openapi: 3.1.0
info:
  title: Inception Chat API
  version: 1.0.0
  description: Inception Labs LLM API — chat, FIM, and edit completions powered by Mercury diffusion language models. OpenAI-compatible request/response shapes with extensions for diffusion-specific features.
  contact:
    name: Inception Labs
    url: https://docs.inceptionlabs.ai
    email: support@inceptionlabs.ai
  license:
    name: Proprietary
  x-logo:
    url: https://docs.inceptionlabs.ai/logo.png
servers:
- url: https://api.inceptionlabs.ai
  description: Production
security:
- BearerAuth: []
tags:
- name: Chat
  description: Chat completion endpoints (OpenAI-compatible).
paths:
  /v1/chat/completions:
    post:
      summary: Create a chat completion
      operationId: createChatCompletion
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ChatCompletionRequest'
            example:
              model: mercury-2
              messages:
              - role: user
                content: What is a diffusion language model?
              max_tokens: 256
              temperature: 0.75
        required: true
      responses:
        '200':
          description: 'Successful response. Returns a JSON object when `stream=false`, or a server-sent events stream of `ChatCompletionChunk` objects (terminated by `data: [DONE]`) when `stream=true`.'
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ChatCompletionResponse'
              example:
                id: chatcmpl-7a2b3c4d5e
                object: chat.completion
                created: 1745798400
                model: mercury-2
                choices:
                - index: 0
                  finish_reason: stop
                  message:
                    role: assistant
                    content: A diffusion language model is a type of language model that uses diffusion to generate text.
                usage:
                  prompt_tokens: 12
                  completion_tokens: 8
                  total_tokens: 20
                  reasoning_tokens: 0
                  cached_input_tokens: 0
            text/event-stream:
              schema:
                $ref: '#/components/schemas/ChatCompletionChunk'
              x-stainless-stream-type: sse
              x-stainless-stream-event-schema:
                $ref: '#/components/schemas/ChatCompletionChunk'
              x-stainless-stream-terminator: '[DONE]'
        '400':
          description: Bad Request — invalid parameters.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: You exceeded the maximum context length for this model of 128000. Please reduce the length of the messages or completion.
                  type: invalid_request_error
                  param: messages
                  code: context_length_exceeded
        '401':
          description: Unauthorized — missing or invalid API key.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: Incorrect API key provided
                  type: authentication_error
                  param: null
                  code: invalid_api_key
        '402':
          description: Payment Required — billing inactive or quota exceeded.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: Account is inactive
                  type: account_error
                  param: null
                  code: account_error
        '404':
          description: Not Found — model not available for this account.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: model `jupyter-2` not found
                  type: invalid_request_error
                  param: model
                  code: model_not_found
        '429':
          description: Too Many Requests — rate limit exceeded.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: Rate limit exceeded. Please try again later.
                  type: rate_limit_error
                  param: null
                  code: rate_limit_reached
        '500':
          description: Internal Server Error.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
              example:
                error:
                  message: The server had an error while processing your request.
                  type: server_error
                  param: null
                  code: server_error
      security:
      - BearerAuth: []
      tags:
      - Chat
      description: Generate a chat completion from a sequence of messages. Returns a `ChatCompletion` object, or a server-sent events stream of `ChatCompletionChunk` deltas when `stream=true`. Supports tool calling, structured output via `response_format`, and reasoning controls.
      x-codeSamples:
      - lang: python
        label: Python
        source: "import os\nfrom inceptionai import Inception\n\nclient = Inception(\n    api_key=os.environ.get(\"INCEPTION_API_KEY\"),  # defaults to this env var; can be omitted\n)\n\nchat_completion = client.chat.completions.create(\n    model=\"mercury-2\",\n    messages=[\n        {\n            \"role\": \"user\",\n            \"content\": \"What is a diffusion language model?\"\n        }\n    ],\n    max_tokens=256,\n    temperature=0.75,\n)\nprint(chat_completion)"
      - lang: typescript
        label: TypeScript
        source: "import Inception from 'inceptionai';\n\nconst client = new Inception({\n  apiKey: process.env['INCEPTION_API_KEY'], // defaults to this env var; can be omitted\n});\n\nconst chatCompletion = await client.chat.completions.create({\n  model: 'mercury-2',\n  messages: [\n    {\n      role: 'user',\n      content: 'What is a diffusion language model?'\n    }\n  ],\n  max_tokens: 256,\n  temperature: 0.75\n});\nconsole.log(chatCompletion);"
components:
  schemas:
    ChatCompletionTool:
      type: object
      title: ChatCompletionTool
      required:
      - type
      - function
      properties:
        type:
          type: string
          const: function
          description: The type of the tool.
