Sail Messages API API

openapi: 3.1.0
info:
  title: Sail Batches API Messages API API
  version: '2026-02-18'
  description: Sail provides OpenAI-compatible Responses and Chat Completions endpoints, plus an Anthropic-compatible Messages endpoint. This reference documents the currently supported subset of fields.
servers:
- url: https://api.sailresearch.com/v1
security:
- BearerAuth: []
tags:
- name: Messages API
  description: Anthropic-compatible Messages API endpoints.
paths:
  /messages:
    post:
      operationId: createMessage
      tags:
      - Messages API
      summary: Create an Anthropic message
      description: Anthropic-compatible Messages endpoint supporting system prompts, tool calling, and streaming (SSE).
      parameters:
      - $ref: '#/components/parameters/AnthropicIdempotencyKey'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CreateMessageRequest'
            example:
              model: zai-org/GLM-5.2-FP8
              max_tokens: 300
              messages:
              - role: user
                content: Give me a short launch checklist for an API integration.
      responses:
        '200':
          description: 'Anthropic-compatible message response. Returns a single JSON object by default, or an Anthropic Server-Sent Events stream when stream: true.'
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/AnthropicMessageResponse'
            text/event-stream:
              schema:
                $ref: '#/components/schemas/AnthropicStreamEvent'
        '400':
          description: Invalid request or unsupported feature.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '401':
          description: Authentication error.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '408':
          description: Timed out waiting for completion.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '500':
          description: Server error.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
  /messages/count_tokens:
    post:
      operationId: countMessageTokens
      tags:
      - Messages API
      summary: Count tokens for an Anthropic message
      description: Counts the input tokens a create-message request would consume, without running the model.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CountMessageTokensRequest'
      responses:
        '200':
          description: Input token count.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/MessageTokenCountResponse'
        '400':
          description: Invalid request or unsupported feature.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '401':
          description: Authentication error.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '429':
          description: Tokenizer at capacity; retry with backoff (Retry-After header set).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '500':
          description: Server error.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
components:
  schemas:
    RequestMetadata:
      type: object
      description: Optional string metadata. completion_window controls scheduling; completion_webhook/webhook_token configure completion webhooks.
      properties:
        completion_window:
          type: string
          description: '[Completion window](/completion-windows) (i.e. latency tier) for the request. Support matrix is available on the [Pricing](/pricing) page. When omitted, defaults to `standard` if that window is supported for the model; otherwise `flex` for async requests when the model supports it, and `asap` in all other cases. See [default behavior](/completion-windows#default-behavior).'
          enum:
          - asap
          - priority
          - standard
          - flex
        completion_webhook:
          type: string
          format: uri
        webhook_token:
          type: string
      additionalProperties:
        type: string
    AnthropicToolResultBlock:
      type: object
      description: Result of a tool call, sent back in a user message.
      required:
      - type
      - tool_use_id
      - content
      properties:
        type:
          type: string
          enum:
          - tool_result
        tool_use_id:
          type: string
        content:
          oneOf:
          - type: string
          - type: array
            items:
              oneOf:
              - $ref: '#/components/schemas/AnthropicTextBlock'
              - $ref: '#/components/schemas/AnthropicImageBlock'
        is_error:
          type: boolean
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    AnthropicToolUseBlock:
      type: object
      description: Tool call emitted by the model. Return its outcome as a tool_result block in a follow-up user message.
      required:
      - type
      - id
      - name
      - input
      properties:
        type:
          type: string
          enum:
          - tool_use
        id:
          type: string
        name:
          type: string
        input:
          type: object
          additionalProperties: true
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    CountMessageTokensRequest:
      type: object
      required:
      - model
      - messages
      description: Same shape as a create-message request; max_tokens is optional and ignored for counting.
      properties:
        model:
          type: string
        max_tokens:
          type: integer
          minimum: 1
          description: Optional here; validated if present, not used for counting.
        messages:
          type: array
          minItems: 1
          items:
            $ref: '#/components/schemas/AnthropicMessageParam'
        system:
          description: 'System prompt: a string, or an array of text blocks.'
