Text Moderation

Submit text for moderation to detect policy, safety, fraud, privacy, and prompt attack risks.

Requirements

Item	Specification
Encoding	UTF-8
Method	POST
Recommended timeout	1s

Request

Request URL

Cluster	Request URL
USA (Virginia)	`http://api-text-fjny.fengkongcloud.com/text/v4`
Singapore	`http://api-text-xjp.fengkongcloud.com/text/v4`

Request Parameters

Parameter	Type	Required	Description
`accessKey`	string	Yes	API authentication key. The default `accessKey` is sent in your onboarding email.
`appId`	string	Yes	Application identifier, such as `web` for your web application or `app` for your mobile app. The default `appId` is sent in your onboarding email. Contact DeepCleer if you need a new `appId`.
`eventId`	string	Yes	Event identifier used to distinguish moderation scenarios in your application, such as `prompt` for user prompts or `modelOutput` for LLM output. The default `eventId` is sent in your onboarding email. Contact DeepCleer if you need a new `eventId`.
`type`	string	Yes	Risk detection types. Detailed type refers to Detection Types. The type field accepts multiple detection categories joined with underscores — for example, set type to `BAN_EROTIC` to moderate for prohibited and pornographic content in a single call. `TEXTRISK` is a bundled general-risk check that covers `POLITY`, `VIOLENT`, `BAN`, `EROTIC`, `DIRTY`, `ADVERT`, `PRIVACY`, `ADLAW`, and `MEANINGLESS`. When `TEXTRISK` is combined with any of these individual categories, the result is their union — so `TEXTRISK_POLITY` behaves the same as `TEXTRISK` alone, because `POLITY` is already included.
`data`	json_object	Yes	Request data content. Max 1 MB. See data Object.
`translationTargetLang`	string	No	Enables online translation of the input before it is moderated. When set, the service translates `text` (and `nickname` / `relateText` if provided) into the target language and applies all policy checks to the translated output. Supported values: `en` (English). Contact DeepCleer to enable this feature.
`acceptLang`	string	Yes	Language for returned labels. Set `en` by default.
`callback`	string	No	Callback HTTP URL. When provided, DeepCleer posts the moderation result to this endpoint as soon as it becomes available.

Detection Types

Value	Description
`POLITY`	Political content
`VIOLENT`	Violence and terrorism
`BAN`	Prohibited content
`EROTIC`	Sexual or pornographic content
`DIRTY`	Profanity and abusive language
`ADVERT`	General advertising
`PRIVACY`	Personally identifiable information
`ADLAW`	Advertising law violations
`MEANINGLESS`	Meaningless or junk text
`FRAUD`	Online fraud and scams
`UNPOACH`	High-value user poaching attempts
`TEXTMINOR`	Content involving minors
`PROMPTATTACK`	Prompt injection and instruction-override attacks
`PUBLICFIGURE`	Public-figure recognition

`data` Object

Parameter	Type	Required	Description
`text`	string	Yes	The text to moderate. Up to 10,000 characters per request; longer inputs are rejected with an error. If `relateText` or `nickname` is also set, all three fields are moderated together in the same request.
`relateText`	string	No	Related context text moderated alongside `text` — for example, the parent comment when moderating a reply. Up to 128 characters; longer values are truncated.
`tokenId`	string	Yes	Stable identifier for the end user, typically your internal user UID (an encrypted UID is fine). Used for behavioral-risk signals such as spam and repeat-offender detection. Alphanumeric with underscores and hyphens, up to 64 characters.
`lang`	string	Yes	Language of the text. Pass `en` by default. For international traffic when the language is combined, pass `auto` to enable automatic language detection. See Supported Languages.
`nickname`	string	No	The end user's display name, moderated alongside `text`. Up to 150 characters; longer values are truncated.
`ip`	string	No	Public IP address of the user who submitted the text. Accepts IPv4 or IPv6.
`deviceId`	string	No	Device-fingerprint identifier issued by DeepCleer.
`dataId`	string	No	Client-side identifier attached to the moderation call. DeepCleer echoes it back with the result, letting you correlate your source record (a post ID, message ID, review ID, etc.) with the moderation verdict — typically used to look up historical decisions in your own database or in the DeepCleer console.
`extra`	json_object	No	Auxiliary parameters. See extra Object.

`extra` Object

Parameter	Type	Required	Description
`receiveTokenId`	string	No	`tokenId` of the message recipient in a one-to-one chat. Alphanumeric with underscores and hyphens, up to 64 characters. Required when `eventId` is `message`.
`topic`	string	No	Identifier for the conversation container — for example, a forum thread, blog post, or book-review section. Strongly recommended when `eventId` is `article`; without it, context-aware moderation cannot be applied.
`atId`	string	No	`tokenId` of the @‑mentioned user in a group-chat message. Alphanumeric with underscores and hyphens, up to 64 characters. Required when `eventId` is `groupChat`.
`room`	string	No	Live-room or game-room ID. Strongly recommended when `eventId` is `groupChat`; without it, context-aware moderation cannot be applied.
`sex`	int	No	User's gender. `0` male, `1` female, `2` unknown.
`passThrough`	json	No	Arbitrary pass-through data. The value is echoed back verbatim in the response, unchanged.

