> ## Documentation Index > Fetch the complete documentation index at: https://sambanova-systems.mintlify.site/docs/llms.txt > Use this file to discover all available pages before exploring further. # Create a message > Anthropic Messages API compatible endpoint. Generates a model response for the supplied conversation. Authentication accepts either the bearer `Authorization: Bearer ` header (SambaNova SDK default) or the `x-api-key` header (Anthropic SDK default); the same API key is used in both cases. When `stream: true` is set, the response is a sequence of Server-Sent Events whose payloads conform to `MessageStreamEvent`; otherwise the response is a single `Message` object. ## OpenAPI ````yaml https://app.stainless.com/api/spec/documented/sambanova/openapi.documented.yml post /messages openapi: 3.1.1 info: title: SambaNova cloud API description: SambaNova cloud API Specification version: 1.2.0 termsOfService: https://sambanova.ai/cloud-end-user-license-agreement contact: email: info@sambanova.ai name: SambaNova information license: name: Apache 2.0 url: https://www.apache.org/licenses/LICENSE-2.0.html servers: - url: https://api.sambanova.ai/v1 security: - api_key: [] externalDocs: description: Find out more in the official SambaNova docs url: https://docs.sambanova.ai/docs/en/api-reference/overview paths: /messages: post: tags: - Messages summary: Create a message description: >- Anthropic Messages API compatible endpoint. Generates a model response for the supplied conversation. Authentication accepts either the bearer `Authorization: Bearer ` header (SambaNova SDK default) or the `x-api-key` header (Anthropic SDK default); the same API key is used in both cases. When `stream: true` is set, the response is a sequence of Server-Sent Events whose payloads conform to `MessageStreamEvent`; otherwise the response is a single `Message` object. operationId: createMessage parameters: - in: header name: anthropic-version required: false description: >- Anthropic API version header sent by the official `anthropic` SDK. Accepted (any value) but currently has no effect on response shape — included for drop-in SDK compatibility. schema: title: Anthropic Version type: string example: '2023-06-01' requestBody: required: true description: Message creation parameters. content: application/json: schema: $ref: '#/components/schemas/MessageCreateRequest' responses: '200': description: >- Successful response. Returns a `Message` object (non-streaming), or a stream of Server-Sent Events whose payloads conform to `MessageStreamEvent` ending with a `message_stop` event (when `stream: true`). content: application/json: schema: oneOf: - $ref: '#/components/schemas/Message' - $ref: '#/components/schemas/MessageStreamEvent' '400': description: >- Bad Request — `invalid_request_error`. Returned for missing or invalid parameters, unsupported feature flags (e.g. `thinking` with `type:"enabled"` or `"adaptive"`), unsupported tool types (`web_search`, `code_execution`, `bash`, `text_editor`, `memory`, `tool_search` variants), image source `type:"url"`, `document` content blocks, or non-object request bodies. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '401': description: Unauthorized — `authentication_error`. Invalid or missing API key. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '403': description: >- Forbidden — `permission_error`. API key lacks access to the requested resource. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '404': description: >- Not Found — `not_found_error`. Model does not exist or is not accessible. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '413': description: >- Payload Too Large — `request_too_large`. Request exceeds the maximum allowed size. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '429': description: Too Many Requests — `rate_limit_error`. Rate limit exceeded. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '500': description: >- Internal Server Error — `api_error`. Unexpected issue on the server side. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '501': description: >- Not Implemented — `not_implemented_error`. Currently returned when `stream: true` is set, until streaming support lands (PRD Phase 4). content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' '503': description: >- Service Unavailable — `overloaded_error`. Backend is temporarily over capacity. content: application/json: schema: $ref: '#/components/schemas/MessageErrorResponse' security: - api_key: [] - x_api_key: [] x-codeSamples: - lang: JavaScript source: |- import SambaNova from 'sambanova'; const client = new SambaNova({ apiKey: process.env['SAMBANOVA_API_KEY'], // This is the default and can be omitted }); const message = await client.messages.create({ max_tokens: 1024, messages: [{ content: 'Hello, Claude!', role: 'user' }], model: 'DeepSeek-V3.1', }); console.log(message); - lang: Python source: |- import os from sambanova import SambaNova client = SambaNova( api_key=os.environ.get("SAMBANOVA_API_KEY"), # This is the default and can be omitted ) for message in client.messages.create( max_tokens=1024, messages=[{ "content": "Hello, Claude!", "role": "user", }], model="DeepSeek-V3.1", ): print(message) components: schemas: MessageCreateRequest: title: Message Create Request type: object description: >- Request body for `POST /messages`. Wire-compatible with the official Anthropic Messages API. Required fields: `model`, `max_tokens`, `messages`. additionalProperties: true properties: model: title: Model description: >- The model ID to use (e.g. gpt-oss-120b). See available [models](https://docs.sambanova.ai/docs/en/models/sambacloud-models) anyOf: - type: string - enum: - Meta-Llama-3.3-70B-Instruct - Meta-Llama-3.2-1B-Instruct - Meta-Llama-3.2-3B-Instruct - Llama-3.2-11B-Vision-Instruct - Llama-3.2-90B-Vision-Instruct - Meta-Llama-3.1-8B-Instruct - Meta-Llama-3.1-70B-Instruct - Meta-Llama-3.1-405B-Instruct - Qwen2.5-Coder-32B-Instruct - Qwen2.5-72B-Instruct - QwQ-32B-Preview - Meta-Llama-Guard-3-8B - DeepSeek-R1 - DeepSeek-R1-0528 - DeepSeek-V3-0324 - DeepSeek-V3.1 - DeepSeek-V3.1-cb - DeepSeek-V3.1-Terminus - DeepSeek-V3.2 - DeepSeek-R1-Distill-Llama-70B - Llama-4-Maverick-17B-128E-Instruct - Llama-4-Scout-17B-16E-Instruct - Qwen3-32B - Qwen3-235B - Llama-3.3-Swallow-70B-Instruct-v0.4 - gpt-oss-120b - ALLaM-7B-Instruct-preview - MiniMax-M2.5 - MiniMax-M2.7 - gemma-3-12b-it example: gpt-oss-120b max_tokens: title: Max Tokens type: integer minimum: 1 description: >- Maximum number of tokens to generate. The combined input + output token count is bounded by the model's context window. example: 1024 messages: title: Messages type: array minItems: 1 items: $ref: '#/components/schemas/MessageInputMessage' description: Conversation turns. system: $ref: '#/components/schemas/MessageSystemPrompt' temperature: title: Temperature type: number minimum: 0 maximum: 2 nullable: true description: >- Sampling temperature in `[0.0, 2.0]`. Higher values produce more random output, lower values more deterministic. Adjust only one of `temperature`, `top_p`, `top_k`. example: 1 top_p: title: Top P type: number minimum: 0 maximum: 1 nullable: true description: >- Nucleus sampling. Considers tokens with cumulative probability mass up to `top_p`. top_k: title: Top K type: integer minimum: 0 nullable: true description: >- Top-k sampling. Considers only the K most likely tokens at each step. Set to 0 to disable. stop_sequences: title: Stop Sequences type: array items: type: string nullable: true description: Custom strings that, when generated, cause the model to stop. stream: title: Stream type: boolean default: false nullable: true description: >- If true, the response is a sequence of Server-Sent Events whose payloads conform to `MessageStreamEvent`. metadata: $ref: '#/components/schemas/MessageMetadata' thinking: $ref: '#/components/schemas/MessageThinkingConfig' tools: title: Tools type: array items: $ref: '#/components/schemas/MessageTool' nullable: true description: Tool definitions the model may call. tool_choice: allOf: - $ref: '#/components/schemas/MessageToolChoice' nullable: true description: Controls how the model selects from `tools`. service_tier: title: Service Tier type: string enum: - auto - standard_only nullable: true description: 'Service-tier preference. **In v1**: silently dropped' container: title: Container type: string nullable: true description: >- Existing code-execution container ID to reuse. **In v1**: silently dropped required: - model - max_tokens - messages example: model: DeepSeek-V3.1 max_tokens: 1024 messages: - role: user content: Hello, Claude! Message: title: Message type: object description: >- Non-streaming response from `POST /messages`. Wire-compatible with the official Anthropic Messages API. properties: id: type: string description: Unique identifier for this message. type: type: string enum: - message const: message role: type: string enum: - assistant const: assistant content: type: array items: $ref: '#/components/schemas/MessageOutputContentBlock' model: type: string description: Model that produced the response. stop_reason: allOf: - $ref: '#/components/schemas/MessageStopReason' nullable: true stop_sequence: type: string nullable: true description: >- The matched stop sequence that triggered termination. Present when `stop_reason` is `stop_sequence`; `null` otherwise. stop_details: allOf: - $ref: '#/components/schemas/MessageStopDetails' nullable: true usage: $ref: '#/components/schemas/MessageUsage' container: allOf: - $ref: '#/components/schemas/MessageContainer' nullable: true required: - id - type - role - content - model - stop_reason - usage MessageStreamEvent: title: Message Stream Event description: >- Discriminated union of SSE events emitted during a streaming `POST /messages` call. Event order: `message_start` -> (`content_block_start` -> `content_block_delta`×N -> `content_block_stop`) × blocks -> `message_delta` -> `message_stop`. A `ping` may appear at any point; an `error` event terminates the stream. oneOf: - $ref: '#/components/schemas/MessageStartEvent' - $ref: '#/components/schemas/MessageContentBlockStartEvent' - $ref: '#/components/schemas/MessageContentBlockDeltaEvent' - $ref: '#/components/schemas/MessageContentBlockStopEvent' - $ref: '#/components/schemas/MessageDeltaEvent' - $ref: '#/components/schemas/MessageStopEvent' - $ref: '#/components/schemas/MessagePingEvent' - $ref: '#/components/schemas/MessageStreamErrorEvent' discriminator: propertyName: type mapping: message_start: $ref: '#/components/schemas/MessageStartEvent' content_block_start: $ref: '#/components/schemas/MessageContentBlockStartEvent' content_block_delta: $ref: '#/components/schemas/MessageContentBlockDeltaEvent' content_block_stop: $ref: '#/components/schemas/MessageContentBlockStopEvent' message_delta: $ref: '#/components/schemas/MessageDeltaEvent' message_stop: $ref: '#/components/schemas/MessageStopEvent' ping: $ref: '#/components/schemas/MessagePingEvent' error: $ref: '#/components/schemas/MessageStreamErrorEvent' MessageErrorResponse: title: Message Error Response type: object description: >- Top-level error envelope returned by the Messages API on any non-2xx response. Shape matches Anthropic's wire format. This envelope is used ONLY on `/messages` and `/messages/count_tokens` properties: type: title: Type type: string description: Always `error` for error envelopes. enum: - error const: error error: $ref: '#/components/schemas/MessageError' request_id: title: Request Id type: string description: >- Opaque request identifier echoed by the server, useful for correlating client-side failures with server-side logs. nullable: true required: - type - error MessageInputMessage: title: Message Input Message type: object description: A turn in the conversation. properties: role: type: string enum: - user - assistant description: >- Conversational role. `user` for the human-side turn, `assistant` for prior model output. content: anyOf: - type: string - title: Content Block Array type: array items: $ref: '#/components/schemas/MessageInputContentBlock' required: - role - content MessageSystemPrompt: title: Message System Prompt description: >- System prompt for the conversation. Accepts either a single string (most common) or an array of text blocks (used when individual segments need `cache_control` markers). Multiple