Configuration reference¶
Every key of the rules file (PAEON_CONFIG), with its type, its default and what it does. The
guides explain how the keys work together; this page is the complete list.
- Every key is optional unless it says required, and a key that is not listed here is refused when the file is loaded: a typing mistake stops the start instead of being ignored.
- A key ending in
_envholds the name of an environment variable. The secret itself goes in the environment (or in the file named byNAME_FILE), never in the rules file. - The executor reads the same format from its own copy of the file and uses the
actionsandnotifierssections.
The environment variables of the processes themselves are in Configuration.
correlation¶
How alerts become incidents. See How it works.
| Key | Type | Default | What it does |
|---|---|---|---|
window_minutes |
number | 30 |
The same alert firing again this soon after its incident closed reopens it. |
host_window_minutes |
number | 5 |
An alert on the same host this soon after an incident opened joins it. |
host_labels |
list of text | [host, instance, hostname] |
Labels that name the host, first match wins. |
hosts |
list of section | Rules that give a host to alerts whose labels carry none. | |
flap |
section | When an incident is flagged as flapping. |
correlation.hosts¶
Gives a host to alerts whose labels carry none (OpenObserve alerts, for example).
| Key | Type | Default | What it does |
|---|---|---|---|
match |
section | required | The alerts this rule gives a host to. |
host |
text | required | The host they are given. |
correlation.hosts.match¶
Matches an alert when every field given matches. An empty matcher matches everything.
| Key | Type | Default | What it does |
|---|---|---|---|
source |
text | The alert's source, exactly: uptime-kuma, alertmanager, openobserve. |
|
title |
text | Regular expression, full match. | |
severity |
unknown | info | warning | critical |
The alert's severity, exactly: info, warning, critical or unknown. |
|
labels |
map of name to text | Exact label values. |
correlation.flap¶
| Key | Type | Default | What it does |
|---|---|---|---|
threshold |
integer | 3 |
Firings within the window. |
window_minutes |
number | 60 |
The window in which the firings are counted. |
maintenance¶
Weekly windows in which alerts are expected. See How it works.
A list of:
A weekly window in which alerts are expected (an updater run, a backup).
| Key | Type | Default | What it does |
|---|---|---|---|
name |
text | required | A name of yours, shown on the incident. |
days |
list of mon | tue | wed | thu | fri | sat | sun |
[mon, tue, wed, thu, fri, sat, sun] |
The days it opens on. Default: every day. |
start |
time (HH:MM) |
required | The local time it opens, HH:MM. |
duration_minutes |
number | required | How long it stays open. It may run past midnight. From 0 to 1440. |
timezone |
text | UTC |
The timezone of start. |
match |
section | The alerts it covers. Empty: all of them. |
maintenance.match¶
Matches an alert when every field given matches. An empty matcher matches everything.
| Key | Type | Default | What it does |
|---|---|---|---|
source |
text | The alert's source, exactly: uptime-kuma, alertmanager, openobserve. |
|
title |
text | Regular expression, full match. | |
severity |
unknown | info | warning | critical |
The alert's severity, exactly: info, warning, critical or unknown. |
|
labels |
map of name to text | Exact label values. |
sources¶
Named monitoring services: where alerts come from, and what the investigator may read in each. See Sources.
A map of a name of yours to a monitoring service. The type key says which one it is, and the other keys depend on it. A source that pushes its alerts posts them to /ingest/<name>.
