateway, archie-state-store and archie-messaging ship at that same version rather than being independently tagged (see RELEASING.md in the source repository for why only archied and archie-agent are independently versioned).

archie-state-store, archie-gateway, archie-messaging, archie-ui and archied are services that must be running and addressable; archie-messaging carries the Telegram, email and webhook channels since the v1.30.0 extraction, so an install (or an upgrade across v1.30.0) needs its unit as well -- see deployments/systemd-user-service.md.

Native self-update

archie-update-install downloads the approved release's Linux/amd64 archive and verifies its SHA256SUMS entry before executing or installing it. Releases must include release.json, all host binaries, and the three updater scripts. Older releases without this contract cannot be installed through the artifact path. --from-source (or ARCHIE_UPDATE_SOURCE_BUILD=1) explicitly clones and builds the selected release instead.

The update check command accepts --channel stable (default), --channel next (includes prereleases), or --channel exact-pin --pin 1.2.3. Set these arguments in the control plane's channel settings alongside the existing check command. The notification destination remains separate from the release channel.

For scheduled, unattended runs, archie-update-install --auto discovers the selected release and refuses if the daemon reports work in flight. It checks again immediately before promotion. A scheduler can set ARCHIE_UPDATE_RELEASE_CHANNEL and ARCHIE_UPDATE_PIN; nothing schedules updates by default. All services stop through their ordinary graceful shutdown path.

release.json supplies the service order and required config sections to both first-install and self-update. A missing unit or config section produces an ordered migration plan and refuses installation. Apply that plan before retrying. Automatic topology edits are intentionally unsupported.

The watchdog switches staged executables by rename, restarts services, and requires the candidate daemon's existing readiness registry to remain healthy for 120 seconds. Agent images come from the release manifest by digest. The image switch updates the control-plane policy, preserving other settings; rollback restores the prior image policy with conflict detection. Source builds use the built image's immutable image ID.

Transactions live in ${XDG_STATE_HOME:-~/.local/state}/archie/update, outside the binary directory. phase, environment, image.json (when applicable), and previous/ describe the transaction. A lock reports the owning process. On daemon boot, an interrupted transaction schedules independent recovery before normal startup. To recover manually, run transaction/watchdog --recover from that state directory. A failed recovery keeps the journal for another attempt.

Database snapshots remain available for manual recovery. Schema migrations and PostgreSQL data are never automatically reversed.