Supported Languages

Value	Language
`en`	English
`zh`	Chinese
`ar`	Arabic
`hi`	Hindi
`es`	Spanish
`fr`	French
`ru`	Russian
`pt`	Portuguese
`id`	Indonesian
`de`	German
`ja`	Japanese
`tr`	Turkish
`vi`	Vietnamese
`it`	Italian
`th`	Thai
`tl`	Filipino
`ko`	Korean
`ms`	Malay

Response

Response Parameters

ℹ️
Parameters other than code, message, and requestId are only guaranteed to be returned when code is 1100.

Parameter	Type	Required	Description
`code`	int	Yes	Response code. See Response Codes.
`message`	string	Yes	Response message corresponding to the `code`.
`requestId`	string	Yes	Unique identifier for this request. We strongly recommend logging it on your side — it's the key piece of information DeepCleer needs to help you investigate a specific moderation decision or tune your integration.
`riskLevel`	string	Yes	Disposition recommendation. `PASS` — content is clean; allow it. `REVIEW` — content is suspicious; better to route to human review. `REJECT` — content violates policy; block it.
`riskLabel1`	string	Yes	Level-1 risk label. Returns `normal` when `riskLevel` is `PASS`; otherwise returns a violation tag such as `POLITY`, `Erotic`.
`riskLabel2`	string	Yes	Level-2 risk label. Returns `normal` when `riskLevel` is `PASS`; otherwise returns a violation tag.
`riskLabel3`	string	Yes	Level-3 risk label. Returns `normal` when `riskLevel` is `PASS`; otherwise returns a violation tag.
`riskDescription`	string	Yes	Human-readable summary of the detected risk, formed by concatenating the three `riskLabel` levels (for example, `Advertising: Contact information`). Returns `Normal` when `riskLevel` is PASS.
`riskDetail`	json_object	Yes	Risk detail information. See riskDetail Object.
`tokenLabels`	json_object	Yes	Account risk profile label information.
`tokenLabels.UGC_account_risk`	json_object	No	UGC content risk.
`tokenLabels.UGC_account_risk.sexy_risk_tokenid`	float	No	Pornographic account risk score (0–1).
`auxInfo`	json_object	Yes	Auxiliary information. See `auxInfo` Object.
`allLabels`	json_array	Yes	All matched risk labels and details. See `allLabels` Array.
`businessLabels`	json_array	Yes	All matched business labels and details. See businessLabels Array.
`tokenProfileLabels`	json_array	No	Account attribute labels.
`tokenRiskLabels`	json_array	No	Account risk labels.
`langResult`	json_object	No	Language information. See langResult Object.
`finalResult`	int	Yes	Whether this is the final result. `1`: final result (can be used directly for downstream processing). `0`: intermediate result (machine moderation; human review will follow).
`resultType`	int	Yes	Current result stage. `0`: machine moderation. `1`: human moderation.

Response Codes

Code	Message
`1100`	Success
`1901`	QPS limit exceeded
`1902`	Invalid parameters
`1903`	Service failure
`1905`	Character count exceeded
`9101`	Unauthorized operation

`riskDetail` Object

Parameter	Type	Required	Description
`matchedLists`	json_array	No	Custom keyword lists that were triggered by this call. Each entry names a list you've configured in the DeepCleer console and enumerates the specific terms from that list that matched the input. Returned only when at least one configured keyword appears in `text`.
`matchedLists[].name`	string	No	Name of the triggered list, as configured in the DeepCleer console.
`matchedLists[].words`	json_array	No	The matched term, as it appeared on the list.
`matchedLists[].words[].word`	string	No	The matched term, as it appeared on the list.
`matchedLists[].words[].position`	int_array	No	Character offsets `[start, end]` of the match in the input text. Zero-indexed; `end` is exclusive.
`riskSegments`	json_array	No	Contiguous spans of the input judged to be high-risk. Returned when political, terrorism, prohibited, or advertising-law content is detected.
`riskSegments[].segment`	string	No	The high-risk span, extracted verbatim from the input.
`riskSegments[].position`	int_array	No	Character offsets `[start, end]` of the span in the input text. Zero-indexed; `end` is exclusive.

