Perspective API tester

Send a comments:analyze request to the drop-in endpoint and read the response as your code will.

POST /v1alpha1/comments:analyze Perspective API request

comment.text 43 / 20,480 bytes

languages spanAnnotations

No key needed here. Ctrl or ⌘ Enter sends.

Request body

{
  "comment": {
    "text": "You are an idiot and nobody wants you here."
  },
  "requestedAttributes": {
    "TOXICITY": {},
    "INSULT": {},
    "PROFANITY": {},
    "THREAT": {}
  },
  "doNotStore": true
}

Response

Send the request to see the response.

The scores land here as bars, and the JSON as the endpoint sends it: attributeScores, languages, detectedLanguages.

The scores are Unprofane's, in Google's format.

Every production attribute, scored

Each scored from 0 to 1, with span scores by sentence on request. Google’s experimental names read as the attribute they became.

Attribute	What it scores	Also answers to
`TOXICITY`	Rude, hostile or unreasonable enough to drive someone out of a conversation.	`TOXICITY_EXPERIMENTAL`
`SEVERE_TOXICITY`	Very hateful or aggressive. Far less moved by mild swearing than TOXICITY.	`SEVERE_TOXICITY_EXPERIMENTAL`
`IDENTITY_ATTACK`	Hateful or demeaning about a person for who they are: race, religion, gender, sexuality, disability.	`IDENTITY_ATTACK_EXPERIMENTAL`
`INSULT`	Insulting, belittling or name-calling, aimed at a person or a group.	`INSULT_EXPERIMENTAL`
`PROFANITY`	Swearing, curse words and other obscene language.	`PROFANITY_EXPERIMENTALOBSCENE`
`THREAT`	Wishes or intends harm on a person or a group.	`THREAT_EXPERIMENTAL`
`SEXUALLY_EXPLICIT`	Sexual acts, body parts or other lewd content.	`SEXUALLY_EXPLICIT_EXPERIMENTALFLIRTATION`

Not served, and refused in Google’s words unless the request sets dropUnsupportedAttributes:

ATTACK_ON_AUTHOR
ATTACK_ON_COMMENTER
INCOHERENT
INFLAMMATORY
LIKELY_TO_REJECT
SPAM
UNSUBSTANTIAL
AFFINITY_EXPERIMENTAL
COMPASSION_EXPERIMENTAL
CURIOSITY_EXPERIMENTAL
NUANCE_EXPERIMENTAL
PERSONAL_STORY_EXPERIMENTAL
REASONING_EXPERIMENTAL
RESPECT_EXPERIMENTAL

Move before December 31, 2026

Change one line and keep your client. Each client below is its guide’s own tested code.

What changes

google-api-python-client

 client = discovery.build( "commentanalyzer", "v1alpha1", developerKey=os.environ["UNPROFANE_KEY"],  # up_test_YOUR_KEY on staging, up_live_YOUR_KEY in production−Remove: discoveryServiceUrl="https://commentanalyzer.googleapis.com/$discovery/rest?version=v1alpha1",+Add: discoveryServiceUrl="https://api.unprofane.com/$discovery/rest?version=v1alpha1", static_discovery=False, )  response = client.comments().analyze(body={ "comment": {"text": "you are a clown"}, "requestedAttributes": {"TOXICITY": {}}, "languages": ["en"], "doNotStore": True, }).execute()

1 line changes. Give discovery.build a discoveryServiceUrl on the Unprofane host and keep every call the same.

Tested with google-api-python-client 2.201.0, on every run of the snippet checks. The Python guide

googleapis

 const client = await google.discoverAPI(−Remove: 'https://commentanalyzer.googleapis.com/$discovery/rest?version=v1alpha1',+Add: 'https://api.unprofane.com/$discovery/rest?version=v1alpha1', ); const res = await client.comments.analyze({ key: process.env.UNPROFANE_KEY, // up_test_YOUR_KEY on staging, up_live_YOUR_KEY in production resource: { comment: { text: 'you are a clown' }, requestedAttributes: { TOXICITY: {} }, languages: ['en'], doNotStore: true, }, });

1 line changes. Build the client with google.discoverAPI from the Unprofane discovery URL and keep the calls.

Tested with googleapis 185.0.0, on every run of the snippet checks. The Node.js guide

Configure, Moderation, Toxic comment filter

 Toxic comment filter: On−Remove: API key:              YOUR_GOOGLE_KEY+Add: API key:              up_test_YOUR_KEY−Remove: Custom endpoint:      (empty)+Add: Custom endpoint:      https://api.unprofane.com/v1alpha1 Toxicity model:       TOXICITY (or empty)

2 lines change. Set Coral’s Custom endpoint to https://api.unprofane.com/v1alpha1 and its API key to an Unprofane key. Coral adds comments:analyze and comments:suggestscore itself.

Tested with Coral 9.11.8 itself, run on October 13, 2026. The Coral Talk guide

lib/discourse_perspective.rb, in your fork

−Remove: GOOGLE_API_DOMAIN = "https://commentanalyzer.googleapis.com"+Add: GOOGLE_API_DOMAIN = "https://api.unprofane.com"

Site settings

 perspective_enabled:        true−Remove: perspective_google_api_key: YOUR_GOOGLE_KEY+Add: perspective_google_api_key: up_test_YOUR_KEY perspective_toxicity_model: standard

2 lines change. The discourse-perspective-api plugin hardcodes Google’s host and has no endpoint setting. Fork it, change one line, rebuild, and put an Unprofane key in perspective_google_api_key. A Discourse you cannot add plugins to cannot run the fork.

Tested with Discourse 2026.10.0-latest itself, run on October 13, 2026. The Discourse guide

Any HTTP client

−Remove: curl -X POST "https://commentanalyzer.googleapis.com/v1alpha1/comments:analyze?key=YOUR_GOOGLE_KEY" \+Add: curl -X POST "https://api.unprofane.com/v1alpha1/comments:analyze?key=up_test_YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{ "comment": {"text": "you are a clown"}, "requestedAttributes": {"TOXICITY": {}, "INSULT": {}}, "languages": ["en"], "doNotStore": true }'

1 line changes. The same POST body and the same response, on a different host, with the key as ?key= or a bearer header.

Tested with the request as printed, on every run of the snippet checks. The cURL guide

Unchanged
The path
The request body
The response
The errors
The ?key= parameter

Get your API key The migration hub: figures, the replay tool and every guide

Every production attribute, scored

Move before December 31, 2026

More free tools