text blocks are joined with newlines and prepended to the conversation as a `role: system` message. anyOf: - type: string - title: System Text Block Array type: array items: $ref: '#/components/schemas/MessageSystemTextBlock' MessageMetadata: title: Message Metadata type: object description: >- Free-form metadata attached to the request. Currently only `user_id` Additional fields are accepted but ignored. properties: user_id: title: User Id type: string description: >- External identifier for the end-user making the request. Mapped internally to the Chat Completions `user` field. nullable: true additionalProperties: true MessageThinkingConfig: title: Message Thinking Config description: >- Controls Anthropic-style extended thinking. **In v1**: only `type:"disabled"` is silently accepted as a no-op; `type:"enabled"` and `type:"adaptive"` return a 400 `invalid_request_error` (`unsupported_parameter`). oneOf: - $ref: '#/components/schemas/MessageThinkingDisabled' - $ref: '#/components/schemas/MessageThinkingEnabled' - $ref: '#/components/schemas/MessageThinkingAdaptive' discriminator: propertyName: type mapping: disabled: $ref: '#/components/schemas/MessageThinkingDisabled' enabled: $ref: '#/components/schemas/MessageThinkingEnabled' adaptive: $ref: '#/components/schemas/MessageThinkingAdaptive' MessageTool: title: Message Tool type: object description: >- User-defined function tool definition. Only custom function tools are supported (Anthropic's `type:"custom"` style or the absent-type Beta style). Anthropic-hosted server tools (`web_search`, `code_execution`, `bash`, `text_editor`, `memory`, `tool_search` variants) return 400 `unsupported_tool_type` if sent. properties: name: type: string description: Tool name. Must match `^[a-zA-Z0-9_-]+$`. type: type: string enum: - custom nullable: true description: >- Tool-type discriminator. May be omitted (defaults to custom) or set to `custom`. Other values return 400 `unsupported_tool_type`. description: type: string nullable: true description: Human-readable description of when the tool should be used. input_schema: type: object additionalProperties: true nullable: true description: >- JSON Schema describing the tool's expected input. Required by the Anthropic spec; accepted as optional by SambaNova. cache_control: allOf: - $ref: '#/components/schemas/MessageCacheControl' nullable: true description: >- Silently dropped on tool definitions (only content and system cache markers participate in caching). input_examples: type: array items: type: object additionalProperties: true nullable: true description: Silently dropped. allowed_callers: type: array items: type: string nullable: true description: Silently dropped. defer_loading: type: boolean nullable: true description: Silently dropped. eager_input_streaming: type: boolean nullable: true description: Silently dropped. strict: type: boolean nullable: true description: Silently dropped. required: - name MessageToolChoice: title: Message Tool Choice description: How the model should choose from the provided tools. oneOf: - $ref: '#/components/schemas/MessageToolChoiceAuto' - $ref: '#/components/schemas/MessageToolChoiceAny' - $ref: '#/components/schemas/MessageToolChoiceNone' - $ref: '#/components/schemas/MessageToolChoiceTool' discriminator: propertyName: type mapping: auto: $ref: '#/components/schemas/MessageToolChoiceAuto' any: $ref: '#/components/schemas/MessageToolChoiceAny' none: $ref: '#/components/schemas/MessageToolChoiceNone' tool: $ref: '#/components/schemas/MessageToolChoiceTool' MessageOutputContentBlock: title: Message Output Content Block description: Typed content block in the model's response. oneOf: - $ref: '#/components/schemas/MessageOutputTextBlock' - $ref: '#/components/schemas/MessageOutputToolUseBlock' - $ref: '#/components/schemas/MessageOutputThinkingBlock' - $ref: '#/components/schemas/MessageOutputRedactedThinkingBlock' - $ref: '#/components/schemas/MessageOutputServerToolUseBlock' - $ref: '#/components/schemas/MessageOutputWebSearchToolResultBlock' - $ref: '#/components/schemas/MessageOutputWebFetchToolResultBlock' - $ref: '#/components/schemas/MessageOutputCodeExecutionToolResultBlock' - $ref: '#/components/schemas/MessageOutputBashCodeExecutionToolResultBlock' - $ref: >- #/components/schemas/MessageOutputTextEditorCodeExecutionToolResultBlock - $ref: '#/components/schemas/MessageOutputToolSearchToolResultBlock' - $ref: '#/components/schemas/MessageOutputContainerUploadBlock' discriminator: propertyName: type mapping: text: $ref: '#/components/schemas/MessageOutputTextBlock' tool_use: $ref: '#/components/schemas/MessageOutputToolUseBlock' thinking: $ref: '#/components/schemas/MessageOutputThinkingBlock' redacted_thinking: $ref: '#/components/schemas/MessageOutputRedactedThinkingBlock' server_tool_use: $ref: '#/components/schemas/MessageOutputServerToolUseBlock' web_search_tool_result: $ref: '#/components/schemas/MessageOutputWebSearchToolResultBlock' web_fetch_tool_result: $ref: '#/components/schemas/MessageOutputWebFetchToolResultBlock' code_execution_tool_result: $ref: '#/components/schemas/MessageOutputCodeExecutionToolResultBlock' bash_code_execution_tool_result: $ref: '#/components/schemas/MessageOutputBashCodeExecutionToolResultBlock' text_editor_code_execution_tool_result: $ref: >- #/components/schemas/MessageOutputTextEditorCodeExecutionToolResultBlock tool_search_tool_result: $ref: '#/components/schemas/MessageOutputToolSearchToolResultBlock' container_upload: $ref: '#/components/schemas/MessageOutputContainerUploadBlock' MessageStopReason: title: Message Stop Reason type: string description: >- Reason the model stopped generating. SambaNova emits `end_turn`, `max_tokens`, `tool_use`, and `stop_sequence`. The remaining values are defined for Anthropic SDK type-parity but never returned: `pause_turn` (server-tool loop limit, not produced); `refusal` (content filter, not exposed); `model_context_window_exceeded` (folded to `max_tokens`). enum: - end_turn - max_tokens - tool_use - pause_turn - refusal - stop_sequence - model_context_window_exceeded MessageStopDetails: title: Message Stop Details type: object description: >- Refusal stop details. Anthropic