Files
JARVISandClaude Opus 5.5 be7d4a4378 multica_agent_host: install the 1Password CLI from 1Password's signed repo (WDWL-37)
Agents get `op` so op:// references resolve at command time and secret
values stay out of prompts and transcripts. Installed per 1Password's
apt instructions (signed-by keyring, debsig policy + keyring,
1password-cli), with the key fingerprint-checked before it is trusted;
EL via their yum repo. multica_op_version pins, "" tracks the repo.
Tag: multica_op.

Optional, default-off hook writes OP_SERVICE_ACCOUNT_TOKEN from the
playbook (ansible-vault) to a 0600 file owned by the daemon user. It is
deliberately not wired into the daemon environment: the user unit
belongs to the multica CLI, and the delivery design is pending.

Rolled out with --tags multica_op to haunted-mansion and spaceship-earth;
second run changed=0, op 2.41.1 as multica, debsig verifies the .deb.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-23 17:11:27 -05:00

186 lines
9.1 KiB
YAML

---
# =============================================================================
# multica_agent_host — defaults
#
# Everything the role needs is a plain variable. The role never reaches into a
# secrets backend itself; see `multica_pat` below.
# =============================================================================
# -----------------------------------------------------------------------------
# Service account
# -----------------------------------------------------------------------------
# The daemon runs as a dedicated unprivileged user with a real home directory.
# It needs a real home (CLI config, agent CLI credentials) and systemd linger,
# so this is NOT a `system: true` account.
multica_manage_user: true
multica_user: multica
multica_group: multica
multica_user_home: "/home/{{ multica_user }}"
multica_user_shell: /bin/bash
# Extra groups for the daemon user (e.g. docker). Left empty by default.
multica_user_extra_groups: []
# -----------------------------------------------------------------------------
# multica CLI — Ansible owns the version
# -----------------------------------------------------------------------------
# Pin it. Re-running the play IS the upgrade path; see multica_disable_auto_update.
multica_version: "0.6.5"
multica_install_dir: /usr/local/bin
multica_release_download_base_url: "https://github.com/multica-ai/multica/releases/download"
# Verify the release tarball against the release's own checksums.txt.
multica_verify_checksum: true
# -----------------------------------------------------------------------------
# Server / workspace
# -----------------------------------------------------------------------------
multica_server_url: "" # REQUIRED, e.g. https://multica.wdw-labs.cloud
multica_app_url: "" # optional frontend URL
multica_workspace_id: "" # optional: pin this profile's default workspace
multica_profile: "" # optional: --profile <name> (isolates config + daemon state)
# -----------------------------------------------------------------------------
# Daemon configuration (mirrors `multica config set` keys)
# -----------------------------------------------------------------------------
# Local disk by default. Repoint this at a mount when one exists, and set
# multica_workspaces_root_mount at the same time so the play fails loudly if
# the mount is missing rather than silently filling the root filesystem.
multica_workspaces_root: "/opt/multica/workspaces"
# Optional guard: assert workspaces_root actually sits on this mount point.
# e.g. "/mnt/utilidor". Empty disables the check.
multica_workspaces_root_mount: ""
multica_workspaces_root_mode: "0750"
multica_device_name: "{{ inventory_hostname.split('.')[0] }}"
multica_runtime_name: "" # "" = leave unset, CLI derives it
multica_max_concurrent_tasks: "" # "" = leave unset
multica_poll_interval: "" # "" = leave unset
multica_heartbeat_interval: "" # "" = leave unset
multica_agent_timeout: "" # "" = leave unset ("0s" disables the cap)
# Ansible owns the version: stop the CLI self-updating out from under the
# role's declared state. Set false only if you want the daemon to self-update.
multica_disable_auto_update: true
# Boot autostart. On Linux `multica daemon autostart enable` writes a systemd
# *user* unit, which only starts at boot when the user has lingering enabled.
multica_autostart: true
multica_enable_linger: true
# -----------------------------------------------------------------------------
# Authentication — backend-agnostic by design
# -----------------------------------------------------------------------------
# REQUIRED. A Multica personal access token: `mul_...` (user PAT) or
# `mcn_...` (Cloud Node PAT). The role takes a VALUE; it does not know or care
# where it came from. Supply it from the playbook with whatever your
# environment already uses:
#
# HashiCorp Vault:
# multica_pat: "{{ lookup('community.hashi_vault.vault_kv2_get',
# 'multica/agent-host').secret.pat }}"
# ansible-vault encrypted var:
# multica_pat: "{{ vaulted_multica_pat }}"
# 1Password:
# multica_pat: "{{ lookup('community.general.onepassword',
# 'Multica Agent Host PAT', field='credential') }}"
# Environment:
# multica_pat: "{{ lookup('env', 'MULTICA_PAT') }}"
#
# Consolidating secret providers later is a one-line playbook change; this role
# is untouched by it.
multica_pat: ""