type: prometheus¶
A Prometheus HTTP API, read with PromQL: Prometheus itself, VictoriaMetrics, Mimir or Thanos. Its alerts arrive through Alertmanager.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | required | The server, with its path prefix if it has one: http://prometheus:9090, http://mimir:8080/prometheus. |
headers |
map of name to text | Headers added to every request, for a tenant: X-Scope-OrgID on Mimir and Loki, THANOS-TENANT, AccountID on VictoriaLogs. Not for secrets. |
|
metrics |
list of text | Metric name patterns (fnmatch) that PromQL queries may read. Empty: none. | |
rules |
boolean | true |
Offer the alerting rules, with their query and their state. |
targets |
boolean | true |
Offer the health of the scrape targets, without their URLs. |
type: alertmanager¶
An Alertmanager: its webhook brings alerts in, and its API says what is silenced.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. |
type: grafana¶
Grafana's alerting: a webhook contact point brings alerts in, and a service account reads the annotations and the alert rules.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. | |
annotations |
boolean | true |
Offer the annotations: alert state changes and deploy marks. |
rules |
boolean | true |
Offer the alert rules. The service account needs the right to read them. |
type: loki¶
Logs in Grafana Loki. The model chooses a stream by its name and gives words to look for; it never writes a selector.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | required | The server, for example http://loki:3100. |
headers |
map of name to text | Headers added to every request, for a tenant: X-Scope-OrgID on Mimir and Loki, THANOS-TENANT, AccountID on VictoriaLogs. Not for secrets. |
|
streams |
map of name to text | Name the model uses, to the LogQL stream selector you wrote, for example {container="api"}. Nothing else is readable. |
|
max_lines |
integer | 100 |
Most lines one search returns. From 1 to 1000. |
type: victorialogs¶
Logs in VictoriaLogs. The model chooses a stream by its name and gives words to look for; it never writes a filter.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | required | The server, for example http://vlogs:9428. |
headers |
map of name to text | Headers added to every request, for a tenant: X-Scope-OrgID on Mimir and Loki, THANOS-TENANT, AccountID on VictoriaLogs. Not for secrets. |
|
streams |
map of name to text | Name the model uses, to the LogsQL filter you wrote, for example {app="api"}. Nothing else is readable. |
|
max_lines |
integer | 100 |
Most lines one search returns. From 1 to 1000. |
type: openobserve¶
An OpenObserve organization, in the rules file instead of the PAEON_OPENOBSERVE_* variables: its alerts are polled, and its logs and metrics are read.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | required | The server, for example http://openobserve:5080. |
org |
text | default |
The organization. |
alerts |
boolean | true |
Poll the triggers stream for alert evaluations. |
poll_seconds |
number | 60 |
How often the alerts are polled. |
stale_seconds |
number | 900 |
Without data for this long, the source is reported stale. |
streams |
list of text | Log streams that may be searched. Nothing else is readable. | |
metrics |
list of text | Metric name patterns (fnmatch) that PromQL queries may read. Empty: none. | |
max_rows |
integer | 100 |
Most rows one search returns. From 1 to 1000. |
type: uptime_kuma¶
uptime-kuma: its Webhook notification brings alerts in, and /metrics says the state of every monitor.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. | |
api_key_env |
text | Variable with an API key, for /metrics. |
type: gatus¶
Gatus: a custom alert brings state changes in, and its API says how each endpoint is.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. |
type: healthchecks¶
Healthchecks (healthchecks.io or your own): a webhook integration brings the checks that go down and up, and a read-only API key reads their state and history.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. https://healthchecks.io if hosted. |
|
api_key_env |
text | Variable with the project's read-only API key. |
type: netdata¶
A Netdata agent: its custom notification brings alert transitions in, and its API reads the active alerts and the charts you list.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. | |
charts |
list of text | Chart id patterns (fnmatch) whose data may be read, for example system.*, disk_space.*. Empty: none. |
type: zabbix¶
Zabbix: a webhook media type brings problems and recoveries in, and an API token of a read-only user reads problems, items and their history.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | The server, as Paeon reaches it. Without it, the source only receives alerts. The frontend's address, without api_jsonrpc.php. |
|
hosts |
list of text | Host name patterns (fnmatch) whose items and history may be read. Empty: none. Problems are listed for every host the token can see. | |
max_values |
integer | 200 |
Most history values one request returns. From 1 to 2000. |
type: beszel¶
A Beszel hub, polled: the alerts that are triggered and the systems that are down. Its notifications carry a sentence only, so nothing is pushed.
| Key | Type | Default | What it does |
|---|---|---|---|
user |
text | User for basic authentication, with password_env. |
|
password_env |
text | Variable with the password of user. |
|
token_env |
text | Variable with a token, sent as Authorization: Bearer. |
|
url |
text | required | The hub, for example http://beszel:8090. |
alerts |
boolean | true |
Poll the triggered alerts and the status of the systems. |
poll_seconds |
number | 60 |
How often the hub is polled. |
stale_seconds |
number | 900 |
Without an answer for this long, the source is stale. |
models¶
Model aliases: a stage names an alias, never a model. See Configuration.