        function:
          $ref: '#/components/schemas/FunctionDefinition'
    ReasoningSummary:
      description: Summary of the model's reasoning process.
      properties:
        content:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          description: The summarized reasoning.
          title: Content
        status:
          default: unavailable
          description: Status of the summary.
          enum:
          - complete
          - unavailable
          - skipped
          title: Status
          type: string
      title: ReasoningSummary
      type: object
      required: []
    ChunkDelta:
      description: Represents a partial message in a streaming response chunk.
      properties:
        content:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          description: The partial content of the message.
          title: Content
        role:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          description: The role of the message sender, typically only in the first chunk.
          title: Role
        tool_calls:
          anyOf:
          - items:
              $ref: '#/components/schemas/ToolCallDelta'
            type: array
          - type: 'null'
          default: null
          description: The tool calls generated by the model in this chunk.
          title: Tool Calls
      title: ChunkDelta
      type: object
      required: []
    ChatCompletionResponse:
      properties:
        id:
          type: string
          title: Id
        object:
          type: string
          const: chat.completion
          title: Object
          default: chat.completion
        created:
          type: integer
          title: Created
        model:
          type: string
          title: Model
        choices:
          items:
            $ref: '#/components/schemas/Choice'
          type: array
          title: Choices
        usage:
          $ref: '#/components/schemas/Usage'
        warning:
          anyOf:
          - type: string
          - type: 'null'
          title: Warning
        reasoning_summary:
          anyOf:
          - $ref: '#/components/schemas/ReasoningSummary'
          - type: 'null'
          description: Summary of reasoning process if requested and available.
      type: object
      required:
      - id
      - created
      - model
      - choices
      - usage
      - object
      title: ChatCompletionResponse
      example:
        id: chatcmpl-7a2b3c4d5e
        object: chat.completion
        created: 1745798400
        model: mercury-2
        choices:
        - index: 0
          finish_reason: stop
          message:
            role: assistant
            content: A diffusion language model is a type of language model that uses diffusion to generate text.
        usage:
          prompt_tokens: 12
          completion_tokens: 8
          total_tokens: 20
          reasoning_tokens: 0
          cached_input_tokens: 0
    FunctionCall:
      properties:
        name:
          anyOf:
          - type: string
          - type: 'null'
          title: Name
        arguments:
          type: string
          title: Arguments
      type: object
      required:
      - arguments
      title: FunctionCall
    FunctionDefinition:
      type: object
      title: FunctionDefinition
      required:
      - name
      properties:
        name:
          type: string
          description: The name of the function to call. Must be a-z, A-Z, 0-9, or contain underscores and dashes, with a maximum length of 64.
        description:
          type: string
          description: A description of what the function does, used by the model to choose when and how to call the function.
        parameters:
          type: object
          additionalProperties: true
          description: The parameters the function accepts, described as a JSON Schema object.
        strict:
          type: boolean
          default: false
          description: Whether to enforce strict schema adherence when generating the function call. When true, the model output is constrained to match the `parameters` schema exactly.
    ErrorResponse:
      type: object
      title: ErrorResponse
      required:
      - error
      properties:
        error:
          $ref: '#/components/schemas/ErrorObject'
      example:
        error:
          message: You exceeded the maximum context length for this model of 128000. Please reduce the length of the messages or completion.
          type: invalid_request_error
          param: messages
          code: context_length_exceeded
    JSONSchema:
      type: object
      title: JSONSchema
      required:
      - name
      properties:
        name:
          type: string
          description: The name of the response format. Must be a-z, A-Z, 0-9, or contain underscores and dashes, with a maximum length of 64.
        description:
          type: string
          description: A description of what the response format is for, used by the model to determine how to respond in the format.
        schema:
          type: object
          additionalProperties: true
          description: The schema for the response format, described as a JSON Schema object.
        strict:
          type: boolean
          default: false
          description: Whether to enforce strict schema adherence when generating the output. When true, the model output is constrained to match the supplied `schema` exactly.