          oneOf:
          - type: string
          - type: array
            items:
              $ref: '#/components/schemas/AnthropicTextBlock'
        temperature:
          type: number
          minimum: 0
          maximum: 1
        top_p:
          type: number
          minimum: 0
          maximum: 1
        output_config:
          $ref: '#/components/schemas/AnthropicOutputConfig'
        tools:
          type: array
          items:
            type: object
            additionalProperties: true
          description: Tool definitions the model may call. Responses include `tool_use` blocks; return outcomes as `tool_result` content blocks in a follow-up message.
        tool_choice:
          type: object
          additionalProperties: true
          description: 'Controls tool use: `auto`, `any`, `tool`, or `none`.'
        thinking:
          type: object
          additionalProperties: true
          description: Extended thinking configuration; translated to the model's reasoning.
        metadata:
          $ref: '#/components/schemas/RequestMetadata'
        context_management:
          type: object
          additionalProperties: true
          description: Anthropic server-side context-editing config (Claude Code sends it automatically). Accepted for compatibility but ignored — Sail forwards the full context each turn.
      additionalProperties: false
    AnthropicRedactedThinkingBlock:
      type: object
      description: Redacted extended-thinking block. Accepted when replayed; the opaque encrypted reasoning is not used.
      required:
      - type
      - data
      properties:
        type:
          type: string
          enum:
          - redacted_thinking
        data:
          type: string
          description: Opaque, encrypted reasoning payload.
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    CreateMessageRequest:
      type: object
      required:
      - model
      - max_tokens
      - messages
      properties:
        model:
          type: string
        max_tokens:
          type: integer
          minimum: 1
        messages:
          type: array
          minItems: 1
          items:
            $ref: '#/components/schemas/AnthropicMessageParam'
        system:
          description: 'System prompt: a string, or an array of text blocks.'
          oneOf:
          - type: string
          - type: array
            items:
              $ref: '#/components/schemas/AnthropicTextBlock'
        temperature:
          type: number
          minimum: 0
          maximum: 1
        top_p:
          type: number
          minimum: 0
          maximum: 1
        output_config:
          $ref: '#/components/schemas/AnthropicOutputConfig'
        tools:
          type: array
          items:
            type: object
            additionalProperties: true
          description: Tool definitions the model may call. Responses include `tool_use` blocks; return outcomes as `tool_result` content blocks in a follow-up message.
        tool_choice:
          type: object
          additionalProperties: true
          description: 'Controls tool use: `auto`, `any`, `tool`, or `none`.'
        thinking:
          type: object
          additionalProperties: true
          description: Extended thinking configuration; translated to the model's reasoning.
        stream:
          type: boolean
          description: When `true`, the response streams as Anthropic Server-Sent Events (`message_start`, `content_block_delta`, `message_stop`, …).
        metadata:
          $ref: '#/components/schemas/RequestMetadata'
        context_management:
          type: object
          additionalProperties: true
          description: Anthropic server-side context-editing config (Claude Code sends it automatically). Accepted for compatibility but ignored — Sail forwards the full context each turn.
      additionalProperties: false
    AnthropicImageBlock:
      type: object
      description: Image content block. Image input is supported only on multimodal models; see the Models page.
      required:
      - type
      - source
      properties:
        type:
          type: string
          enum:
          - image
        source:
          oneOf:
          - type: object
            required:
            - type
            - media_type
            - data
            properties:
              type:
                type: string
                enum:
                - base64
              media_type:
                type: string
                description: Image MIME type, e.g. image/jpeg, image/png, image/webp, image/gif.
              data:
                type: string
                description: 'Base64-encoded image bytes (no data: prefix).'
            additionalProperties: false
          - type: object
            required:
            - type
            - url
            properties:
              type:
                type: string
                enum:
                - url
              url:
                type: string
                description: Public http(s) URL of the image.