compatibility only — `refusal` is never emitted as a stop_reason by SambaNova (content filtering is not exposed at the API layer). properties: type: type: string enum: - refusal const: refusal category: type: string enum: - cyber - bio required: - type MessageUsage: title: Message Usage type: object description: Token accounting for the request. properties: input_tokens: type: integer description: Total tokens in the prompt (system + messages + tools). output_tokens: type: integer description: Total tokens generated by the model. cache_creation_input_tokens: type: integer nullable: true description: >- Tokens written to prompt cache. Absent in v1; emitted once prompt caching wiring lands (CP-2897). cache_read_input_tokens: type: integer nullable: true description: >- Tokens read from prompt cache. Absent in v1; emitted once prompt caching wiring lands (CP-2897). server_tool_use: type: object additionalProperties: true nullable: true description: >- Server-tool usage metrics (e.g. `web_search_requests`). Anthropic compatibility only — SambaNova does not run server tools, so this field is never emitted. service_tier: type: string nullable: true description: >- Service tier that processed the request. Anthropic compatibility only — SambaNova is single-tier and never emits this field. cache_creation: type: object additionalProperties: true nullable: true description: >- Anthropic SDK alias for cache write metrics. Always `null` in SambaNova responses; use `cache_creation_input_tokens` instead. inference_geo: type: string nullable: true description: >- Geographic region that served the request. Anthropic compatibility only - SambaNova does not expose geo routing, always `null`. required: - input_tokens - output_tokens MessageContainer: title: Message Container type: object description: >- Code-execution container reference. Anthropic compatibility only — SambaNova does not run server-side code execution, so this field is never emitted on responses. properties: id: type: string expires_at: type: string description: ISO-8601 timestamp. required: - id - expires_at MessageStartEvent: title: Message Start Event type: object description: >- First event of a stream. Carries the initial Message envelope (empty `content[]`, `stop_reason: null`) and token usage from prompt processing. properties: type: type: string enum: - message_start const: message_start message: $ref: '#/components/schemas/Message' required: - type - message MessageContentBlockStartEvent: title: Content Block Start Event type: object description: Opens a new content block. One per block in `content[]`. properties: type: type: string enum: - content_block_start const: content_block_start index: type: integer description: Zero-based index of the block within `content[]`. content_block: $ref: '#/components/schemas/MessageOutputContentBlock' required: - type - index - content_block MessageContentBlockDeltaEvent: title: Content Block Delta Event type: object description: Incremental update to the currently open content block. properties: type: type: string enum: - content_block_delta const: content_block_delta index: type: integer description: Zero-based index of the block within `content[]`. delta: $ref: '#/components/schemas/MessageContentBlockDelta' required: - type - index - delta MessageContentBlockStopEvent: title: Content Block Stop Event type: object description: Closes the current content block. properties: type: type: string enum: - content_block_stop const: content_block_stop index: type: integer description: Zero-based index of the block within `content[]`. required: - type - index MessageDeltaEvent: title: Message Delta Event type: object description: >- Penultimate event of the stream. Carries the final `stop_reason`, optional `stop_sequence`, and final usage counts. properties: type: type: string enum: - message_delta const: message_delta delta: type: object properties: stop_reason: $ref: '#/components/schemas/MessageStopReason' stop_sequence: type: string nullable: true description: >- Custom stop sequence that triggered termination. Field is emitted but value is always `null` in v1 (backend collapses `StopSequenceHit` and `EndOfText` into the same finish_reason). stop_details: allOf: - $ref: '#/components/schemas/MessageStopDetails' nullable: true description: >- Refusal stop details. Field is emitted but value is always `null` in v1 — `refusal` is not produced as a stop_reason. required: - stop_reason usage: $ref: '#/components/schemas/MessageDeltaUsage' required: - type - delta - usage MessageStopEvent: title: Message Stop Event type: object description: Final event of the stream. No fields beyond `type`. properties: type: type: string enum: - message_stop const: message_stop required: - type MessagePingEvent: title: Message Ping Event type: object description: Keepalive heartbeat. May appear at any point in the stream. properties: type: type: string enum: - ping const: ping required: - type MessageStreamErrorEvent: title: Message Stream Error Event type: object description: Streamed error envelope. Terminates the stream. properties: type: type: string enum: - error const: error error: $ref: '#/components/schemas/MessageError' required: - type - error MessageError: title: Message Error type: object description: >- Inner error object carried inside a `MessageErrorResponse`. The `type` value follows Anthropic's published error taxonomy. properties: type: title: Type type: string description: Error category. Values follow Anthropic's taxonomy. enum: - invalid_request_error - authentication_error - permission_error - not_found_error - request_too_large - rate_limit_error - api_error - overloaded_error - not_implemented_error message: title: Message type: string description: Human-readable explanation of the error. required: - type - message MessageInputContentBlock: title: Message Input Content Block description: Typed content block inside a message's `content` array. oneOf: - $ref: '#/components/schemas/MessageInputTextBlock' - $ref: '#/components/schemas/MessageInputImageBlock' - $ref: '#/components/schemas/MessageInputVideoBlock' - $ref: '#/components/schemas/MessageInputToolUseBlock' - $ref: '#/components/schemas/MessageInputToolResultBlock' - $ref: '#/components/schemas/MessageInputServerToolUseBlock' - $ref: '#/components/schemas/MessageInputSearchResultBlock' - $ref: '#/components/schemas/MessageInputWebSearchToolResultBlock' - $ref: '#/components/schemas/MessageInputWebFetchToolResultBlock' - $ref: '#/components/schemas/MessageInputCodeExecutionToolResultBlock' - $ref: '#/components/schemas/MessageInputBashCodeExecutionToolResultBlock' - $ref: >- #/components/schemas/MessageInputTextEditorCodeExecutionToolResultBlock - $ref: '#/components/schemas/MessageInputToolSearchToolResultBlock' - $ref: '#/components/schemas/MessageInputThinkingBlock' - $ref: '#/components/schemas/MessageInputRedactedThinkingBlock' - $ref: '#/components/schemas/MessageInputContainerUploadBlock' - $ref: '#/components/schemas/MessageInputDocumentBlock' discriminator: propertyName: type mapping: text: $ref: '#/components/schemas/MessageInputTextBlock' image: $ref: '#/components/schemas/MessageInputImageBlock' video: $ref: '#/components/schemas/MessageInputVideoBlock' tool_use: $ref: '#/components/schemas/MessageInputToolUseBlock' tool_result: $ref: '#/components/schemas/MessageInputToolResultBlock' server_tool_use: $ref: '#/components/schemas/MessageInputServerToolUseBlock' search_result: $ref: '#/components/schemas/MessageInputSearchResultBlock' web_search_tool_result: $ref: '#/components/schemas/MessageInputWebSearchToolResultBlock' web_fetch_tool_result: $ref: '#/components/schemas/MessageInputWebFetchToolResultBlock' code_execution_tool_result: $ref: '#/components/schemas/MessageInputCodeExecutionToolResultBlock' bash_code_execution_tool_result: $ref: '#/components/schemas/MessageInputBashCodeExecutionToolResultBlock' text_editor_code_execution_tool_result: $ref: >- #/components/schemas/MessageInputTextEditorCodeExecutionToolResultBlock tool_search_tool_result: $ref: '#/components/schemas/MessageInputToolSearchToolResultBlock' thinking: $ref: '#/components/schemas/MessageInputThinkingBlock' redacted_thinking: $ref: '#/components/schemas/MessageInputRedactedThinkingBlock' container_upload: $ref: '#/components/schemas/MessageInputContainerUploadBlock' document: $ref: '#/components/schemas/MessageInputDocumentBlock' MessageSystemTextBlock: title: Message System Text Block type: object description: >- A text segment within a structured `system` prompt array. Multiple text blocks are concatenated (with newlines) and prepended to the conversation as a `role: system` message at the chat-completions layer. properties: type: const: text enum: - text title: Type type: string text: title: Text type: string description: Plain-text content of the system prompt segment. cache_control: $ref: '#/components/schemas/MessageCacheControl' citations: title: Citations type: array description: 'Optional citations. **In v1**: silently dropped' items: type: object additionalProperties: true nullable: true required: - type - text MessageThinkingDisabled: title: Message Thinking Disabled type: object description: >- Disables Anthropic-style extended thinking. **In v1**: silently accepted as a no-op properties: type: title: Type type: string enum: - disabled const: disabled required: - type MessageThinkingEnabled: title: Message Thinking Enabled type: object description: >- Enables Anthropic-style extended thinking with a fixed budget. **In v1**: returns a 400 `invalid_request_error` (`unsupported_parameter`). properties: type: title: Type type: string enum: - enabled const: enabled budget_tokens: title: Budget Tokens type: integer minimum: 1024 description: >- Maximum tokens the model may spend on extended thinking before producing the final answer. required: - type - budget_tokens MessageThinkingAdaptive: title: Message Thinking Adaptive type: object description: >- Enables Anthropic-style adaptive extended thinking. **In v1**: returns a 400 `invalid_request_error` (`unsupported_parameter`). properties: type: title: Type type: string enum: - adaptive const: adaptive budget_tokens: title: Budget Tokens type: integer minimum: 1024 nullable: true description: >- Optional upper bound on tokens spent on adaptive thinking. When omitted, the backend chooses based on prompt complexity. required: - type MessageCacheControl: title: Message Cache Control type: object description: >- Marks the preceding content block (or system text block) as a prompt- cache breakpoint. Marker positions are collected by the adapter; their wiring into the router's longest-prefix matching **In v1**: position is recorded; the `ttl` value is ignored. properties: type: title: Type type: string enum: - ephemeral const: ephemeral description: Cache breakpoint type. Only `ephemeral` is supported by Anthropic. ttl: title: TTL type: string description: >- Optional time-to-live hint (e.g. `"5m"`, `"1h"`). **Currently ignored** in v1 nullable: true required: - type MessageToolChoiceAuto: title: Message Tool Choice (Auto) type: object description: Let the model decide whether and which tool to use. properties: type: type: string enum: - auto const: auto disable_parallel_tool_use: type: boolean nullable: true description: Silently dropped. required: - type MessageToolChoiceAny: title: Message Tool Choice (Any) type: object description: Require the model to call one of the provided tools. properties: type: type: string enum: - any const: any disable_parallel_tool_use: type: boolean nullable: true description: Silently dropped. required: - type MessageToolChoiceNone: title: Message Tool Choice (None) type: object description: Forbid the model from calling any tool. properties: type: type: string enum: - none const: none required: - type MessageToolChoiceTool: title: Message Tool Choice (Tool) type: object description: Force the model to call a specific tool by name. properties: type: type: string enum: - tool const: tool name: type: string description: Name of the required tool. disable_parallel_tool_use: type: boolean nullable: true description: Silently dropped. required: - type - name MessageOutputTextBlock: title: Message Output Text Block type: object description: Plain-text segment of the model's response. properties: type: type: string enum: - text const: text