`auxInfo` Object

Parameter	Type	Required	Description
`filteredText`	string	No	Text with risky segments replaced by asterisks (`*`).
`passThrough`	json_object	No	Pass-through field. Same value as `data.extra.passThrough` in the request.
`contactResult`	json_array	No	Contact information detection results, including WeChat, QQ, and phone numbers.
`contactResult[].contactType`	int	No	Contact type. `0`: phone number. `1`: QQ number. `2`: WeChat ID. `3`: other.
`contactResult[].contactString`	string	No	Detected contact information string.
`contextText`	string	No	Returned when context recognition is enabled.
`unauthorizedType`	string	No	Moderation types that are not enabled. Contact DeepCleer to enable this feature.

`allLabels` Array

Each element in the array:

Parameter	Type	Required	Description
`riskLabel1`	string	Yes	Level 1 risk label.
`riskLabel2`	string	Yes	Level 2 risk label.
`riskLabel3`	string	Yes	Level 3 risk label.
`riskDescription`	string	Yes	Risk description.
`riskLevel`	string	Yes	Risk level: `PASS`, `REVIEW` or `REJECT`.
`probability`	float	Yes	Confidence score (0–1). Higher values indicate greater confidence.
`riskDetail`	json_object	Yes	Risk detail with same structure as riskDetail Object.

`businessLabels` Array

Each element in the array:

Parameter	Type	Required	Description
`businessLabel1`	string	Yes	Level 1 business label.
`businessLabel2`	string	Yes	Level 2 business label.
`businessLabel3`	string	Yes	Level 3 business label.
`businessDescription`	string	Yes	Business label description.
`probability`	float	Yes	Confidence score (0–1).
`businessDetail`	json_object	Yes	Business detail information.

Token Labels

Both tokenProfileLabels and tokenRiskLabels share the same structure:

Parameter	Type	Required	Description
`label1`	string	No	Level 1 label.
`label2`	string	No	Level 2 label.
`label3`	string	No	Level 3 label.
`description`	string	No	Label description.
`timestamp`	int	No	Label timestamp. 13-digit Unix timestamp in milliseconds.

`langResult` Object

Parameter	Type	Required	Description
`detectedLang`	string	No	Language detection result. Returned when `lang` is set to `auto` on an international text product. Value is a standard language code (e.g., `en`, `ar`).
`translatedText`	string	No	Translation result. Returned when `translationTargetLang` is provided in the request.

Risk Label Reference

Level 1 Label	Identifier	Type	Required `type` Value
Political	`Politics`	Regulatory	`TEXTRISK`
Violence & Terrorism	`Violence`	Regulatory	`TEXTRISK`
Pornographic	`Erotic`	Regulatory	`TEXTRISK`
Prohibited	`Prohibit`	Regulatory	`TEXTRISK`
Abusive	`Abuse`	Regulatory	`TEXTRISK`
Advertising	`Ads`	Regulatory	`TEXTRISK`
Blocklist	`Blacklist`	Regulatory	`TEXTRISK`

Examples

Request Example

{
  "accessKey": "YOUR_ACCESS_KEY",
  "appId": "default",
  "eventId": "text",
  "type": "TEXTRISK",
  "data": {
    "text": "A biological father and daughter engaged in a sexual act.",
    "tokenId": "user_ID_001",
    "ip": "118.89.214.89",
    "deviceId": "device_fingerprint_id",
    "nickname": "user_nickname",
    "extra": {
      "topic": "12345",
      "atId": "username1",
      "room": "test_room_001",
      "receiveTokenId": "user_ID_002"
    }
  }
}

Response Example

{
    "allLabels": [
        {
            "probability": 0.851256158682228,
            "riskDescription": "Erotic:EroticDescription:IncestDescription:IncestOverview",
            "riskDetail": {},
            "riskLabel1": "Erotic",
            "riskLabel2": "EroticDescription",
            "riskLabel3": "IncestDescription",
            "riskLabel4": "IncestOverview",
            "riskLevel": "REJECT"
        }
    ],
    "auxInfo": {
        "filteredText": "*********************************************************"
    },
    "businessLabels": [],
    "code": 1100,
    "finalResult": 1,
    "langResult": {
        "detectedLang": "en"
    },
    "message": "success",
    "requestId": "41b2a64ef2c0bf47ecc10c615ed4dcb0",
    "resultType": 0,
    "riskDescription": "Erotic:EroticDescription:IncestDescription:IncestOverview",
    "riskDetail": {},
    "riskLabel1": "Erotic",
    "riskLabel2": "EroticDescription",
    "riskLabel3": "IncestDescription",

Requirements

Request

Request URL

Request Parameters

Detection Types

data Object

extra Object

Supported Languages

Response

Response Parameters

Response Codes

riskDetail Object

auxInfo Object

allLabels Array

businessLabels Array

Token Labels

langResult Object

Risk Label Reference

Examples

Request Example

Response Example

`data` Object

`extra` Object

`riskDetail` Object

`auxInfo` Object

`allLabels` Array

`businessLabels` Array

`langResult` Object