Skip to content

Connection types

A connection holds the address and credentials of one system: a Slack workspace, a Postgres database, a Google service account. Monitors, flows, jobs, syncs and notification channels refer to it by id, so no secret lives in a monitor or a job. This page lists every connection type, grouped as the Connections page groups them, with each field as the API takes it.

For the ideas behind connections, read Connections and secrets.

On the Connections page, pick a type and select Connect. The form shows the type’s fields. Or call the API:

POST /api/v1/connections
Content-Type: application/json
{
"type": "postgres",
"name": "orders-db",
"description": "Orders database, read replica",
"config": {
"host": "db.internal",
"database": "orders",
"user": "sreinsights_ro",
"password": "…",
"ssl": true
}
}
Field Type Required Default What it does
type string Yes A type id from this page. An unknown type returns 400 UNKNOWN_INTEGRATION.
name string, 1–100 Yes How monitors, channels, routes and alerts refer to it, e.g. orders-db.
description string, ≤ 500 No What it’s for, who owns it.
clientId uuid No none (shared) The client that owns it. See Shared and client connections.
secretMode sreinsights or runner No sreinsights Where the secrets live. See Keep secrets on runners.
config object No {} The type’s fields, secrets included, as listed on this page.

Rules for config:

  • Unknown fields are refused, so a typo can’t silently store nothing.
  • On create, required fields must be present. Fields that apply only in one mode are required only in that mode. Defaults fill the rest.
  • On update (PATCH /api/v1/connections/:id), every field is optional. An omitted secret is kept. A secret set to null is removed. Removing a secret the current mode needs returns 400 SECRET_REQUIRED.
  • Fields with a pattern (a Slack webhook URL, a Jira project key) are checked against it and the error says what is expected.

GET /api/v1/integrations returns the whole catalog: every type with its fields, channel settings, actions and what it supports. GET /api/v1/connections and GET /api/v1/connections/:id return your connections.

DELETE /api/v1/connections/:id deletes a connection. Its encrypted secrets are erased at once and its name is free to use again. Channels that used it stop delivering. A push source’s ingest token stops working, so the source’s webhooks are refused.

Fields marked secret on this page are encrypted per organization: AES-256-GCM under a data key that is wrapped by the instance’s SREINSIGHTS_MASTER_KEY. The API never returns them. A connection shows each secret only as set or not set, for example { "secrets": { "password": true } }. To change a secret, enter it again.

When a runner leases a run that uses the connection, SREInsights seals the secrets to that runner’s own key, bound to that lease.

A connection that only runners use can keep its secrets on the runners instead. SREInsights then never sees them. In the form, pick Keep them on the runners under Secrets. In the API, send "secretMode": "runner" and leave the secret fields out.

Each runner that uses the connection reads every secret field from an environment variable:

SREINSIGHTS_SECRET_<CONNECTION>_<FIELD>
  • <CONNECTION> is the connection’s name in upper case, with every run of characters other than A–Z and 0–9 turned into one _, and leading and trailing _ removed.
  • <FIELD> is the field name with an _ before each capital that follows a lower-case letter or digit, in upper case.
Connection name Field Variable
orders-db password SREINSIGHTS_SECRET_ORDERS_DB_PASSWORD
Rancher prod bearerToken SREINSIGHTS_SECRET_RANCHER_PROD_BEARER_TOKEN
aws billing secretAccessKey SREINSIGHTS_SECRET_AWS_BILLING_SECRET_ACCESS_KEY

The form and the API (runnerVariables on the connection) show the exact names. They follow the connection’s name, so renaming the connection renames them. Switching an existing connection to runner mode erases the copy stored in SREInsights at once.

Which types can keep secrets on runners:

Non-secret settings may hold ${name} placeholders, for example a base URL of https://${env}.acme.com or a host of ${db_host}. Each check or flow step that uses the connection fills them in on the runner, from, in order:

  1. The step’s or check’s connectionVars, for example { "env": "staging" }.
  2. The run’s variables: a saved flow value, a parameter, or a webhook trigger field with the same name.

A variable nobody set fails the step, and the error names it. Secrets can’t hold variables. A number field may hold a variable only as its whole value, such as ${port}. The connection lists its variables in variables.

The form offers variables on types that have flow actions. See Expressions for the ${…} syntax inside steps.

Some fields apply only while another field has a certain value. Slack, for example, connects with an incoming webhook or a bot token:

{ "type": "slack", "name": "slack-ops", "config": { "mode": "bot", "botToken": "xoxb-…", "defaultChannel": "#alerts" } }

The tables on this page say which mode a field belongs to. A field of another mode is not shown and not required.