A map of a name of yours to:
| Key | Type | Default | What it does |
|---|---|---|---|
backend |
anthropic | openai |
required | The API the model is called through: Anthropic's Messages API, or the OpenAI chat completions format that most gateways and local servers speak. |
model |
text | required | The model's name, as that API expects it. |
base_url |
text | Defaults to the provider's own API. Use it for a gateway. | |
api_key_env |
text | Name of the environment variable (or NAME_FILE) with the key. | |
max_tokens |
integer | 2000 |
Output ceiling; on recent models it includes thinking. |
timeout_seconds |
number | 60 |
How long one call may take. |
effort |
low | medium | high | xhigh | max |
Anthropic backend only: output_config.effort. |
|
price |
section | What a call costs. Budgets are counted from it. |
models.price¶
Price per million tokens at call time, in the currency of your budget.
| Key | Type | Default | What it does |
|---|---|---|---|
input_per_mtok |
number | 0 |
Price of a million input tokens. 0: this model's calls cost nothing and no budget stops them. |
output_per_mtok |
number | 0 |
Price of a million output tokens. |
cached_input_per_mtok |
number | Price of input read from the cache; input price if unset. |
triage¶
Stage 1: a cheap first look at each incident. See Configuration.
| Key | Type | Default | What it does |
|---|---|---|---|
enabled |
boolean | false |
Switches the stage on. |
model |
text | triage |
An alias from models. |
confidence_threshold |
number | 0.7 |
A verdict under this confidence is not trusted: the incident is treated as real. From 0 to 1. |
settle_seconds |
number | 60 |
Wait this long so that related alerts join the incident. |
max_per_hour |
integer | 20 |
Storm guard. |
daily_budget |
number | What triage may spend in a day, in the currency of price. Spent: incidents are treated as real, with no call. |
|
history_days |
integer | 30 |
How far back the alert's own history goes in what the model is given. |
skip |
list of section | Page-level alerts: never triaged, always treated as real. |
triage.skip¶
Matches an alert when every field given matches. An empty matcher matches everything.
| Key | Type | Default | What it does |
|---|---|---|---|
source |
text | The alert's source, exactly: uptime-kuma, alertmanager, openobserve. |
|
title |
text | Regular expression, full match. | |
severity |
unknown | info | warning | critical |
The alert's severity, exactly: info, warning, critical or unknown. |
|
labels |
map of name to text | Exact label values. |
investigation¶
Stage 2: a strong model with read-only tools. See Configuration.
| Key | Type | Default | What it does |
|---|---|---|---|
enabled |
boolean | false |
Switches the stage on. |
model |
text | investigate |
An alias from models. |
fallback_model |
text | An alias tried once, from the start, when the first model fails or declines. The tool budget is shared between the two. | |
verdicts |
list of noise | transient | known_issue | real |
[real, known_issue] |
Triage verdicts that are investigated. Without triage, every incident is. |
settle_seconds |
number | 60 |
Wait this long after the incident opened, so that related alerts join it. |
max_tool_calls |
integer | 25 |
Tool calls one investigation may make. From 1 to 200. |
max_seconds |
number | 600 |
How long one investigation may last. |
max_result_chars |
integer | 12000 |
A tool result longer than this is cut before the model reads it. |
max_per_hour |
integer | 6 |
Storm guard. |
budget_per_incident |
number | What one investigation may spend, in the currency of price. |
|
daily_budget |
number | Counts every model call of the day, all stages. | |
tools |
section | What the investigator may read. Each tool is off until it is configured. | |
ntfy |
section | Where the analysis is sent. Short for one ntfy channel. | |
notify |
list of text | Names from notifiers: every channel listed gets the message. |
investigation.tools¶
| Key | Type | Default | What it does |
|---|---|---|---|
openobserve |
section | Log search and PromQL in OpenObserve. | |
docker |
section | Container list, inspect and logs. | |
knowledge |
section | Full-text search in your runbooks and notes. | |
uptime_kuma |
section | The state of every monitor. | |
git |
section | Recent commits of working copies. | |
ssh |
section | Prepared commands on other hosts. |
investigation.tools.openobserve¶
What the investigator may read in OpenObserve (with the PAEON_OPENOBSERVE_* account).
