> ## Documentation Index > Fetch the complete documentation index at: https://apidoc.cometapi.com/llms.txt > Use this file to discover all available pages before exploring further. # Tạo một moderation > Sử dụng CometAPI POST /v1/moderations để kiểm tra văn bản hoặc đầu vào multimodal bằng một yêu cầu moderation tương thích OpenAI. Sử dụng endpoint này để kiểm tra nội dung do người dùng tạo trước khi bạn gửi nó đến một endpoint model. Gửi một yêu cầu moderation tương thích OpenAI với giá trị `model` và `input`. Sử dụng khóa API CometAPI trong bearer header: `Authorization: Bearer $COMETAPI_KEY`. ## Nội dung request | Field | Type | Required | Description | | ------- | ----------------- | -------: | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `input` | string hoặc array | Yes | Nội dung cần kiểm tra. Dùng một string cho một đầu vào văn bản, một mảng string để kiểm tra văn bản theo lô, hoặc các phần multimodal kiểu OpenAI như văn bản cộng với `image_url`. | | `model` | string | Yes | model ID moderation. Sử dụng `omni-moderation-latest` cho moderation văn bản và hình ảnh trừ khi bạn có yêu cầu cụ thể về model moderation. | Đối với moderation multimodal, hãy gửi đầu vào multimodal kiểu OpenAI với một model hỗ trợ nó, chẳng hạn như `omni-moderation-latest`. URL hình ảnh công khai phải có thể được máy chủ CometAPI tải xuống. Để thử nghiệm hình ảnh có thể sao chép, hãy sử dụng một URL dữ liệu base64. ## Ví dụ request Gửi một chuỗi văn bản duy nhất khi bạn chỉ cần phân loại một đầu vào văn bản: ```bash theme={null} curl https://api.cometapi.com/v1/moderations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": "I want to bake cookies for my family." }' ``` Gửi một mảng chuỗi khi bạn muốn kiểm tra nhiều đầu vào văn bản trong một request: ```bash theme={null} curl https://api.cometapi.com/v1/moderations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ "I want to bake cookies.", "I want to kill someone." ] }' ``` Gửi văn bản cộng với một URL hình ảnh khi đầu vào moderation của bạn bao gồm một hình ảnh mà CometAPI có thể tải xuống: ```bash theme={null} curl https://api.cometapi.com/v1/moderations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ { "type": "text", "text": "...text to classify goes here..." }, { "type": "image_url", "image_url": { "url": "https://www.gstatic.com/webp/gallery/1.png" } } ] }' ``` Sử dụng một URL dữ liệu base64 khi bạn cần một request hình ảnh tự chứa: ```bash theme={null} curl https://api.cometapi.com/v1/moderations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ { "type": "text", "text": "...text to classify goes here..." }, { "type": "image_url", "image_url": { "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAAAAAA6fptVAAAADElEQVR4nGP4//8/AAX+Av4N70a4AAAAAElFTkSuQmCC" } } ] }' ``` ## Cấu trúc phản hồi Phản hồi bao gồm `id`, `model`, `results`, và `usage`. Mỗi mục trong `results` cho biết đầu vào tương ứng có bị gắn cờ hay không, các giá trị boolean của danh mục, điểm số danh mục, và các loại đầu vào được áp dụng cho từng danh mục. Với một request văn bản theo lô, `results` chứa một mục cho mỗi chuỗi đầu vào. Sử dụng `usage` cho các trường tính phí và giám sát. ## OpenAPI ````yaml api/openapi/content-moderation/create-moderation.openapi.json POST /v1/moderations openapi: 3.1.0 info: title: Create Moderation version: 1.0.0 description: Check content with CometAPI's OpenAI-compatible moderation endpoint. servers: - url: https://api.cometapi.com security: - bearerAuth: [] paths: /v1/moderations: post: summary: Create a moderation description: >- Checks text or multimodal input with a moderation model. The response is OpenAI-compatible and returns one `results` item for each moderated input. operationId: createModeration requestBody: required: true content: application/json: schema: type: object required: - model - input properties: model: type: string description: >- Required moderation model ID. Use `omni-moderation-latest` for text and image moderation unless you have a specific moderation model requirement. example: omni-moderation-latest input: description: >- Content to check. Use a string for one text input, an array of strings for batch text checks, or OpenAI-style multimodal parts such as text plus `image_url` when the selected model supports it. oneOf: - type: string description: One text input to classify. example: ...text to classify goes here... - type: array description: Batch text inputs to classify in one request. items: type: string example: - First text to classify. - Second text to classify. - type: array description: >- OpenAI-style multimodal input parts. Use a text part with an image_url part for image moderation. Public image URLs must be downloadable by CometAPI servers; use a base64 data URL for a self-contained image request. items: oneOf: - type: object required: - type - text properties: type: type: string enum: - text description: Text input part. text: type: string description: Text to classify. additionalProperties: false - type: object required: - type - image_url properties: type: type: string enum: - image_url description: Image input part. image_url: type: object required: - url properties: url: type: string description: >- Publicly downloadable image URL or base64 data URL. additionalProperties: true description: >- Image container for image input. Provide `url` with an