A connection without a clientId is shared by the whole organization. A connection with one belongs to that client.

  • A job or monitor of client A may use A’s connections and shared ones. Its job destination must be A’s own.
  • A job or monitor without a client uses shared connections only.
  • Moving a connection to another client (or to shared) is refused with 409 CONNECTION_IN_USE while a job or monitor that couldn’t use it there still does.

See Clients.

Alert channels. SREInsights delivers notices to them itself. A notification channel is one of these connections plus its channel settings, such as a Jira project or a Slack channel.

Type id Label Channel settings Carries reports
pagerduty PagerDuty none No, alerts only
jira Jira Cloud projectKey, issueType, resolveTransitionId No, alerts only
slack Slack channel Yes
teams Microsoft Teams none Yes
email Email (SMTP) to Yes
aws_sns AWS SNS topic none Yes
discord Discord none Yes
gchat Google Chat none Yes
telegram Telegram chatId Yes
webhook Webhook none Yes

Channels that carry reports also take report digests and test messages. Adding a PagerDuty or Jira channel to a report is refused.

Every channel gets one notice per alert change people should hear about: firing, firing again (reopened), escalated, acknowledged and resolved. Chat, email and webhook messages share the same wording: a headline such as Firing · CRITICAL · checkout API: Check failing, then the client, why it fired, the value and the labels, and links to the alert in SREInsights, in its source, and to its runbook when it has one.

Type pagerduty. Opens, acknowledges and resolves one PagerDuty incident per alert.

Field Label Type Required Default Notes
routingKey Events API v2 integration key secret For alerts The 32-character key of the service’s Events API v2 integration. Needed for alerts and to trigger or resolve.
region PagerDuty region select: us, eu No us EU accounts sign in at *.eu.pagerduty.com.
apiToken REST API token secret No Only for flows that list incidents.
defaultSource Default source for flow incidents text No e.g. sreinsights.

What a notice does: every event uses the alert’s fingerprint as dedup_key, so a re-fire or an escalation updates the same incident. Firing, firing again and escalated send trigger; acknowledged sends acknowledge; resolved sends resolve. Severities map as critical → critical, high → error, medium → warning, low and info → info. The incident’s source is the alert’s cluster, service or host label, or sreinsights. Its custom details carry the alert’s labels and value, and its links point to the alert and its runbook.

Flow actions:

Action Label What it does Parameters (* required)
triggerIncident Trigger incident Fires an event. The same dedup key updates one incident instead of opening another. summary*, severity (critical, error, warning, info; default error), dedupKey, source, details
resolveIncident Resolve incident Resolves the incident with a dedup key. dedupKey*
listIncidents List incidents Incidents from the REST API; needs the API token. statuses (default triggered,acknowledged), limit (default 25)

Type jira. Opens a Jira issue when an alert fires, comments when it fires again, and comments on or transitions it when it resolves.

Field Label Type Required Default Notes
baseUrl Jira site URL url Yes https://acme.atlassian.net. Must start with https://.
email Bot account email text Yes The account the API token belongs to.
apiToken API token secret Yes From id.atlassian.com → Security → API tokens, signed in as the bot. The bot needs Browse, Create issues, Add comments and Transition issues in the project.

Channel settings:

Field Label Required Default Notes
projectKey Project key Yes e.g. OPS.
issueType Issue type No Task
resolveTransitionId Transition to run when an alert resolves No A numeric transition id, e.g. 31. Empty: only comment when an alert resolves.

What a notice does:

  • Firing: creates an issue with the summary [SEV1] <alert title> (critical SEV1 to info SEV5), the labels sreinsights, sreinsights-sev-<severity> and one that identifies the alert, and a description with the alert’s details. The issue key is saved on the alert.
  • Firing again, escalated: comments on the open issue (“Still firing.”, “Severity raised to …”, “Firing again. This alert has reopened … times.”).
  • Acknowledged: nothing changes in Jira.
  • Resolved: comments “Resolved in SREInsights after …”, then runs the transition if the channel has one.

Flow actions:

Action Label What it does Parameters (* required)
getIssue Get issue An issue by key, e.g. OPS-123. key*, fields
createIssue Create issue Opens an issue. projectKey, issueType (default Task), summary, description, labels
addComment Add comment Comments on an issue. key, body
searchIssues Search issues Runs a JQL search. jql*, maxResults (default 25), fields
transitionIssue Transition issue Moves an issue by transition id. key, transitionId

Type slack. Posts alerts and flow messages through an incoming webhook or a bot.