# Force `multica login --token` even when the host already reports an
# authenticated session against multica_server_url.
multica_force_relogin: false
# -----------------------------------------------------------------------------
# Agent CLIs
# -----------------------------------------------------------------------------
# Which provider CLIs this host should offer. Each must be a key in
# multica_agent_cli_catalog. The daemon registers one runtime per CLI it finds.
multica_agent_clis:
- claude
# Some agent CLIs are installed by piping a vendor script into a shell. That is
# a real supply-chain decision, so it is opt-in and off by default.
multica_allow_remote_install_scripts: false
multica_agent_cli_catalog:
claude:
provider: claude # matches `provider` in `multica runtime list`
binary: claude
method: npm
package: "@anthropic-ai/claude-code"
# Pinned: npm "latest" is 2.1.306 while "stable" is 2.1.304, so an empty
# value silently tracks whichever npm serves that day. 2.1.304 is the
# version running on haunted-mansion.
package_version: "2.1.304"
version_command: "claude --version"
# Subscription auth needs a browser once per host. Not automatable here.
manual_login: true
manual_login_hint: "run `claude` once as {{ multica_user }} and complete the login"
hermes:
provider: hermes
binary: hermes
method: script
install_url: "https://hermes-agent.nousresearch.com/install.sh"
install_dir: "{{ multica_user_home }}/.hermes/hermes-agent"
version_command: "hermes --version"
manual_login: true
manual_login_hint: "configure Hermes credentials as {{ multica_user }} (Hermes reads 1Password natively; confirm against its docs)"
# Keys MERGED (recursively) into the daemon user's ~/.claude/settings.json.
# Merged, not templated: Claude Code writes that file itself. Keys not named
# here are left alone. Empty = don't touch the file.
#
# An npm-installed Claude Code updates itself in place, which silently breaks
# the package_version pin above (and the next play downgrades it again).
# Setting env.DISABLE_AUTOUPDATER: "1" here is what makes the pin hold.
multica_claude_settings: {}
# -----------------------------------------------------------------------------
# 1Password CLI (op)
# -----------------------------------------------------------------------------
# Lets agents resolve op:// references with `op run` / `op inject` / `op read`,
# so secret values never pass through a prompt or a transcript. Installed from
# 1Password's signed apt/yum repository.
multica_op_install: true
# "" = whatever the repository serves, upgraded on every run. Set e.g. "2.41.1"
# to pin. 1Password's apt index carries only the newest build, so a pin older
# than that cannot be installed on a fresh host.
multica_op_version: ""
multica_op_key_url: "https://downloads.1password.com/linux/keys/1password.asc"
# Code signing for 1Password <codesign@1password.com>. The play refuses a key
# at multica_op_key_url with any other fingerprint.
multica_op_key_fingerprint: "3FEF9748469ADBE15DA7CA80AC2D62742012EA22"
multica_op_debsig_policy_url: "https://downloads.1password.com/linux/debian/debsig/1password.pol"
multica_op_apt_base_url: "https://downloads.1password.com/linux/debian"
multica_op_rpm_base_url: "https://downloads.1password.com/linux/rpm/stable"
# Service account token delivery. Off by default. When on, the token is written
# to multica_op_token_file (0600, owned by the daemon user) and nowhere else.
# The role does NOT export it into the daemon environment; see the README.
# Supply the VALUE from the playbook, like multica_pat, e.g.
# multica_op_service_account_token: "{{ vault_op_service_account_tokens[inventory_hostname_short] }}"
multica_op_deliver_token: false
multica_op_service_account_token: ""
multica_op_token_file: "{{ multica_user_home }}/.config/op-service-account/token"
# -----------------------------------------------------------------------------
# Node.js (installed only when a selected agent CLI needs npm)
# -----------------------------------------------------------------------------
# Full semver — this is a pin, not a major-version track.
multica_node_version: "24.22.1"
multica_node_dist_base_url: "https://nodejs.org/dist"
multica_node_prefix: /usr/local/lib/nodejs
# npm global prefix for the daemon user, so agent CLI installs need no root.
multica_npm_prefix: "{{ multica_user_home }}/.npm-global"
# -----------------------------------------------------------------------------
# Verification
# -----------------------------------------------------------------------------
multica_verify: true
# How long to wait for runtimes to appear online after the daemon starts.
multica_verify_retries: 24
multica_verify_delay: 5