text: type: string citations: type: array items: type: object additionalProperties: true nullable: true description: Not emitted in v1. required: - type - text MessageOutputToolUseBlock: title: Message Output Tool Use Block type: object description: Tool call generated by the model. properties: type: type: string enum: - tool_use const: tool_use id: type: string description: Unique identifier for this tool call. name: type: string description: Name of the tool being called. input: type: object additionalProperties: true description: Tool inputs as a JSON object. caller: type: object additionalProperties: true nullable: true description: Anthropic routing metadata. Always `null` in SambaNova responses. required: - type - id - name - input MessageOutputThinkingBlock: title: Message Output Thinking Block type: object description: Extended-reasoning trace from the model. Emitted by reasoning models. properties: type: type: string enum: - thinking const: thinking thinking: type: string signature: type: string nullable: true required: - type - thinking MessageOutputRedactedThinkingBlock: title: Message Output Redacted Thinking Block type: object description: >- Anthropic compatibility only — SambaNova does not produce encrypted thinking output. Never emitted in responses. properties: type: type: string enum: - redacted_thinking const: redacted_thinking data: type: string required: - type - data MessageOutputServerToolUseBlock: title: Message Output Server Tool Use Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side tools. Never emitted in responses; defined for Anthropic SDK type-parity. properties: type: type: string enum: - server_tool_use const: server_tool_use id: type: string name: type: string input: type: object additionalProperties: true required: - type - id - name - input MessageOutputWebSearchToolResultBlock: title: Message Output Web Search Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `web_search`. Never emitted in responses. properties: type: type: string enum: - web_search_tool_result const: web_search_tool_result tool_use_id: type: string content: type: array items: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputWebFetchToolResultBlock: title: Message Output Web Fetch Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `web_fetch`. Never emitted in responses. properties: type: type: string enum: - web_fetch_tool_result const: web_fetch_tool_result tool_use_id: type: string content: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputCodeExecutionToolResultBlock: title: Message Output Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `code_execution`. Never emitted in responses. properties: type: type: string enum: - code_execution_tool_result const: code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputBashCodeExecutionToolResultBlock: title: Message Output Bash Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side bash code execution. Never emitted in responses. properties: type: type: string enum: - bash_code_execution_tool_result const: bash_code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputTextEditorCodeExecutionToolResultBlock: title: Message Output Text Editor Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side text-editor code execution. Never emitted in responses. properties: type: type: string enum: - text_editor_code_execution_tool_result const: text_editor_code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputToolSearchToolResultBlock: title: Message Output Tool Search Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `tool_search`. Never emitted in responses. properties: type: type: string enum: - tool_search_tool_result const: tool_search_tool_result tool_use_id: type: string content: type: object additionalProperties: true required: - type - tool_use_id - content MessageOutputContainerUploadBlock: title: Message Output Container Upload Block type: object description: >- Anthropic compatibility only — SambaNova does not produce container_upload blocks (these come from Anthropic's server-side `code_execution` tool). Never emitted in responses. properties: type: type: string enum: - container_upload const: container_upload file_id: type: string required: - type - file_id MessageContentBlockDelta: title: Content Block Delta description: Incremental update to an open content block. oneOf: - $ref: '#/components/schemas/MessageContentBlockTextDelta' - $ref: '#/components/schemas/MessageContentBlockInputJsonDelta' - $ref: '#/components/schemas/MessageContentBlockThinkingDelta' - $ref: '#/components/schemas/MessageContentBlockSignatureDelta' discriminator: propertyName: type mapping: text_delta: $ref: '#/components/schemas/MessageContentBlockTextDelta' input_json_delta: $ref: '#/components/schemas/MessageContentBlockInputJsonDelta' thinking_delta: $ref: '#/components/schemas/MessageContentBlockThinkingDelta' signature_delta: $ref: '#/components/schemas/MessageContentBlockSignatureDelta' MessageDeltaUsage: title: Message Delta Usage type: object description: >- Final token accounting emitted in the closing `message_delta` event of a stream. properties: output_tokens: type: integer description: Total tokens generated (final count). input_tokens: type: integer nullable: true description: Total tokens in the prompt (echoed from `message_start`). cache_creation_input_tokens: type: integer nullable: true description: Tokens written to prompt cache. Absent in v1; cache_read_input_tokens: type: integer nullable: true description: Tokens read from prompt cache. Absent in v1; server_tool_use: type: object additionalProperties: true nullable: true description: >- Server-tool usage metrics. Anthropic compatibility only — SambaNova does not run server tools, so this field is never emitted. required: - output_tokens MessageInputTextBlock: title: Message Input Text Block type: object description: Plain-text segment of a message. properties: type: type: string enum: - text const: text text: type: string cache_control: $ref: '#/components/schemas/MessageCacheControl' citations: title: Citations type: array items: type: object additionalProperties: true nullable: true required: - type - text MessageInputImageBlock: title: Message