Field Label Type Required Default Notes
mode Connect with select: webhook, bot No webhook Webhook: one channel, nothing to install. Bot: any channel the bot is invited to.
webhookUrl Incoming webhook URL secret Yes, webhook mode Starts with https://hooks.slack.com/.
botToken Bot token secret Yes, bot mode Starts with xoxb-. The app needs the chat:write scope.
defaultChannel Default channel text No, bot mode #alerts or C0123ABCD.

Channel setting: channel (Channel (bot connections)), where alerts go with a bot connection. Empty: the connection’s default channel.

What a notice contains: a message coloured by severity (green when resolved) with the headline in bold, the details, a line with the source, when it started and how often it reopened, and buttons to open the alert, its source and its runbook. A bot that isn’t in the channel fails with “Invite the bot to …”.

Flow actions:

Action Label What it does Parameters (* required)
sendMessage Send message Posts a message. Bot connections can choose the channel and reply in a thread. text*, channel (bot mode), threadTs (bot mode)

Type teams. Posts alerts and flow messages as cards to a Teams channel.

Field Label Type Required Default Notes
webhookUrl Webhook URL secret Yes From the channel’s … → Workflows → “Post to a channel when a webhook request is received”. Must start with https://.
format Card format select: adaptive, messagecard No adaptive adaptive for Workflows webhooks; messagecard for legacy Office 365 connectors.

What a notice contains: a card with the headline as its title, the details, the alert’s labels as facts, and buttons to open the alert, its source and its runbook.

Flow actions:

Action Label What it does Parameters (* required)
sendMessage Send message Posts a card. title, text*

Type email. Emails alerts and flow messages through your own SMTP server or provider (SES, SendGrid, Postmark, Microsoft 365).

Field Label Type Required Default Notes
host SMTP host text Yes e.g. smtp.sendgrid.net.
port Port number No 587
secure TLS from the start (port 465) boolean No false Off: STARTTLS on 587.
username Username text No
password Password secret No
from From address text Yes SREInsights <alerts@acme.com>.

Channel setting: to (Recipients, required), comma-separated, e.g. oncall@acme.com, sre@acme.com.

What a notice contains: one email per notice. The subject is the headline. The body comes as plain text and HTML, with the details, value, labels and links. Every email of one alert threads together in mail clients.

Flow actions:

Action Label What it does Parameters (* required)
send Send email Sends an email; subject and body accept ${…} templates. to, subject, body*, html (default false)

Type aws_sns. Publishes alerts to an Amazon SNS topic, which delivers them to its subscribers: email addresses, SMS, queues. The connection has no secrets. SREInsights publishes with its own AWS credentials, so its role needs sns:Publish on the topic.

Field Label Type Required Default Notes
topicArn Topic ARN text Yes e.g. arn:aws:sns:us-east-2:123456789012:alerts. A FIFO topic ends in .fifo.
region Region text Yes The topic’s region, as in its ARN. A region that doesn’t match the ARN fails.

What a notice contains: one message per notice. The subject is plain ASCII of at most 100 characters, which SNS requires, such as [high] checkout API: Check failing, [resolved] … or [acknowledged] …. Characters SNS refuses are replaced or dropped. The body has the headline, the state, when it was first seen, the client, why it fired, the labels and the links. A FIFO topic gets the alert as its message group, so one alert’s notices stay in order.

If SREInsights has no AWS credentials, delivery fails with a message saying so.

Type discord. Posts alerts and flow messages to a Discord channel.

Field Label Type Required Default Notes
webhookUrl Webhook URL secret Yes Channel settings → Integrations → Webhooks → New Webhook. Starts with https://discord.com/api/webhooks/.
username Post as text No SREInsights

What a notice contains: an embed with the headline as its title, linked to the alert, coloured by severity, with the details and links. Mentions are turned off.

Flow actions:

Action Label What it does Parameters (* required)
sendMessage Send message Posts a message. content*, username
sendEmbed Send embed Posts an embed with a title, description and colour. title*, description, color, url

Type gchat. Posts alerts and flow messages to a Google Chat space.

Field Label Type Required Default Notes
webhookUrl Webhook URL secret Yes In the space: Apps & integrations → Webhooks → Add webhook. Starts with https://chat.googleapis.com/.
defaultThreadKey Default thread key (flow messages) text No e.g. nightly-checks.

What a notice contains: plain text with the headline, details and links. Every notice of one alert goes to the same thread.

Flow actions:

Action Label What it does Parameters (* required)
sendMessage Send message Posts text; the same thread key keeps messages in one thread. text*, threadKey
sendCard Send card Posts a card with a header, text and a link button. title, subtitle, text, buttonText, buttonUrl, threadKey

Type telegram. Sends alerts and flow messages to a Telegram chat through a bot.

