Reference

Everything radarmodel-ai 0.1 does, in one page. The README on GitHub says the same in the repository.

Commands

radarmodel-ai init  [-config FILE]           write a starter config
radarmodel-ai check [-config FILE] [flags]   scan once, compare, report
radarmodel-ai watch [-config FILE] [flags]   scan every interval until stopped
radarmodel-ai show  [-config FILE] SOURCE    list the stored models of a source
radarmodel-ai version

The config is radarmodel.json in the current folder, or the file in $RADARMODEL_CONFIG, or -config FILE.

Flags for check and watch:

Flag	What it does
`-min info\|warning\|breaking`	Lowest severity shown at the terminal. Warning by default; the rest is counted
`-json`	Print the report as JSON
`-no-alerts`	Skip webhook, Slack and email
`-dry-run`	`check` only: compare, but keep the stored snapshots

Exit codes

Code	Meaning
0	Scanned, nothing breaking
1	At least one breaking change
2	A provider couldn’t be scanned, another run holds the state folder, or the config is wrong

Breaking wins: a run with a breaking change and a failed provider exits 1, and the report shows both.

Configuration

{
  "interval": "1h",
  "timeout": "30s",
  "min_fraction": 0.5,
  "confirm_removals": true,
  "state_dir": ".radarmodel",
  "providers": [
    {"type": "openai"},
    {"type": "anthropic"},
    {"type": "google"},
    {"type": "openrouter"},
    {"type": "openai", "name": "proxy", "base_url": "https://llm-proxy.example.com", "api_key_env": "PROXY_KEY"}
  ],
  "watch": ["openai:gpt-4o", "anthropic:claude-sonnet-4-5", "google:gemini-2.5-flash", "openrouter:anthropic/*", "proxy:gpt-4o"],
  "alerts": {
    "min_severity": "warning",
    "webhook": {"url_env": "RADAR_WEBHOOK_URL"},
    "slack": {"url_env": "SLACK_WEBHOOK_URL"},
    "email": {"host": "smtp.example.com", "port": 587, "from": "radar@example.com",
              "to": ["ops@example.com"], "username_env": "SMTP_USER", "password_env": "SMTP_PASS"}
  }
}

Setting	Default	What it does
`providers`	none	The lists to read. `type` is `openai`, `anthropic`, `google` or `openrouter`
`providers[].name`	the type	Name used in reports, watch entries and file names. Two entries of one type need different names
`providers[].base_url`	the provider’s API	A proxy or compatible endpoint. `file://` reads saved answers from disk (Linux and macOS)
`providers[].api_key_env`	`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`	Variable holding the key. OpenRouter needs none
`watch`	none	Models your app calls, as `source:model`. `` matches anything, slashes included; a source of `` means every provider
`interval`	`1h`	Time between scans in `watch` mode. At least `1m`
`timeout`	`30s`	Limit for one HTTP request
`min_fraction`	`0.5`	A list with fewer than this share of the last one’s models is a failed scan. `0` turns it off
`confirm_removals`	`true`	A watched removal is breaking only when a second scan agrees. `false` reports it on the first
`state_dir`	`.radarmodel`	Where snapshots are kept, relative to the config file
`alerts.min_severity`	`warning`	Lowest severity that sends an alert. A failed scan always sends one
`alerts.webhook`	none	`url_env`, or `url` for a URL with no token. Gets the JSON report
`alerts.slack`	none	`url_env` naming the incoming-webhook URL. Gets a text summary
`alerts.email`	none	`host`, `port` (587), `from`, `to`, and `username_env` and `password_env` if the server needs them. STARTTLS when offered

Unknown fields are errors, and so are watch entries naming a source that isn’t configured, source names that differ only in case or punctuation, and secret variables that are named but not set.

What counts as what

Only watched models can be breaking or warnings. Every other change is info.

Change to a watched model	Severity
Gone from the list, on two scans in a row	breaking
Gone from the list on one scan	warning
Context or output limit went down	breaking
Lifecycle became `retired` (Anthropic)	breaking
Lifecycle left `active`, such as `deprecated` (Anthropic)	warning
Got or changed a retirement date (Anthropic, OpenRouter)	warning
A limit is no longer published	warning
Price went up, or a price appeared or vanished	warning
Lost a supported method (Google)	warning
Named in `watch` but not in the list	warning, every run
New models, price cuts, bigger limits, name changes	info

When a watched model disappears and a model with the same display name appears in the same scan, the report adds “possibly renamed to …”.

What each provider shows

Provider	What’s compared
OpenAI	Model IDs and owner
Anthropic	Display name, input and output limits, lifecycle, retirement date
Google	Display name, version, input and output limits, supported methods
OpenRouter	Name, context length, max output, input and output price per token, retirement date

What a failed scan is

A failed scan reports an error for that provider, changes nothing in its stored snapshot, and lets the other providers scan as usual. These are all failed scans:

an HTTP error. 408, 429 and 5xx get three attempts in all, honouring Retry-After up to 30 seconds
a body that isn’t valid JSON, is cut off, or has no model list
an empty list
a page of a paged list failing, or a paged answer where one page was expected
fewer models than min_fraction of the last list
a redirect to another host, refused so keys can’t follow it
a timeout

Files

state_dir holds latest/SOURCE.json, the snapshot the next scan compares with, and the last 50 snapshots in history/SOURCE/. Each file is written whole or not at all. A lock file keeps two runs from using the folder at once; one older than 30 minutes is taken to be left from a crash.