    ResponseFormatJSONSchema:
      type: object
      title: ResponseFormatJSONSchema
      required:
      - type
      - json_schema
      properties:
        type:
          type: string
          const: json_schema
          description: The type of response format being defined. Always `json_schema`.
        json_schema:
          $ref: '#/components/schemas/JSONSchema'
    Message:
      properties:
        role:
          type: string
          enum:
          - system
          - user
          - assistant
          - tool
          - function
          title: Role
          description: The role of the message sender (system, user, assistant, tool, function).
        content:
          anyOf:
          - type: string
          - items:
              additionalProperties: true
              type: object
            type: array
          - type: 'null'
          title: Content
          description: The content of the message, can be string or array of content items.
        tool_calls:
          anyOf:
          - items:
              $ref: '#/components/schemas/ToolCall'
            type: array
          - type: 'null'
          title: Tool Calls
          description: The tool calls generated by the model.
        tool_call_id:
          anyOf:
          - type: string
          - type: 'null'
          title: Tool Call Id
          description: The ID of the tool call this message is responding to (for role='tool').
      type: object
      required:
      - role
      title: Message
    ChatCompletionRequest:
      properties:
        messages:
          items:
            $ref: '#/components/schemas/Message'
          type: array
          title: Messages
        model:
          type: string
          title: Model
          description: The model to use for the chat completion.
        max_tokens:
          type: integer
          title: Max Tokens
          description: Maximum number of tokens to generate.
          minimum: 1
          maximum: 50000
          default: 16384
        temperature:
          type: number
          title: Temperature
          description: What sampling temperature to use, between 0.5 and 1. Higher values make the output more random; lower values make it more focused and deterministic. Values outside this range are reset to 0.75 and a `warning` is returned in the response.
          default: 0.75
          minimum: 0.5
          maximum: 1.0
        stop:
          items:
            type: string
          type: array
          title: Stop
          description: A list of sequences where the API will stop generating further tokens. The returned text will not contain the stop sequences.
        tools:
          type: array
          title: Tools
          description: A list of tools the model may call. Use this to provide functions the model can generate JSON arguments for.
          items:
            $ref: '#/components/schemas/ChatCompletionTool'
        tool_choice:
          type: string
          enum:
          - auto
          - required
          - none
          title: Tool Choice
          description: 'Controls tool selection: ''auto'' (model decides, default when tools present), ''required'' (model must call one), or ''none'' (model must not call one).'
        stream:
          type: boolean
          title: Stream
          description: Whether to stream the response.
          default: false
        stream_options:
          $ref: '#/components/schemas/StreamOptions'
          description: Options that control streaming behavior.
        diffusing:
          type: boolean
          title: Diffusing
          description: Whether to show the diffusion effect in the streamed response.
          default: false
        extra_body:
          additionalProperties: true
          type: object
          title: Extra Body
          description: Extra parameters passed to the model (OpenAI-compat passthrough).
          x-stainless-param: extra_params
        realtime:
          type: boolean
          title: Realtime
          description: Enable flag for more realtime workloads that require lower TTFT/TTFAT.
          default: false
        response_format:
          title: Response Format
          description: An object specifying the format that the model must output.
          oneOf:
          - $ref: '#/components/schemas/ResponseFormatText'
          - $ref: '#/components/schemas/ResponseFormatJSONObject'
          - $ref: '#/components/schemas/ResponseFormatJSONSchema'
        reasoning_summary:
          type: boolean
          title: Reasoning Summary
          description: Request a summary of the model's reasoning process.
          default: false
        reasoning_summary_wait:
          type: boolean
          title: Reasoning Summary Wait
          description: Wait for all reasoning summaries to complete before finishing the response.
          default: false
        reasoning_effort:
          type: string
          title: Reasoning Effort
          description: Constrains the effort spent on reasoning before the model responds.
          enum:
          - instant
          - low
          - medium
          - high
          default: medium
      type: object
      required:
      - messages
      - model
      title: ChatCompletionRequest
    ResponseFormatText:
      type: object
      title: ResponseFormatText
      required:
      - type
      properties:
        type:
          type: string
          const: text
          description: The type of response format being defined. Always `text`. The model responds with plain text and output is not constrained to JSON.