Field Label Type Required Default Notes
botToken Bot token secret Yes From @BotFather. Add the bot to the chat first.
defaultChatId Default chat id text No -1001234567890 or @channel.

Channel setting: chatId (Chat id). Empty: the connection’s default chat. Group ids start with -100.

What a notice contains: plain text with the headline, details and links, without link previews. A wrong chat id, or a bot that isn’t in the chat, fails at once with Telegram’s reason.

Flow actions:

Action Label What it does Parameters (* required)
sendMessage Send message Sends a message to a chat, group or channel. text*, chatId, parseMode (none, Markdown, MarkdownV2, HTML; default none), disableNotification (default false)

Type webhook. POSTs alert changes, or a flow’s own payload, as JSON to your endpoint.

Field Label Type Required Default Notes
url Endpoint URL url Yes Starts with https:// or http://.
signingSecret Signing secret secret No SREInsights sends X-SREInsights-Signature: sha256=<HMAC-SHA256 of the raw body with this secret>.
authorization Authorization header secret No Sent as is, e.g. Bearer ….

What a notice contains: the notice as JSON, { "version": 1, "kind": "firing", "sequence": 1, "alert": { … } }, with the headers User-Agent: SREInsights-Webhook/1, X-SREInsights-Event: <kind> and X-SREInsights-Delivery: <alert id>:<sequence>. A delivery can arrive twice when a response is lost, so deduplicate on X-SREInsights-Delivery. A report arrives as { "version": 1, "kind", "id", "subject", "text", "url", "color" }.

Flow actions:

Action Label What it does Parameters (* required)
post POST payload Sends { title, message, sentAt, ...extra } as JSON. title, message*, extra, timeoutMs (default 30000)

Alert sources. These tools send their own alerts to SREInsights by webhook. Each alert they send becomes one SREInsights alert, labelled connection=<connection name>. It resolves when the source says it recovered, and your routes, silences and escalation apply as for any other alert. See Receive alerts from other tools.

When you create a Kibana, Dynatrace or Alertmanager connection, SREInsights returns its ingest URL and a token. The token (pit_…) is shown once, in ingestToken of the create response. SREInsights keeps only its hash.

POST /api/v1/ingest/:type/:connectionId
Authorization: Bearer pit_…
Content-Type: application/json
  • :type is kibana, dynatrace or alertmanager, and must match the connection’s type.
  • The answer is 202 { "status": "success", "data": { "received": <signals>, "alerts": [ … ] } }.
  • A missing or wrong token, or a connection that doesn’t exist, returns the same 401 INGEST_UNAUTHORIZED, so the endpoint never confirms which connections exist.
  • A body that isn’t what the source sends returns 400 INGEST_PAYLOAD_INVALID with what to fix in the source’s setup.

The connection’s setup drawer on the Connections page shows the URL, the steps and the payload or configuration to paste, filled in. To replace the token, use Create a new token? there, or POST /api/v1/connections/:id/ingest-token. The old token stops working at once, so the source can’t send alerts until it has the new one.

Type kibana. Existing Kibana rules raise SREInsights alerts. Kibana calls the webhook when a rule’s alert starts and when it recovers. Each alert of a rule (one per host or group) is its own SREInsights alert.

Field Label Type Required Default Notes
kibanaUrl Kibana URL url No Links alerts back to their rule when the rule doesn’t send its own link.
defaultSeverity Severity for rules without a severity tag select: severities No high Tag a rule severity:critical (or high, medium, low, info) to set it per rule.

Set up a Webhook connector in Kibana (Stack Management → Connectors) with method POST, the ingest URL and the headers Authorization: Bearer <token> and Content-Type: application/json. On each rule, add an action with it once for the firing action group and once for Recovered, both with this body:

{
"rule_id": "{{rule.id}}",
"rule_name": "{{rule.name}}",
"alert_id": "{{alert.id}}",
"action_group": "{{alert.actionGroup}}",
"reason": "{{context.reason}}",
"tags": "{{rule.tags}}",
"space": "{{rule.spaceId}}",
"url": "{{rule.url}}"
}

Rule tags such as team:payments or env=prod become alert labels.

Type dynatrace. Dynatrace problems raise SREInsights alerts and resolve when Dynatrace closes them. A runner can also fetch open problems with a Dynatrace check, in case a webhook was lost.