| Key | Type | Default | What it does |
|---|---|---|---|
streams |
list of text | Log streams that may be searched. Nothing else is readable. | |
metrics |
list of text | Metric name patterns (fnmatch) that PromQL queries may read. Empty: none. | |
max_rows |
integer | 100 |
Rows one log search may return, at most. From 1 to 1000. |
investigation.tools.docker¶
What the investigator may read from Docker, through a socket proxy limited to GET.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | For example http://dockerproxy:2375. |
logs |
list of text | Container name patterns whose logs may be read (fnmatch). | |
labels |
list of text | Label keys that inspect may return. The others are left out. |
|
mount_sources |
list of text | Host path prefixes that inspect may show as mount sources. |
|
max_log_lines |
integer | 200 |
Log lines one call may return, at most. From 1 to 2000. |
investigation.tools.knowledge¶
| Key | Type | Default | What it does |
|---|---|---|---|
paths |
list of text | Directories of markdown: runbooks, notes, a README. |
investigation.tools.uptime_kuma¶
The current state of every monitor, read from uptime-kuma's /metrics endpoint.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | For example http://uptime-kuma:3001. |
api_key_env |
text | Environment variable (or NAME_FILE) with an API key. |
investigation.tools.git¶
Local clones whose history may be read (the deployment repository, for example).
| Key | Type | Default | What it does |
|---|---|---|---|
repos |
map of name to text | Name the model uses, to the path of a working copy. | |
max_commits |
integer | 30 |
Commits one call may return, at most. From 1 to 200. |
investigation.tools.ssh¶
Other hosts, through commands you list. Pair it with a forced command on the far side.
| Key | Type | Default | What it does |
|---|---|---|---|
hosts |
map of name to section | Name the model uses, to the host and the commands it may run there. | |
identity_file |
text | required | The private key Paeon connects with. |
known_hosts_file |
text | required | Host keys are checked strictly against this file. |
timeout_seconds |
number | 20 |
How long one command may run. From 0 to 120. |
investigation.tools.ssh.hosts¶
| Key | Type | Default | What it does |
|---|---|---|---|
target |
text | required | user@host. |
port |
integer | 22 |
The SSH port. From 1 to 65535. |
commands |
map of name to text | required | Name the model uses, to the command line you wrote. The model chooses a name; it never writes a command or an argument. |
investigation.ntfy¶
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | The ntfy server, for example https://ntfy.example.org. |
topic |
text | required | The topic messages are published to. |
token_env |
text | Environment variable (or NAME_FILE) with an access token. Leave it out for an open topic. |
postmortems¶
Drafts written once an incident stays resolved. See Configuration.
| Key | Type | Default | What it does |
|---|---|---|---|
enabled |
boolean | false |
Switches the stage on. |
dir |
text | postmortems |
The directory drafts are written to. |
git |
section | Commit and push each draft. The directory must be inside a git working copy. | |
model |
text | postmortem |
An alias from models. |
daily_budget |
number | Counts every model call of the day, all stages. Spent: the draft is written without the model's prose. |
postmortems.git¶
Commit each draft in the working copy that holds postmortems.dir.
| Key | Type | Default | What it does |
|---|---|---|---|
commit |
boolean | true |
Commit each draft. |
push |
boolean | false |
Push after the commit, to the branch's own upstream. |
author_name |
text | Paeon |
The commit's author. |
author_email |
text | paeon@localhost |
The commit's author address. |
identity_file |
text | A deploy key for the push, if the remote is reached over SSH. | |
known_hosts_file |
text | Host keys for the push over SSH, checked strictly. |
digest¶
The daily summary. See Configuration.
| Key | Type | Default | What it does |
|---|---|---|---|
hour |
integer | 8 |
The hour, in timezone, from which the day's digest is sent. From 0 to 23. |
timezone |
text | UTC |
The timezone of hour, and of the times the console shows. |
ntfy |
section | Short for one ntfy channel. | |
notify |
list of text | Names from notifiers: every channel listed gets the message. |
|
feedback |
integer | 5 |
After the digest, one short message with right / wrong buttons for each of this many verdicts, those triage would have demoted first. Needs the executor, which signs the buttons. 0: none. From 0 to 20. |
digest.ntfy¶
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | The ntfy server, for example https://ntfy.example.org. |
topic |
text | required | The topic messages are published to. |
token_env |
text | Environment variable (or NAME_FILE) with an access token. Leave it out for an open topic. |
notifiers¶
Named notification channels. A stage sends to those its notify lists. See Notifications.