    ChatCompletionChunk:
      description: Model for a single chunk in a streaming chat completion response.
      properties:
        id:
          title: Id
          type: string
        object:
          const: chat.completion.chunk
          default: chat.completion.chunk
          title: Object
          type: string
        created:
          title: Created
          type: integer
        model:
          title: Model
          type: string
        choices:
          items:
            $ref: '#/components/schemas/ChunkChoice'
          title: Choices
          type: array
        reasoning_summary:
          anyOf:
          - $ref: '#/components/schemas/ReasoningSummary'
          - type: 'null'
          default: null
          description: Summary of reasoning process (included in final chunk if available).
      required:
      - id
      - created
      - model
      - choices
      - object
      title: ChatCompletionChunk
      type: object
    ChunkChoice:
      description: Represents a choice in a streaming response chunk.
      properties:
        index:
          title: Index
          type: integer
        delta:
          $ref: '#/components/schemas/ChunkDelta'
        finish_reason:
          type:
          - string
          - 'null'
          enum:
          - stop
          - length
          - tool_calls
          - content_filter
          - null
          description: The reason the model stopped generating tokens.
      required:
      - index
      - delta
      title: ChunkChoice
      type: object
    ResponseFormatJSONObject:
      type: object
      title: ResponseFormatJSONObject
      required:
      - type
      properties:
        type:
          type: string
          const: json_object
          description: The type of response format being defined. Always `json_object`.
    Choice:
      properties:
        index:
          type: integer
          title: Index
        message:
          $ref: '#/components/schemas/Message'
        finish_reason:
          type:
          - string
          - 'null'
          enum:
          - stop
          - length
          - tool_calls
          - content_filter
          - null
          description: The reason the model stopped generating tokens.
      type: object
      required:
      - index
      - message
      - finish_reason
      title: Choice
    FunctionCallDelta:
      description: Streaming counterpart to FunctionCall; arguments arrives chunk-by-chunk and may be null.
      properties:
        name:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          title: Name
        arguments:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          title: Arguments
      title: FunctionCallDelta
      type: object
      required: []
    ToolCall:
      properties:
        index:
          anyOf:
          - type: integer
          - type: 'null'
          title: Index
          default: 0
        id:
          type: string
          title: Id
        type:
          type: string
          const: function
          title: Type
          default: function
        function:
          $ref: '#/components/schemas/FunctionCall'
      type: object
      required:
      - id
      - function
      - type
      title: ToolCall
    ErrorObject:
      type: object
      title: ErrorObject
      required:
      - message
      - type
      properties:
        message:
          type: string
          description: Human-readable error message.
        type:
          type: string
          description: Error category, e.g. `invalid_request_error`, `rate_limit_error`.
        param:
          type:
          - string
          - 'null'
          description: Offending parameter, if applicable.
        code:
          type:
          - string
          - 'null'
          description: Machine-readable error code.
    Usage:
      properties:
        prompt_tokens:
          type: integer
          title: Prompt Tokens
        reasoning_tokens:
          type: integer
          title: Reasoning Tokens
        completion_tokens:
          type: integer
          title: Completion Tokens
        total_tokens:
          type: integer
          title: Total Tokens
        cached_input_tokens:
          type: integer
          title: Cached Input Tokens
      type: object
      required:
      - prompt_tokens
      - reasoning_tokens
      - completion_tokens
      - total_tokens
      - cached_input_tokens
      title: Usage
    StreamOptions:
      properties:
        include_usage:
          type: boolean
          title: Include Usage
          description: 'If true, an additional chunk is streamed before the `data: [DONE]` message containing the token usage statistics for the entire request. Inside that chunk, `choices` is an empty array and the `usage` field is populated.'
          default: false
      type: object
      title: StreamOptions
      description: Options for controlling streaming behavior. Only used when `stream=true`.
      required: []
    ToolCallDelta:
      description: Streaming counterpart to ToolCall; per OpenAI SSE spec, id/type only appear in the first chunk per index.
      properties:
        index:
          anyOf:
          - type: integer
          - type: 'null'
          default: 0
          title: Index
        id:
          anyOf:
          - type: string
          - type: 'null'
          default: null
          title: Id
        type:
          anyOf:
          - const: function
            type: string
          - type: 'null'
          default: null
          title: Type
        function:
          anyOf:
          - $ref: '#/components/schemas/FunctionCallDelta'
          - type: 'null'
          default: null
      title: ToolCallDelta
      type: object
      required: []
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      description: 'API key provided as a Bearer token: `Authorization: Bearer <api_key>`. Get an API key at https://platform.inceptionlabs.ai.'