Field Label Type Required Default Notes
tenantUrl Environment URL url No e.g. https://abc12345.live.dynatrace.com. Needed for a runner to fetch open problems.
apiToken API token secret No A token with the problems.read scope, for a runner to fetch open problems.
availabilitySeverity Availability problems select: severities No critical Also used for monitoring-unavailable problems.
errorSeverity Error problems select: severities No high
slowdownSeverity Slowdown and resource problems select: severities No medium
otherSeverity Custom and info problems select: severities No low

In Dynatrace, open Settings → Integration → Problem notifications → Add notification → Custom integration. Use the ingest URL, add the headers Authorization: Bearer <token> and Content-Type: application/json, turn on “Call webhook if problem is closed”, and paste this payload:

{
"pid": "{PID}",
"problemId": "{ProblemID}",
"state": "{State}",
"title": "{ProblemTitle}",
"severity": "{ProblemSeverity}",
"impact": "{ProblemImpact}",
"entities": {ImpactedEntities},
"tags": "{Tags}",
"url": "{ProblemURL}"
}

Type alertmanager. Prometheus alerts, including Rancher Monitoring’s, arrive through an Alertmanager webhook. Each Prometheus alert becomes its own SREInsights alert, per pod or container.

Field Label Type Required Default Notes
defaultSeverity Severity for alerts without a severity label select: severities No high

The alert’s severity label sets its severity: critical and page → critical; high, error and major → high; warning, warn and medium → medium; low and minor → low; info, informational and none → info. Other labels become alert labels.

Add a webhook receiver to Alertmanager (in Rancher: Monitoring → Routes and Receivers, or an AlertmanagerConfig), and keep send_resolved on so alerts resolve in SREInsights too:

receivers:
- name: sreinsights
webhook_configs:
- url: <ingest URL>
send_resolved: true
http_config:
authorization:
type: Bearer
credentials: <token>

Systems your runners query or act on. SREInsights never connects to them itself: the runner does, with the secrets sealed to it for the run. Their secrets can stay on the runners. See Check types and Flow steps.

Type id Label Used by
http HTTP API http checks and steps, integration actions
postgres PostgreSQL postgres checks and steps, integration actions
mysql MySQL mysql checks and steps, integration actions
mongodb MongoDB mongodb checks and steps, integration actions
redis Redis redis checks and steps, integration actions
elasticsearch Elasticsearch elasticsearch checks and steps
prometheus Prometheus promql checks and steps
aws AWS integration actions
github GitHub integration actions, ADAM’s code lookups
gitlab GitLab integration actions, ADAM’s code lookups
linear Linear integration actions
sentry Sentry integration actions
docker_registry Docker registry script checks and steps (registryConnectionId)

The Dynatrace connection under Where alerts come from is also used by dynatrace checks. An integration check or step runs any action of any connection that offers it, including the chat, email, webhook, PagerDuty and Jira actions above.

HTTP API, Elasticsearch and Prometheus connections share an authType field. The fields of the other modes are not shown and not required.

authType Fields Offered by
none none HTTP API, Elasticsearch, Prometheus
bearer bearerToken (Bearer token, secret, required) HTTP API, Prometheus
basic username (Username, required), password (Password, secret) HTTP API, Elasticsearch, Prometheus
header headerName (Header name, required, e.g. X-Api-Key), headerValue (Header value, secret, required) HTTP API, Prometheus
apiKey apiKey (API key, secret, required): a base64 API key from Kibana → Stack Management → API keys, with read on the indices you query Elasticsearch

Type http. A base URL with default headers and credentials, so HTTP checks and flow steps call an API without keeping secrets in the monitor.

Field Label Type Required Default Notes
baseUrl Base URL url Yes https://api.acme.internal or https://${env}.acme.com. A check’s URL that starts with / is a path on it.
headers Default headers textarea No One per line, as Name: value. Headers in a check override these.
authType and its fields Authentication select No none none, bearer, basic, header.
insecureTls Skip TLS certificate verification boolean No false Only for internal services with self-signed certificates.
Action Label What it does Parameters (* required)
request Request Sends a request relative to the base URL. Non-2xx answers fail the step. method* (GET, POST, PUT, PATCH, DELETE; default GET), path*, query, headers, body, timeoutMs (default 30000)

Types postgres and mysql (MySQL or MariaDB). Query a database from a runner: replication lag, stuck jobs, row counts.

Field Label Type Required Default Notes
host Host text Yes db.internal or ${db_host}.
port Port number No 5432 (Postgres), 3306 (MySQL)
database Database text Yes
user User text Yes Checks run read-only. Flow query actions run with this user’s rights, so grant only what flows need.
password Password secret No
ssl Use TLS boolean No false
Action Label What it does Parameters (* required)
query Run SQL Runs any statement, including writes, and returns the rows. sql*, params (bind values, a JSON array), timeoutMs (default 30000)

Type mongodb. Find, count, aggregate or write documents from a runner.