A map of a name of yours to a channel. The type key says which service it is, and the other keys depend on it. A stage sends to the channels its notify lists.
type: ntfy¶
A topic on an ntfy server. The only service whose buttons can act in one tap.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | The ntfy server, for example https://ntfy.example.org. |
topic |
text | required | The topic messages are published to. |
token_env |
text | Environment variable (or NAME_FILE) with an access token. Leave it out for an open topic. |
type: telegram¶
A chat with a Telegram bot. Buttons are links under the message.
| Key | Type | Default | What it does |
|---|---|---|---|
chat_id |
text | required | The chat, group or channel the bot posts to. |
token_env |
text | required | The bot's token, from BotFather. |
api_url |
text | https://api.telegram.org |
Change it only for a Bot API server of your own. |
type: discord¶
A channel's webhook. Buttons are link buttons under the message.
| Key | Type | Default | What it does |
|---|---|---|---|
webhook_env |
text | required | The webhook URL. It is a secret: whoever has it can post. |
username |
text | The name the message is posted under. Default: the webhook's own. |
type: slack¶
An incoming webhook. Buttons are links in the text: a Slack button, even one that only opens a link, expects an app that answers the click.
| Key | Type | Default | What it does |
|---|---|---|---|
webhook_env |
text | required | The webhook URL. It is a secret: whoever has it can post. |
type: email¶
Mail through an SMTP server. Buttons are links in the body.
| Key | Type | Default | What it does |
|---|---|---|---|
host |
text | required | The SMTP server. |
port |
integer | 587 |
Its port: 587 with starttls, 465 with tls. From 1 to 65535. |
security |
starttls | tls | none |
starttls |
starttls (port 587), tls from the first byte (port 465), or none for a relay on your own network. |
username |
text | The login, if the server asks for one. Goes with password_env. |
|
password_env |
text | Environment variable (or NAME_FILE) with the password. | |
sender |
text | required | The From address, a bare address. |
to |
list of text | required | The addresses the mail is sent to, one or more. |
type: webhook¶
A POST of the message as JSON, for what Paeon has no adapter for (n8n, Node-RED, a script of yours). Buttons are a list of labels and links in the body.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | Where the message is posted. | |
url_env |
text | Instead of url, when the URL holds a secret. |
|
authorization_env |
text | The value of the Authorization header. |
type: gotify¶
An application of a Gotify server. Links are lines of the message.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | The server, without /message. |
token_env |
text | required | The application's token. |
type: pushover¶
Pushover. Links are in the message; the page a tap opens is its one extra URL.
| Key | Type | Default | What it does |
|---|---|---|---|
token_env |
text | required | The application's API token. |
user_env |
text | required | The user or group key. |
device |
text | Limit delivery to one of your devices, by its name. |
type: teams¶
Microsoft Teams, through a Workflows webhook. Buttons open the link.
| Key | Type | Default | What it does |
|---|---|---|---|
webhook_env |
text | required | The workflow's URL. It is a secret: whoever has it can post. |
type: matrix¶
A room on a Matrix homeserver. Links are in the message.
| Key | Type | Default | What it does |
|---|---|---|---|
homeserver |
text | required | The homeserver's base URL, for example https://matrix.example.org. |
room_id |
text | required | The room's id (!abc:example.org). |
token_env |
text | required | An access token of the account that posts. |
type: apprise¶
Any service Apprise knows, for what Paeon has no adapter of its own. Needs the optional install (paeon[apprise]). Links are lines of the message.
| Key | Type | Default | What it does |
|---|---|---|---|
urls_env |
text | required | One or more Apprise URLs, separated by spaces or new lines. They hold the services' secrets. |
actions¶
Stages 3 and 4: plans, approval, execution. See Acting.