Input Image Block type: object description: >- Image content. Only `source.type:"base64"` is supported in v1; URL sources return 400. properties: type: type: string enum: - image const: image source: oneOf: - $ref: '#/components/schemas/MessageInputImageSourceBase64' - $ref: '#/components/schemas/MessageInputImageSourceUrl' discriminator: propertyName: type mapping: base64: $ref: '#/components/schemas/MessageInputImageSourceBase64' url: $ref: '#/components/schemas/MessageInputImageSourceUrl' cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - source MessageInputVideoBlock: title: Message Input Video Block type: object description: Video content. properties: type: type: string enum: - video const: video source: oneOf: - $ref: '#/components/schemas/MessageInputVideoSourceBase64' - $ref: '#/components/schemas/MessageInputVideoSourceUrl' discriminator: propertyName: type mapping: base64: $ref: '#/components/schemas/MessageInputVideoSourceBase64' url: $ref: '#/components/schemas/MessageInputVideoSourceUrl' cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - source MessageInputToolUseBlock: title: Message Input Tool Use Block type: object description: A prior assistant turn that invoked a tool. properties: type: type: string enum: - tool_use const: tool_use id: type: string description: >- Unique identifier for the tool call (used to correlate `tool_result`). name: type: string description: Name of the tool being invoked. input: type: object additionalProperties: true description: Tool inputs as a JSON object. cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - id - name - input MessageInputToolResultBlock: title: Message Input Tool Result Block type: object description: Result of a prior tool call. properties: type: type: string enum: - tool_result const: tool_result tool_use_id: type: string description: ID of the `tool_use` block this result corresponds to. content: anyOf: - type: string - title: Tool Result Content Array type: array items: $ref: '#/components/schemas/MessageInputToolResultContent' is_error: type: boolean nullable: true description: Silently dropped in v1. cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id MessageInputServerToolUseBlock: title: Message Input Server Tool Use Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side tools. A prior assistant turn that invoked an Anthropic-hosted tool (web_search, code_execution, etc.). Accepted in conversation history (e.g. replaying an Anthropic-served session) but never originates from a SambaNova response. New `server_tool_use`-type tool definitions on outgoing requests are rejected with 400 `unsupported_tool_type`. properties: type: type: string enum: - server_tool_use const: server_tool_use id: type: string name: type: string input: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - id - name - input MessageInputSearchResultBlock: title: Message Input Search Result Block type: object description: >- Inline search result content. In v1 the `title`, `source`, and `content[]` text are extracted into a text block; citations are dropped. properties: type: type: string enum: - search_result const: search_result title: type: string source: type: string content: type: array items: $ref: '#/components/schemas/MessageInputTextBlock' citations: type: object additionalProperties: true nullable: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type MessageInputWebSearchToolResultBlock: title: Message Input Web Search Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `web_search`. Echo of a prior Anthropic-served `web_search` tool call; accepted in conversation history but never originates from a SambaNova response. When present, only `title` (`url`) per result is extracted into a tool message. properties: type: type: string enum: - web_search_tool_result const: web_search_tool_result tool_use_id: type: string content: type: array items: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputWebFetchToolResultBlock: title: Message Input Web Fetch Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `web_fetch`. Echo of a prior Anthropic-served `web_fetch` tool call; accepted in conversation history but never originates from a SambaNova response. When present, only the text content is extracted. properties: type: type: string enum: - web_fetch_tool_result const: web_fetch_tool_result tool_use_id: type: string content: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputCodeExecutionToolResultBlock: title: Message Input Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `code_execution`. Echo of a prior Anthropic-served `code_execution` tool call; accepted in conversation history but never originates from a SambaNova response. When present, only `stdout`, `stderr`, and `return_code` are extracted; image output is dropped. properties: type: type: string enum: - code_execution_tool_result const: code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputBashCodeExecutionToolResultBlock: title: Message Input Bash Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side bash code execution. Echo of a prior Anthropic-served bash tool call; accepted in conversation history but never originates from a SambaNova response. Same lossy extraction as `code_execution_tool_result`. properties: type: type: string enum: - bash_code_execution_tool_result const: bash_code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputTextEditorCodeExecutionToolResultBlock: title: Message Input Text Editor Code Execution Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side text-editor code execution. Echo of a prior Anthropic-served text-editor tool call; accepted in conversation history but never originates from a SambaNova response. When present, only file content is extracted; metadata (line count, file type) is dropped. properties: type: type: string enum: - text_editor_code_execution_tool_result const: text_editor_code_execution_tool_result tool_use_id: type: string content: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputToolSearchToolResultBlock: title: Message Input Tool Search Tool Result Block type: object description: >- Anthropic compatibility only — SambaNova does not run server-side `tool_search`. Echo