Field Label Type Required Default Notes
uri Connection URI secret Yes mongodb+srv://user:…@cluster.mongodb.net.
database Database text Yes
Action Label What it does Parameters (* required)
find Find documents Returns matching documents. collection*, filter, projection, sort, limit (default 100)
insertOne Insert document Inserts one document. collection, document
updateMany Update documents Applies an update, which must use an operator such as $set, to every match. collection, filter, update*, upsert (default false)
countDocuments Count documents Counts documents matching a filter. collection*, filter
aggregate Aggregate Runs an aggregation pipeline. collection, pipeline

Type redis. Read keys, queue lengths and server info from Redis or Valkey; flows can also write.

Field Label Type Required Default Notes
host Host text Yes redis.internal.
port Port number No 6379
username Username text No
password Password secret No
db Database index number No 0
tls Use TLS boolean No false
Action Label What it does Parameters (* required)
get Get key Reads a string value; null when the key is absent. key*
set Set key Writes a string value, optionally with a TTL. key, value, ttlSeconds (default 0, none)
del Delete keys Deletes keys and reports how many existed. keys* (comma-separated)
incr Increment counter Atomically increments a numeric key. key*, by (default 1)
publish Publish message Publishes to a pub/sub channel. channel, message

Type elasticsearch. Count or search documents from a runner, for example 5xx log lines in the last 5 minutes. The same cluster Kibana reads. No flow actions; use an elasticsearch check or step.

Field Label Type Required Default Notes
url Elasticsearch URL url Yes The Elasticsearch endpoint, not Kibana’s, e.g. https://elastic.acme.internal:9200.
authType and its fields Authentication select No apiKey none, apiKey, basic.
insecureTls Skip TLS certificate verification boolean No false

Type prometheus. PromQL from a runner in your cluster. With Rancher Monitoring, use the in-cluster Prometheus service. Each pod or container gets its own alert (see per-series rules). No flow actions; use a promql check or step.

Field Label Type Required Default Notes
url Prometheus URL url Yes e.g. http://rancher-monitoring-prometheus.cattle-monitoring-system.svc:9090.
authType and its fields Authentication select No none none, bearer, basic, header.
insecureTls Skip TLS certificate verification boolean No false

Type aws. CloudWatch metrics and alarms, Cost Explorer spend and Budgets, read from a runner with an IAM access key. Read-only: nothing here changes AWS. The key’s policy needs cloudwatch:GetMetricStatistics, cloudwatch:DescribeAlarms, ce:GetCostAndUsage, ce:GetCostForecast and budgets:ViewBudget.

Field Label Type Required Default Notes
accessKeyId Access key id text Yes AKIA….
secretAccessKey Secret access key secret Yes
sessionToken Session token (temporary credentials) secret No
region Default region text Yes us-east-1 Where CloudWatch metrics live. Cost Explorer and Budgets are global.
accountId Account id (for Budgets) text No 12 digits, e.g. 123456789012.
Action Label What it does Parameters (* required)
cloudwatchMetric CloudWatch metric One statistic of a metric over the last minutes. Output: value (latest datapoint), datapoints, unit. namespace, metricName, dimensions, statistic (Average, Sum, Maximum, Minimum, SampleCount; default Average), periodSeconds (default 300), minutes (default 15), region
cloudwatchAlarms CloudWatch alarms Alarms in a state, optionally by name prefix. Output: alarms, count. state (ALARM, OK, INSUFFICIENT_DATA; default ALARM), namePrefix, region
costAndUsage Cost and usage Spend over the last days from Cost Explorer, optionally per service. Output: total, currency, days[], services[], latest. days (default 7), granularity (DAILY, MONTHLY), metric (default UnblendedCost), groupBy (default SERVICE)
costForecast Cost forecast Cost Explorer’s forecast for the rest of the month. Output: forecast, currency, periodEnd. metric (default UNBLENDED_COST)
budgets Budgets Every budget’s limit, actual spend and forecast. Output: budgets[], overBudget, count. namePrefix

Type github. Read and write GitHub issues and pull requests, dispatch workflows, and read code for ADAM.

Field Label Type Required Default Notes
token Access token secret Yes ghp_… or github_pat_…. A fine-grained token with access to the repositories flows use. Contents: read is enough for code lookups.
baseUrl API URL (GitHub Enterprise) url No Default https://api.github.com.