            additionalProperties: false
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    AnthropicUsage:
      type: object
      required:
      - input_tokens
      - output_tokens
      properties:
        input_tokens:
          type: integer
        output_tokens:
          type: integer
      additionalProperties: false
    ErrorObject:
      type: object
      required:
      - message
      - type
      properties:
        message:
          type: string
        type:
          type: string
        param:
          oneOf:
          - type: string
          - type: 'null'
        code:
          oneOf:
          - type: string
          - type: integer
          - type: 'null'
      additionalProperties: true
    AnthropicTextBlock:
      type: object
      required:
      - type
      - text
      properties:
        type:
          type: string
          enum:
          - text
        text:
          type: string
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    MessageTokenCountResponse:
      type: object
      required:
      - input_tokens
      properties:
        input_tokens:
          type: integer
      additionalProperties: false
    AnthropicMessageResponse:
      type: object
      required:
      - id
      - type
      - role
      - content
      - model
      - stop_reason
      - stop_sequence
      - usage
      properties:
        id:
          type: string
        type:
          type: string
          enum:
          - message
        role:
          type: string
          enum:
          - assistant
        content:
          type: array
          minItems: 1
          items:
            oneOf:
            - $ref: '#/components/schemas/AnthropicTextBlock'
            - $ref: '#/components/schemas/AnthropicToolUseBlock'
            - $ref: '#/components/schemas/AnthropicThinkingBlock'
        model:
          type: string
        stop_reason:
          type: string
          enum:
          - end_turn
          - max_tokens
          - tool_use
          - stop_sequence
          - refusal
        stop_sequence:
          oneOf:
          - type: string
          - type: 'null'
        usage:
          $ref: '#/components/schemas/AnthropicUsage'
      additionalProperties: false
    ErrorResponse:
      type: object
      required:
      - error
      properties:
        error:
          $ref: '#/components/schemas/ErrorObject'
      additionalProperties: false
    AnthropicStreamEvent:
      type: object
      description: 'One Server-Sent Event emitted when stream: true — the event''s JSON data payload; the SSE `event:` line carries the same `type`.'
      required:
      - type
      properties:
        type:
          type: string
          enum:
          - message_start
          - content_block_start
          - content_block_delta
          - content_block_stop
          - message_delta
          - message_stop
          - ping
          - error
      additionalProperties: true
    AnthropicThinkingBlock:
      type: object
      description: Extended-thinking reasoning block. When continuing a thinking conversation, replay it unchanged (including its signature).
      required:
      - type
      - thinking
      properties:
        type:
          type: string
          enum:
          - thinking
        thinking:
          type: string
        signature:
          type: string
          description: Opaque signature returned with the block; replay unchanged.
        cache_control:
          $ref: '#/components/schemas/AnthropicCacheControl'
      additionalProperties: false
    AnthropicOutputConfigFormat:
      type: object
      required:
      - type
      - schema
      properties:
        type:
          type: string
          enum:
          - json_schema
        schema:
          type: object
          additionalProperties: true
        name:
          type: string
        strict:
          type: boolean
      additionalProperties: false
    AnthropicMessageParam:
      type: object
      required:
      - role
      - content
      properties:
        role:
          type: string
          enum:
          - user
          - assistant
          - system
          - developer
          description: user/assistant per Anthropic; system and developer are also accepted (Claude Code injects mid-conversation system/developer messages).
        content:
          oneOf:
          - type: string
          - type: array
            minItems: 1
            items:
              oneOf:
              - $ref: '#/components/schemas/AnthropicTextBlock'
              - $ref: '#/components/schemas/AnthropicImageBlock'
              - $ref: '#/components/schemas/AnthropicToolUseBlock'
              - $ref: '#/components/schemas/AnthropicToolResultBlock'
              - $ref: '#/components/schemas/AnthropicThinkingBlock'
              - $ref: '#/components/schemas/AnthropicRedactedThinkingBlock'
      additionalProperties: false
    AnthropicOutputConfig:
      type: object
      properties:
        format:
          $ref: '#/components/schemas/AnthropicOutputConfigFormat'
        effort:
          type: string
          enum:
          - none
          - minimal
          - low
          - medium
          - high
          - xhigh
          - max
          description: Reasoning effort for the model (`max` is treated as `xhigh`). Overrides any effort derived from `thinking`.
      additionalProperties: false
    AnthropicCacheControl:
      type: object
      description: 'Prompt-cache hint (e.g. { "type": "ephemeral" }). Accepted on content blocks for Anthropic SDK compatibility but ignored — no cache read/write; see the Prompt caching note.'
      properties:
        type:
          type: string
          enum:
          - ephemeral
      additionalProperties: true
  parameters:
    AnthropicIdempotencyKey:
      name: anthropic-idempotency-key
      in: header
      required: false
      schema:
        type: string
        maxLength: 255
      description: Anthropic-compatible idempotency header. Same semantics as Idempotency-Key; used on the /messages endpoint to match Anthropic's conventions.
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      bearerFormat: API Key