HTTPS URL or a base64 data URL. additionalProperties: false example: - type: text text: ...text to classify goes here... - type: image_url image_url: url: https://www.gstatic.com/webp/gallery/1.png example: I want to check this text before sending it to a model. examples: Text: summary: Text input value: model: omni-moderation-latest input: I want to bake cookies for my family. Batch text: summary: Batch text input value: model: omni-moderation-latest input: - I want to bake cookies. - Plan a birthday party for my friend. Image URL: summary: Text plus image URL description: Use an image URL that CometAPI servers can download. value: model: omni-moderation-latest input: - type: text text: ...text to classify goes here... - type: image_url image_url: url: https://www.gstatic.com/webp/gallery/1.png Image data URL: summary: Text plus image data URL description: Use a base64 data URL for a self-contained image request. value: model: omni-moderation-latest input: - type: text text: ...text to classify goes here... - type: image_url image_url: url: >- data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAAAAAA6fptVAAAADElEQVR4nGP4//8/AAX+Av4N70a4AAAAAElFTkSuQmCC responses: '200': description: Moderation decisions and token usage. content: application/json: schema: $ref: '#/components/schemas/CreateModerationResponse' example: id: modr-1594 model: omni-moderation-latest results: - flagged: false categories: harassment: false harassment/threatening: false hate: false hate/threatening: false illicit: false illicit/violent: false self-harm: false self-harm/intent: false self-harm/instructions: false sexual: false sexual/minors: false violence: false violence/graphic: false category_scores: harassment: 0.0001 harassment/threatening: 0.0001 hate: 0.0001 hate/threatening: 0.0001 illicit: 0.0001 illicit/violent: 0.0001 self-harm: 0.0001 self-harm/intent: 0.0001 self-harm/instructions: 0.0001 sexual: 0.0001 sexual/minors: 0.0001 violence: 0.0001 violence/graphic: 0.0001 category_applied_input_types: harassment: - text harassment/threatening: - text hate: - text hate/threatening: - text illicit: - text illicit/violent: - text self-harm: - text self-harm/intent: - text self-harm/instructions: - text sexual: - text sexual/minors: - text violence: - text violence/graphic: - text usage: prompt_tokens: 12 completion_tokens: 0 total_tokens: 12 input_tokens: 12 output_tokens: 0 prompt_tokens_details: cached_tokens: 0 completion_tokens_details: reasoning_tokens: 0 input_tokens_details: null claude_cache_creation_1_h_tokens: 0 claude_cache_creation_5_m_tokens: 0 '400': description: The request body is invalid or missing required input. '401': description: The API key is missing or invalid. '503': description: >- The selected moderation model is unavailable for the current account. x-codeSamples: - lang: Shell label: Text source: | curl https://api.cometapi.com/v1/moderations \ -X POST \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": "I want to bake cookies for my family." }' - lang: Shell label: Batch text source: | curl https://api.cometapi.com/v1/moderations \ -X POST \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ "I want to bake cookies.", "I want to kill someone." ] }' - lang: Shell label: Image URL source: | curl https://api.cometapi.com/v1/moderations \ -X POST \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ { "type": "text", "text": "...text to classify goes here..." }, { "type": "image_url", "image_url": { "url": "https://www.gstatic.com/webp/gallery/1.png" } } ] }' - lang: Shell label: Image data URL source: | curl https://api.cometapi.com/v1/moderations \ -X POST \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $COMETAPI_KEY" \ -d '{ "model": "omni-moderation-latest", "input": [ { "type": "text", "text": "...text to classify goes here..." }, { "type": "image_url", "image_url": { "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAAAAAA6fptVAAAADElEQVR4nGP4//8/AAX+Av4N70a4AAAAAElFTkSuQmCC" } } ] }' - lang: Python label: Text source: | import os from openai import OpenAI client = OpenAI( base_url="https://api.cometapi.com/v1", api_key=os.environ["COMETAPI_KEY"], ) response = client.moderations.create( model="omni-moderation-latest", input="I want to bake cookies for my family.", ) print(response) - lang: Python label: Image data URL source: | import os from openai import OpenAI client = OpenAI( base_url="https://api.cometapi.com/v1", api_key=os.environ["COMETAPI_KEY"], ) response = client.moderations.create( model="omni-moderation-latest", input=[ {"type": "text", "text": "...text to classify goes here..."}, { "type": "image_url", "image_url": { "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAAAAAA6fptVAAAADElEQVR4nGP4//8/AAX+Av4N70a4AAAAAElFTkSuQmCC", }, }, ], ) print(response) - lang: Python label: Batch text source: | import os from openai import OpenAI client = OpenAI( base_url="https://api.cometapi.com/v1", api_key=os.environ["COMETAPI_KEY"], ) response = client.moderations.create( model="omni-moderation-latest", input=[ "I want to bake cookies.", "I want to kill someone.", ], ) for result in response.results: print(result.flagged) - lang: Python label: Image URL source: | import os from openai import OpenAI client = OpenAI( base_url="https://api.cometapi.com/v1", api_key=os.environ["COMETAPI_KEY"], ) response = client.moderations.create( model="omni-moderation-latest", input=[ {"type": "text", "text": "...text to classify goes here..."