Every action takes owner* and repo*.

Action Label What it does Other parameters (* required)
getRepo Get repository Repository details.
getIssue Get issue An issue by number. issueNumber*
createIssue Create issue Opens an issue. title*, body, labels
addIssueComment Comment on issue Adds a comment to an issue or pull request. issueNumber, body
getPullRequest Get pull request A pull request by number. pullNumber*
dispatchWorkflow Dispatch workflow Starts a workflow with a workflow_dispatch trigger, e.g. a rollback. workflowId, ref (default main), inputs
fetchCode Fetch code excerpts Read-only. Finds the files a stack trace names and returns the lines around each. hints*, ref, contextLines (default 60)
getPullRequestDiff Get pull request diff Read-only. A pull request’s changed files with their patches. pullNumber*

Type gitlab. Trigger pipelines, manage issues and read code for ADAM, on GitLab.com or your own GitLab.

Field Label Type Required Default Notes
baseUrl GitLab URL url No https://gitlab.com
token Access token secret Yes glpat-….
defaultProject Default project text No group/project or 12345.

Every action takes an optional project, which overrides the default project.

Action Label What it does Other parameters (* required)
triggerPipeline Trigger pipeline Starts a pipeline on a branch or tag. ref* (default main), variables
getPipeline Get pipeline A pipeline’s status by id. pipelineId*
createIssue Create issue Opens an issue. title*, description, labels
listPipelines List pipelines Recent pipelines, newest first. status (any, running, pending, success, failed, canceled; default any), perPage (default 20)
fetchCode Fetch code excerpts Read-only. Finds the files a stack trace names and returns the lines around each. hints*, ref, contextLines (default 60)
getMergeRequestDiff Get merge request diff Read-only. A merge request’s changed files with their diffs. mergeRequestIid*
addMergeRequestNote Comment on merge request Adds a note to a merge request. mergeRequestIid, body

Type linear. Create, search and comment on Linear issues.

Field Label Type Required Default Notes
apiKey API key secret Yes lin_api_….
defaultTeamId Default team id text No Run List teams to find it.
Action Label What it does Parameters (* required)
listTeams List teams Teams and their ids.
createIssue Create issue Opens an issue on a team. title*, description, teamId, priority (0 none, 1 urgent, 2 high, 3 medium, 4 low; default 0)
searchIssues Search issues Full-text search. term*, limit (default 25)
addComment Add comment Comments on an issue. issueId, body

Type sentry. Count and resolve Sentry issues, for example to fail a monitor when new errors appear.

Field Label Type Required Default Notes
baseUrl Sentry URL url No https://sentry.io
authToken Auth token secret Yes sntrys_…. event:read to list; event:write to resolve.
organization Organization slug text Yes e.g. acme.
defaultProject Default project slug text No e.g. backend.
Action Label What it does Parameters (* required)
listIssues List issues A project’s issues, with Sentry’s search syntax. project, query (default is:unresolved), statsPeriod (1h, 24h, 14d, 30d; default 24h), limit (default 25)
getIssue Get issue One issue by id. issueId*
resolveIssue Resolve issue Marks an issue resolved, ignored or unresolved. issueId*, status (default resolved)

Type docker_registry. Credentials for pulling private images in Docker script checks (runtime: docker, registryConnectionId). No flow actions.

Field Label Type Required Default Notes
registry Registry host text Yes ghcr.io or registry.acme.internal:5000.
username Username text Yes
password Password or access token secret Yes

Accounts data jobs read, and where they write what they collect, including sync sources and destinations. SREInsights never calls them itself: the job’s code reads them from its input file on the runner. There is no test button, and the fields are only checked for shape. Their secrets can stay on the runners. See Job types and Sync spec.

A job names each connection by its role, such as google, shopify or destination:

Type id Label Used by (role)
google_service_account Google service account search_console, ga4, url_inspection (google); syncs from a Google Sheet read as a service account (source)
google_ads Google Ads google_ads (google_ads)
shopify_admin Shopify Admin shopify_orders, shopify_customers, shopify_order_lines, shopify_catalog, onpage_audit (shopify)
s3_destination S3 destination every collecting job type (destination, required); syncs into the lake
aws_secrets_manager AWS Secrets Manager etsy_catalog, etsy_orders (etsy); ebay_catalog, ebay_orders (ebay)
http_destination HTTP API (sync destination) syncs (destination)
postgres_destination Postgres table (sync destination) syncs (destination)

A client’s job must write to the client’s own destination. See Shared and client connections.