of a prior Anthropic-served `tool_search` tool call; accepted in conversation history but never originates from a SambaNova response. When present, an empty string is emitted to the tool message (no plain-text fields). properties: type: type: string enum: - tool_search_tool_result const: tool_search_tool_result tool_use_id: type: string content: type: object additionalProperties: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - tool_use_id - content MessageInputThinkingBlock: title: Message Input Thinking Block type: object description: Extended-reasoning trace from a prior assistant turn. properties: type: type: string enum: - thinking const: thinking thinking: type: string signature: type: string required: - type - thinking - signature MessageInputRedactedThinkingBlock: title: Message Input Redacted Thinking Block type: object description: >- Anthropic compatibility only — SambaNova does not produce encrypted thinking output. Echo of a prior Anthropic-served response where `thinking.display:"omitted"` was set. Accepted in conversation history but never originates from a SambaNova response. Silently dropped on input. properties: type: type: string enum: - redacted_thinking const: redacted_thinking data: type: string required: - type - data MessageInputContainerUploadBlock: title: Message Input Container Upload Block type: object description: >- Anthropic compatibility only — SambaNova does not produce container_upload blocks (these come from Anthropic's server-side `code_execution` tool). Accepted in conversation history but never originates from a SambaNova response. Silently dropped on input. properties: type: type: string enum: - container_upload const: container_upload file_id: type: string cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - file_id MessageInputDocumentBlock: title: Message Input Document Block type: object description: >- PDF or document content. **Returns 400** — no document-extraction pipeline available. properties: type: type: string enum: - document const: document source: type: object additionalProperties: true title: type: string nullable: true context: type: string nullable: true citations: type: object additionalProperties: true nullable: true cache_control: $ref: '#/components/schemas/MessageCacheControl' required: - type - source MessageContentBlockTextDelta: title: Content Block Text Delta type: object description: Incremental text chunk for an open text content block. properties: type: type: string enum: - text_delta const: text_delta text: type: string required: - type - text MessageContentBlockInputJsonDelta: title: Content Block Input JSON Delta type: object description: >- Incremental fragment of a tool_use block's `input` JSON. Concatenate successive `partial_json` strings to reconstruct the full input object. properties: type: type: string enum: - input_json_delta const: input_json_delta partial_json: type: string required: - type - partial_json MessageContentBlockThinkingDelta: title: Content Block Thinking Delta type: object description: >- Incremental thinking chunk for an open thinking block. Emitted by reasoning models. properties: type: type: string enum: - thinking_delta const: thinking_delta thinking: type: string required: - type - thinking MessageContentBlockSignatureDelta: title: Content Block Signature Delta type: object description: >- Signature for an open thinking block. Emitted at the end of a thinking stream (paired with the closing `content_block_stop`); the `signature` value may be an empty string when the backend has no signed payload to attach. properties: type: type: string enum: - signature_delta const: signature_delta signature: type: string required: - type - signature MessageInputImageSourceBase64: title: Message Input Image Source (Base64) type: object description: Inline image data encoded as base64. properties: type: type: string enum: - base64 const: base64 media_type: type: string enum: - image/jpeg - image/png - image/webp description: MIME type of the image bytes. data: type: string description: Base64-encoded image bytes (no `data:` URI prefix). required: - type - media_type - data MessageInputImageSourceUrl: title: Message Input Image Source (URL) type: object description: >- HTTPS URL pointing to an image. **Returns 400 in v1** — URL fetching is blocked. Use `type:"base64"` instead. properties: type: type: string enum: - url const: url url: type: string format: uri required: - type - url MessageInputVideoSourceBase64: title: Message Input Video Source (Base64) type: object description: Inline video data encoded as base64. properties: type: type: string enum: - base64 const: base64 media_type: type: string enum: - video/mp4 description: MIME type of the video bytes. data: type: string description: Base64-encoded video bytes (no `data:` URI prefix). required: - type - media_type - data MessageInputVideoSourceUrl: title: Message Input Video Source (URL) type: object description: HTTPS URL pointing to a video. properties: type: type: string enum: - url const: url url: type: string format: uri required: - type - url MessageInputToolResultContent: title: Message Input Tool Result Content description: Content of a `tool_result` block. only text is forwarded. oneOf: - $ref: '#/components/schemas/MessageInputTextBlock' - $ref: '#/components/schemas/MessageInputImageBlock' discriminator: propertyName: type mapping: text: $ref: '#/components/schemas/MessageInputTextBlock' image: $ref: '#/components/schemas/MessageInputImageBlock' securitySchemes: api_key: type: http description: >- SambaNova API key, sent as a bearer token in the `Authorization` header (`Authorization: Bearer `). Default authentication scheme used by the SambaNova SDK across every OpenAI compatible endpoint. scheme: bearer bearerFormat: apiKey x_api_key: type: apiKey description: >- SambaNova API key, sent in the `x-api-key` header. Accepted on the Messages API routes (`/messages`, `/messages/count_tokens`) so that callers can point the official Anthropic SDK at SambaNova by swapping `base_url` — no code changes. The credential value is the same as the bearer `api_key` scheme; only the transport differs. in: header name: x-api-key ````