}, { "type": "image_url", "image_url": { "url": "https://www.gstatic.com/webp/gallery/1.png", }, }, ], ) print(response) - lang: JavaScript label: Text source: | import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://api.cometapi.com/v1", apiKey: process.env.COMETAPI_KEY, }); const response = await client.moderations.create({ model: "omni-moderation-latest", input: "I want to bake cookies for my family.", }); console.log(response); - lang: JavaScript label: Image data URL source: | import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://api.cometapi.com/v1", apiKey: process.env.COMETAPI_KEY, }); const response = await client.moderations.create({ model: "omni-moderation-latest", input: [ { type: "text", text: "...text to classify goes here..." }, { type: "image_url", image_url: { url: "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAAAAAA6fptVAAAADElEQVR4nGP4//8/AAX+Av4N70a4AAAAAElFTkSuQmCC", }, }, ], }); console.log(response); - lang: JavaScript label: Batch text source: | import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://api.cometapi.com/v1", apiKey: process.env.COMETAPI_KEY, }); const response = await client.moderations.create({ model: "omni-moderation-latest", input: [ "I want to bake cookies.", "I want to kill someone.", ], }); for (const result of response.results) { console.log(result.flagged); } - lang: JavaScript label: Image URL source: | import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://api.cometapi.com/v1", apiKey: process.env.COMETAPI_KEY, }); const response = await client.moderations.create({ model: "omni-moderation-latest", input: [ { type: "text", text: "...text to classify goes here..." }, { type: "image_url", image_url: { url: "https://www.gstatic.com/webp/gallery/1.png", }, }, ], }); console.log(response); components: schemas: CreateModerationResponse: type: object required: - id - model - results - usage properties: id: type: string description: Moderation request ID. example: modr-1594 model: type: string description: Model used for moderation. example: omni-moderation-latest results: type: array description: >- Moderation decisions. For batch text input, this array contains one result per input string. items: $ref: '#/components/schemas/ModerationResult' usage: $ref: '#/components/schemas/ModerationUsage' description: Token usage for the moderation request. additionalProperties: true ModerationResult: type: object required: - flagged - categories - category_scores - category_applied_input_types properties: flagged: type: boolean description: Whether this input item was flagged by any moderation category. example: false categories: $ref: '#/components/schemas/ModerationCategories' description: Per-category boolean decision flags. category_scores: $ref: '#/components/schemas/ModerationCategoryScores' description: Per-category confidence scores between 0 and 1. category_applied_input_types: $ref: '#/components/schemas/ModerationCategoryAppliedInputTypes' description: >- For each category, the input types (`text`, `image`) that contributed to the decision. additionalProperties: true ModerationUsage: type: object description: >- Token usage for the moderation request. Additional provider accounting fields can be present. additionalProperties: true properties: prompt_tokens: type: integer description: Input tokens counted for the request. example: 12 completion_tokens: type: integer description: Output tokens counted for the request. example: 0 total_tokens: type: integer description: Total tokens counted for the request. example: 12 input_tokens: type: integer description: Input token count when reported separately. example: 12 output_tokens: type: integer description: Output token count when reported separately. example: 0 prompt_tokens_details: type: - object - 'null' description: Detailed prompt token accounting when provided. additionalProperties: true completion_tokens_details: type: - object - 'null' description: Detailed completion token accounting when provided. additionalProperties: true input_tokens_details: type: - object - 'null' description: Detailed input token accounting when provided. additionalProperties: true claude_cache_creation_1_h_tokens: type: integer description: >- One-hour cache creation tokens when reported by the provider accounting layer. example: 0 claude_cache_creation_5_m_tokens: type: integer description: >- Five-minute cache creation tokens when reported by the provider accounting layer. example: 0 ModerationCategories: type: object description: Boolean moderation category decisions for one input item. additionalProperties: false properties: harassment: type: boolean description: Whether the input matched the harassment moderation category. example: false harassment/threatening: type: boolean description: >- Whether the input matched the harassment/threatening moderation category. example: false hate: type: boolean description: Whether the input matched the hate moderation category. example: false hate/threatening: type: boolean description: Whether the input matched the hate/threatening moderation category. example: false illicit: type: boolean description: Whether the input matched the illicit moderation category. example: false illicit/violent: type: boolean description: Whether the input matched the illicit/violent moderation category. example: false self-harm: type: boolean description: Whether the input matched the self-harm moderation category. example: false self-harm/intent: type: boolean description: Whether the input matched the self-harm/intent moderation category. example: false self-harm/instructions: type: boolean description: >- Whether the input matched the self-harm/instructions moderation category. example: false sexual: type: boolean description: Whether the input matched the sexual moderation category. example: false sexual/minors: type: boolean description: Whether the input matched the sexual/minors moderation category. example: false violence: type: boolean description: Whether the input matched the violence moderation category. example: false violence/graphic: type: boolean description: Whether the input matched the violence/graphic moderation category. example: false ModerationCategoryScores: type: object description: Moderation category confidence scores for one input item. additionalProperties: false properties: harassment: type: number format: float description: Confidence score for the harassment moderation category. example: 0.0001 harassment/threatening: type: number format: float description: Confidence score for the harassment/threatening moderation category. example: 0.0001 hate: type: number format: float description: Confidence score for the hate moderation category. example: 0.0001 hate/threatening: type: number format: float description: Confidence score for the hate/threatening moderation category. example: 0.0001 illicit: type: number format: float description: Confidence score for the illicit moderation category. example: 0.0001 illicit/violent: type: number format: float description: Confidence score for the illicit/violent moderation category. example: 0.0001 self-harm: type: number format: float description: Confidence score for the self-harm moderation category. example: 0.0001 self-harm/intent: type: number format: float description: Confidence score for the self-harm/intent moderation category. example: 0.0001 self-harm/instructions: type: number format: float description: Confidence score for the self-harm/instructions moderation category. example: 0.0001 sexual: type: number format: float description: Confidence score for the sexual moderation category. example: 0.0001 sexual/minors: type: number format: float description: Confidence score for the sexual/minors moderation category. example: 0.0001 violence: type: number format: float description: Confidence score for the violence moderation category. example: 0.0001 violence/graphic: type: number format: float description: Confidence score for the violence/graphic moderation category. example: 0.0001 ModerationCategoryAppliedInputTypes: type: object description: Input part types considered for each moderation category. additionalProperties: type: array items: type: string properties: harassment: type: array description: Input part types that were evaluated for the harassment category. items: type: string enum: - text - image example: - text harassment/threatening: type: array description: >- Input part types that were evaluated for the harassment/threatening category. items: type: string enum: - text - image example: - text hate: type: array description: Input part types that were evaluated for the hate category. items: type: string enum: - text - image example: - text hate/threatening: type: array description: >- Input part types that were evaluated for the hate/threatening category. items: type: string enum: - text - image example: - text illicit: type: array description: Input part types that were evaluated for the illicit category. items: type: string enum: - text - image example: - text illicit/violent: type: array description: >- Input part types that were evaluated for the illicit/violent category. items: type: string enum: - text - image example: - text self-harm: type: array description: Input part types that were evaluated for the self-harm category. items: type: string enum: - text - image example: - text self-harm/intent: type: array description: >- Input part types that were evaluated for the self-harm/intent category. items: type: string enum: - text - image example: - text self-harm/instructions: type: array description: >- Input part types that were evaluated for the self-harm/instructions category. items: type: string enum: - text - image example: - text sexual: type: array description: Input part types that were evaluated for the sexual category. items: type: string enum: - text - image example: - text sexual/minors: type: array description: Input part types that were evaluated for the sexual/minors category. items: type: string enum: - text - image example: - text violence: type: array description: Input part types that were evaluated for the violence category. items: type: string enum: - text - image example: - text violence/graphic: type: array description: >- Input part types that were evaluated for the violence/graphic category. items: type: string enum: - text - image example: - text securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: CometAPI API key description: >- CometAPI API key used for model requests. Send it as `Authorization: Bearer $COMETAPI_KEY`. ````