Stage 3 and 4: the catalog, who may approve what, and where the executor is. The same file is read by the core and by the executor (paeon executor). The executor trusts only its own copy: mount it read-only there, from git.
| Key | Type | Default | What it does |
|---|---|---|---|
enabled |
boolean | false |
Switches plans on. Without it nothing is proposed to the executor. |
executor_url |
text | http://127.0.0.1:8081 |
Where the core reaches the executor. |
public_url |
text | Where your phone reaches the executor's /act path. Without it there are no buttons, and nothing above suggest can happen. |
|
docker_url |
text | Executor only: a Docker socket proxy that allows start, stop and restart. | |
docker_read_url |
text | Executor only: where it lists and inspects containers, when the proxy above allows nothing but the actions. Defaults to docker_url. |
|
lock_dir |
text | locks |
One lock file per container, shared with other repairers. |
approval_minutes |
number | 30 |
How long a Go link works. After it the plan expires and nothing runs. From 0 to 1440. |
buttons |
one_tap | page |
one_tap |
What a button in a notification does. one_tap: the phone sends the link in the background, and the only answer is the next notification. page: it opens the executor's page, which says what the link does, asks to confirm, and then shows what happened. Two taps, and an answer that needs no notification. |
settle_seconds |
number | 30 |
A runbook waits this long: the alert may clear by itself. |
max_per_hour |
integer | 6 |
Plans executed per hour, at most. |
default_level |
observe | suggest | approve | auto |
suggest |
The level of a plan no policy matches. |
ntfy |
section | The topic for approvals. Give it a token of its own. | |
notify |
list of text | Names from notifiers: where approvals and results are sent. Every channel listed receives the Go link: list only what you alone read. |
|
catalog |
map of name to section | The actions that exist here, each for the targets you name. An action that is not listed cannot be proposed. | |
policies |
list of section | Who decides: the level of each (alert, action, target). The first match wins. | |
autonomy |
section | When auto is earned, and when it is not trusted. |
|
runbooks |
list of section | Known alerts and their fix, run without a model. | |
coolify |
section | Executor only: the Coolify instance coolify_restart uses. |
|
repository |
section | Executor only: the deployment repository config_change opens pull requests on. |
actions.ntfy¶
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | The ntfy server, for example https://ntfy.example.org. |
topic |
text | required | The topic messages are published to. |
token_env |
text | Environment variable (or NAME_FILE) with an access token. Leave it out for an open topic. |
actions.catalog¶
One action of the closed set, switched on for the containers you name.
| Key | Type | Default | What it does |
|---|---|---|---|
targets |
list of text | required | Container name patterns (fnmatch) it may be applied to. |
deny |
list of text | Patterns refused even when a target pattern matches. | |
verify |
section | How the executor decides that the action worked. |
actions.catalog.verify¶
How the executor decides that an action worked. Written by you, never by a model.
| Key | Type | Default | What it does |
|---|---|---|---|
wait_seconds |
number | 15 |
Wait this long after the action before the first check. From 0 to 600. |
timeout_seconds |
number | 120 |
How long the checks may go on before the action is said to have failed. From 0 to 3600. |
http |
text | A URL that must answer below 400 (the service's health endpoint). |
actions.policies¶
The autonomy level of one (alert class, action, target). The first match wins.
| Key | Type | Default | What it does |
|---|---|---|---|
alert |
text | .* |
Alert title, regular expression, full match. |
action |
text | * |
A catalog action, or * for all. |
target |
text | * |
Container name pattern (fnmatch). |
level |
observe | suggest | approve | auto |
required | observe: record only. suggest: say what would be done. approve: ask, and run on your Go. auto: run, and tell you after. |
actions.autonomy¶
| Key | Type | Default | What it does |
|---|---|---|---|
promote_after |
integer | 5 |
Approved and verified runs before auto is proposed. |
window_days |
integer | 90 |
The period in which those runs are counted. |
min_confidence |
number | 0.8 |
Under it, a plan never runs without approval. From 0 to 1. |
actions.runbooks¶
A known signature and its fix: a hit skips both models.
| Key | Type | Default | What it does |
|---|---|---|---|
name |
text | required | A name of yours, recorded with the plan. |
match |
section | required | The alert this runbook answers. |
steps |
list of section | required | The actions, in order. Each must be allowed by the catalog. |
mode |
escalate | sequence |
escalate |
escalate: stop at the first step that fixes it. sequence: run them all. |
actions.runbooks.match¶
Matches an alert when every field given matches. An empty matcher matches everything.