Type google_service_account. A Google Cloud service account key, for jobs that read Search Console, Google Analytics 4 and URL Inspection, and for syncs that read a Google Sheet.

Field Label Type Required Default Notes
clientEmail Service account email text No Ends in .iam.gserviceaccount.com. Add this address as a user of each Search Console and GA4 property the jobs read.
key JSON key secret text Yes The whole JSON key file: Google Cloud console → IAM & Admin → Service accounts → Keys → Add key → JSON. Must contain "type": "service_account".

Type google_ads. Google Ads API access through a manager account, for jobs that read campaign and search term performance.

Field Label Type Required Default Notes
developerToken Developer token secret Yes Google Ads → Tools → API Center, in the manager account.
clientId OAuth client id text Yes Ends in .apps.googleusercontent.com.
clientSecret OAuth client secret secret Yes
refreshToken Refresh token secret Yes Of a Google user with access to the accounts, for the scope https://www.googleapis.com/auth/adwords.
loginCustomerId Manager account id text No The manager (MCC) account, e.g. 123-456-7890. Leave empty when the user has direct access.

Type shopify_admin. A Shopify store’s Admin API, for jobs that read orders, customers and the store’s pages.

Field Label Type Required Default Notes
shopDomain Store domain text Yes The store’s myshopify.com address, e.g. acme.myshopify.com, not its own domain.
accessToken Admin API access token secret Yes shpat_…. A custom app with read_orders, read_customers, read_products and read_content.
apiVersion API version text No 2025-10 YYYY-MM.

Type s3_destination. A bucket and prefix where jobs write what they collect, with credentials for that prefix only.

Field Label Type Required Default Notes
bucket Bucket text Yes An S3 bucket name, e.g. acme-data-lake.
prefix Prefix text No e.g. jobs/, no leading /. Jobs write under this prefix only.
region Region text Yes us-east-1
accessKeyId Access key id text Yes An IAM user allowed s3:PutObject (and what else the job needs) on arn:aws:s3:::<bucket>/<prefix>* only.
secretAccessKey Secret access key secret Yes
sessionToken Session token (temporary credentials) secret No Only with temporary credentials from STS.

Type aws_secrets_manager. Secrets under one name prefix in AWS Secrets Manager, for jobs that take a store’s credentials from there. The Etsy and eBay job types read a store’s tokens from <secret prefix>store/<key>. The Stores page connects an Etsy or eBay store by signing in and writes its tokens there. See Stores.

Field Label Type Required Default Notes
region Region text Yes Where the secrets are, e.g. us-east-2.
secretPrefix Secret name prefix text Yes Ends in /, e.g. acme/etsy/. Jobs read secrets under this prefix only.
accessKeyId Access key id text Yes An IAM user allowed secretsmanager:GetSecretValue on the secrets under the prefix, and PutSecretValue only where a job type says it writes one back.
secretAccessKey Secret access key secret Yes
sessionToken Session token (temporary credentials) secret No Only with temporary credentials from STS.
lockBucket Lock bucket text No For a token that rotates on every refresh (Etsy): the S3 bucket of the lock taken before refreshing it, e.g. acme-locks. The key needs s3:GetObject and s3:PutObject on the lock object only.
lockPrefix Lock key prefix text No locks/etsy/ The lock of store <key> is <prefix><key>.

Type http_destination. An API that takes a sync’s rows in batches, by the SREInsights sync protocol, and answers row by row. A sync destination is its own type, so a monitor’s connection can never be written through, and a sync’s can’t be queried by a monitor.

Field Label Type Required Default Notes
baseUrl Base URL url Yes e.g. https://api.example.com, without a query. The sync adds its path to this. http:// only for an address only your own runner reaches.
token Bearer token secret Yes Sent as Authorization: Bearer …, only to this address: redirects are refused.
extraHeaders Extra headers secret text No One Name: value a line, such as a Cloudflare Access service token.
timeoutMs Timeout per batch (ms) number No 30000

Type postgres_destination. One Postgres table a sync writes its rows to, updating each by its key, as a user that can write only that table.

Field Label Type Required Default Notes
host Host text Yes
port Port number No 5432
database Database text Yes
user User text Yes A user with SELECT, INSERT and UPDATE on the table and nothing else.
password Password secret No
ssl TLS select: require, verify-full, disable No require verify-full also checks the server’s certificate against the runner’s trusted CAs.
schema Schema text No public A lower-case schema name.
table Table text Yes A lower-case table name. It needs a sync_key text primary key column; a synced_at timestamptz column is set when a row is written.

The Sync spec has the protocol, the table’s SQL and the lake destination.