loom/loom.toml.example at main · marin-community/loom · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
# loom.toml — the authored source of truth for every credential/setting a
# loom deploy needs. Copy to `loom.toml` (repo root) and fill in by hand, or
# let `loom setup github-app` / `loom setup secrets` fill it in for you —
# either way, NEVER commit the real one (the repo .gitignore already excludes
# it; this `.example` has no real secrets, so it's safe to commit).
#
# `loom.toml` is not read directly by anything else — run
# `loom config render-env` to turn it into `deploy/standalone/.env`, the file
# docker compose and other deployment tooling consume.
# Re-run it after any edit here; the generated `.env` is not meant to be
# hand-edited.

# The public domain whose DNS A/AAAA record points at the deploy host. Caddy
# obtains a Let's Encrypt certificate for it automatically. `localhost` for a
# local try-out (Caddy then uses its own internal CA — the browser warns;
# expected).
domain = "loom.example.com"

# GitHub login seeded as the first approved user on a fresh database. This is
# the only account that can log in until it approves others. Approved users are
# also who may drive the `@loom` trigger — add more in Settings → Approved users.
owner_github = "your-github-login"

# The GitHub App loom acts through (webhook receiver, REST identity, and
# "Sign in with GitHub"). `loom setup github-app` fills these four in for
# you — hand-typing the private key PEM is impractical, so leave them unset
# here and run the wizard instead unless you registered the App by hand.
# github_app_id = ""
# github_app_private_key = ""
# github_webhook_secret = ""
# github_client_id = ""
# github_client_secret = ""

# Paste-once agent secrets — `loom setup secrets` prompts for these (hidden
# input) instead of you typing them in cleartext here. anthropic_api_key feeds
# the Claude runtime, openai_api_key the Codex runtime; set only the ones you
# use (either can also be an interactive `claude` / `codex login` instead).
# anthropic_api_key = ""
# openai_api_key = ""

# Slack Socket Mode integration (the `/marinbot` / `@marinbot` trigger) —
# optional; unset leaves it off. Both are secrets: an app-level token
# (`xapp-…`, needs the `connections:write` scope) and a bot-user OAuth token
# (`xoxb-…`). See docs/slack-trigger.md.
# slack_app_token = ""
# slack_bot_token = ""

# Deploy-mechanism knobs — every one has a sensible compose-level default, so
# leave them commented out unless you need to override one. (You can also
# override any of these, or anything above, for a single invocation by
# exporting the same-named env var instead of editing this file — see
# `loom config render-env --help`.)
# tls_email = "you@example.com"        # LOOM_TLS_EMAIL: ACME contact for cert-expiry notices
# tls_cert_file = "/path/to/cert.pem"  # Bring your own cert (e.g. `tailscale cert`)
# tls_key_file = "/path/to/key.pem"    #   instead of Caddy's automatic https
# host_uid = "1000"                    # HOST_UID: uid the image runs as
# host_gid = "1000"                    # HOST_GID: gid the image runs as
# image = "loom:latest"                # LOOM_IMAGE: override the image tag

# Optional tool installs the entrypoint runs on every server boot, after the
# pinned Claude/Codex runtime installs — fail-soft, unset by default. The
# first is arbitrary shell: the generic hook for tools loom knows nothing
# about; stage any companion skill under /home/app/.local/share/loom/skills/
# so the second can point at it (absolute paths and $$ for a literal $ —
# Compose expands $VAR in .env values against the host, so $HOME resolves
# to the wrong home). The second is a comma-separated name=source list
# `loom skills install` copies into every installed harness's global
# skills dir. The canonical example installs open-code-review's delegate
# skill — the harness's own model reviews, OCR handles file selection and
# rules, no OCR-side LLM needed. See deploy/README.md "Optional tool
# installs".
# install_cmd = "npm install -g @alibaba-group/open-code-review@1.12.11 && mkdir -p /home/app/.local/share/loom/skills/open-code-review-delegate && curl -fsSL https://raw.githubusercontent.com/alibaba/open-code-review/v1.12.11/skills/open-code-review-delegate/SKILL.md -o /home/app/.local/share/loom/skills/open-code-review-delegate/SKILL.md"
# install_skills = "open-code-review-delegate=/home/app/.local/share/loom/skills/open-code-review-delegate"