| Key | Type | Default | What it does |
|---|---|---|---|
source |
text | The alert's source, exactly: uptime-kuma, alertmanager, openobserve. |
|
title |
text | Regular expression, full match. | |
severity |
unknown | info | warning | critical |
The alert's severity, exactly: info, warning, critical or unknown. |
|
labels |
map of name to text | Exact label values. |
actions.runbooks.steps¶
| Key | Type | Default | What it does |
|---|---|---|---|
action |
text | required | A catalog action. |
target |
text | required | The container, or Coolify resource, it is applied to. |
actions.coolify¶
Executor only: the Coolify resources coolify_restart may restart, by a name of yours.
| Key | Type | Default | What it does |
|---|---|---|---|
url |
text | required | For example http://coolify:8000. |
token_env |
text | required | An API token that can deploy. |
resources |
map of name to section | Name of yours, to the Coolify resource it stands for. A plan names the resource by this name. |
actions.coolify.resources¶
| Key | Type | Default | What it does |
|---|---|---|---|
uuid |
text | required | The resource's UUID in Coolify. |
kind |
application | service |
application |
What it is in Coolify. |
actions.repository¶
Executor only: a clone of the deployment repository that belongs to Paeon. Paeon checks out branches in it and never merges anything: a change is a branch pushed with the key you name and, with github, a pull request. Merging it is your approval.
| Key | Type | Default | What it does |
|---|---|---|---|
path |
text | required | A working copy nothing else uses. |
remote |
text | origin |
The remote branches are fetched from and pushed to. |
base |
text | main |
The branch every change starts from. |
files |
map of name to section | Name of yours, to a file in which values may be changed. | |
github |
section | Open a pull request for each change. Without it the branch is only pushed. | |
identity_file |
text | A deploy key for fetch and push over SSH. | |
known_hosts_file |
text | Host keys for SSH, checked strictly. | |
author_name |
text | Paeon |
The commit's author. |
author_email |
text | paeon@localhost |
The commit's author address. |
actions.repository.files¶
One file of the deployment repository in which config_change may change values.
| Key | Type | Default | What it does |
|---|---|---|---|
path |
text | required | Relative to the repository. YAML only. |
keys |
list of text | required | Dotted key patterns (fnmatch) that may be changed, for example services.*.mem_limit. The key must already exist and hold a single value. |
values |
text | [A-Za-z0-9][A-Za-z0-9_.:/@+=-]{0,99} |
What a new value must look like, full match. |
actions.repository.github¶
| Key | Type | Default | What it does |
|---|---|---|---|
repo |
text | required | owner/name. |
token_env |
text | required | A token that can open pull requests on that repository only. |
api_url |
text | https://api.github.com |
Change it only for GitHub Enterprise. |
ui¶
The console. See The console.
The console under /ui. It has no login of its own: the reverse proxy protects it.
| Key | Type | Default | What it does |
|---|---|---|---|
enabled |
boolean | false |
Serves the console under /ui. |
require_header |
text | A header your proxy sets after authentication (for example X-authentik-username). A request without it is refused: a second net against a route that bypasses the proxy, not a login: whoever reaches the port can send the header. | |
proxy_secret |
section | A header and a secret value your proxy adds (in Traefik, a headers middleware with customRequestHeaders). A request without that exact value is refused: what reaches Paeon's port without passing the proxy cannot open the console. |
|
public_url |
text | Where you reach the console (https://paeon.example.org). Notifications link to the incident page there. |
ui.proxy_secret¶
A header only your proxy can send: its value is a secret the proxy adds to each request.
| Key | Type | Default | What it does |
|---|---|---|---|
header |
text | X-Paeon-Proxy |
The header your proxy adds. |
value_env |
text | PAEON_UI_PROXY_SECRET |
Name of the environment variable (or NAME_FILE) with the value, 24 characters or more. |
telemetry¶
Traces of Paeon's own work. See Configuration.
Traces of Paeon's own work, over OTLP/HTTP. Facts about each step, never its content.
| Key | Type | Default | What it does |
|---|---|---|---|
otlp_endpoint |
text | required | The OTLP/HTTP base, without /v1/traces. OpenObserve: http://openobserve:5080/api/default. |
authorization_env |
text | Environment variable (or NAME_FILE) with the whole Authorization header, for example Basic <base64 of user:password>. |
|
headers |
map of name to text | Other headers, none of them secret (stream-name). | |
service_name |
text | paeon |
The service.name the traces carry. |
interval_seconds |
number | 30 |
How often finished spans are sent. From 5 to 3600. |