llama-stack-mirror/source/openapi.yaml

components:
  responses: {}
  schemas:
    AgenticSystemCreateRequest:
      additionalProperties: false
      properties:
        available_tools:
          items:
            additionalProperties: false
            properties:
              input_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              output_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              parameters:
                additionalProperties:
                  oneOf:
                  - type: 'null'
                  - type: boolean
                  - type: number
                  - type: string
                  - type: array
                  - type: object
                type: object
              tool_name:
                oneOf:
                - enum:
                  - web_search
                  - math
                  - image_gen
                  - code_interpreter
                  type: string
                - type: string
            required:
            - tool_name
            - input_shields
            - output_shields
            type: object
          type: array
        executable_tools:
          items:
            type: string
          type: array
          uniqueItems: true
        input_shields:
          items:
            $ref: '#/components/schemas/ShieldConfig'
          type: array
        instructions:
          type: string
        memory_bank_uuids:
          items:
            type: string
          type: array
        model:
          enum:
          - llama3_8b_chat
          - llama3_70b_chat
          type: string
        output_shields:
          items:
            $ref: '#/components/schemas/ShieldConfig'
          type: array
        uuid:
          type: string
      required:
      - uuid
      - instructions
      - model
      - available_tools
      - executable_tools
      - memory_bank_uuids
      - input_shields
      - output_shields
      type: object
    AgenticSystemCreateResponse:
      additionalProperties: false
      properties:
        agent_uuid:
          type: string
      required:
      - agent_uuid
      type: object
    AgenticSystemExecuteRequest:
      additionalProperties: false
      properties:
        agent_uuid:
          type: string
        messages:
          items:
            $ref: '#/components/schemas/Message'
          type: array
        stream:
          default: false
          type: boolean
        turn_history:
          items:
            $ref: '#/components/schemas/AgenticSystemTurn'
          type: array
      required:
      - agent_uuid
      - messages
      - turn_history
      - stream
      type: object
    AgenticSystemExecuteResponse:
      additionalProperties: false
      properties:
        turn:
          $ref: '#/components/schemas/AgenticSystemTurn'
      required:
      - turn
      title: non-stream response from the agentic system.
      type: object
    AgenticSystemExecuteResponseStreamChunk:
      additionalProperties: false
      properties:
        event_type:
          enum:
          - step_start
          - step_end
          - step_progress
          title: The type of event.
          type: string
        response_text_delta:
          type: string
        retrieved_document:
          additionalProperties: false
          properties:
            content:
              contentEncoding: base64
              type: string
            metadata:
              additionalProperties:
                oneOf:
                - type: 'null'
                - type: boolean
                - type: number
                - type: string
                - type: array
                - type: object
              type: object
            mime_type:
              type: string
            uuid:
              type: string
          required:
          - uuid
          - content
          - metadata
          - mime_type
          type: object
        step_type:
          enum:
          - model_inference
          - tool_execution
          - safety_filtering
          - memory_retrieval
          title: The type of execution step.
          type: string
        step_uuid:
          type: string
        stop_reason:
          enum:
          - not_stopped
          - finished_ok
          - max_tokens
          title: Stop reasons are used to indicate why the model stopped generating
            text.
          type: string
        tool_call:
          additionalProperties: false
          properties:
            arguments:
              additionalProperties:
                oneOf:
                - type: 'null'
                - type: boolean
                - type: number
                - type: string
                - type: array
                - type: object
              type: object
            tool_name:
              type: string
          required:
          - tool_name
          - arguments
          title: A tool call is a request to a tool.
          type: object
        tool_response_delta:
          additionalProperties: false
          properties:
            content:
              oneOf:
              - type: string
              - $ref: '#/components/schemas/Attachment'
              - items:
                  oneOf:
                  - type: string
                  - $ref: '#/components/schemas/Attachment'
                type: array
            tool_name:
              type: string
          required:
          - tool_name
          - content
          type: object
        violation:
          additionalProperties: false
          properties:
            details:
              type: string
            suggested_user_response:
              type: string
            violation_type:
              type: string
          required:
          - violation_type
          - details
          type: object
      required:
      - event_type
      - step_uuid
      - step_type
      title: Streamed agent execution response.
      type: object
    AgenticSystemTurn:
      additionalProperties: false
      properties:
        response_message:
          $ref: '#/components/schemas/Message'
        steps:
          items:
            oneOf:
            - additionalProperties: false
              properties:
                logprobs:
                  additionalProperties:
                    oneOf:
                    - type: 'null'
                    - type: boolean
                    - type: number
                    - type: string
                    - type: array
                    - type: object
                  type: object
                step_type:
                  default: model_inference
                  enum:
                  - model_inference
                  - tool_execution
                  - safety_filtering
                  - memory_retrieval
                  title: The type of execution step.
                  type: string
                text:
                  type: string
                uuid:
                  type: string
              required:
              - step_type
              - uuid
              - text
              type: object
            - additionalProperties: false
              properties:
                step_type:
                  default: tool_execution
                  enum:
                  - model_inference
                  - tool_execution
                  - safety_filtering
                  - memory_retrieval
                  title: The type of execution step.
                  type: string
                tool_calls:
                  items:
                    additionalProperties: false
                    properties:
                      arguments:
                        additionalProperties:
                          oneOf:
                          - type: 'null'
                          - type: boolean
                          - type: number
                          - type: string
                          - type: array
                          - type: object
                        type: object
                      tool_name:
                        type: string
                    required:
                    - tool_name
                    - arguments
                    title: A tool call is a request to a tool.
                    type: object
                  type: array
                tool_responses:
                  items:
                    additionalProperties: false
                    properties:
                      content:
                        oneOf:
                        - type: string
                        - $ref: '#/components/schemas/Attachment'
                        - items:
                            oneOf:
                            - type: string
                            - $ref: '#/components/schemas/Attachment'
                          type: array
                      tool_name:
                        type: string
                    required:
                    - tool_name
                    - content
                    type: object
                  type: array
                uuid:
                  type: string
              required:
              - step_type
              - uuid
              - tool_calls
              - tool_responses
              type: object
            - additionalProperties: false
              properties:
                step_type:
                  default: safety_filtering
                  enum:
                  - model_inference
                  - tool_execution
                  - safety_filtering
                  - memory_retrieval
                  title: The type of execution step.
                  type: string
                uuid:
                  type: string
                violation:
                  additionalProperties: false
                  properties:
                    details:
                      type: string
                    suggested_user_response:
                      type: string
                    violation_type:
                      type: string
                  required:
                  - violation_type
                  - details
                  type: object
              required:
              - step_type
              - uuid
              type: object
            - additionalProperties: false
              properties:
                documents:
                  items:
                    additionalProperties: false
                    properties:
                      content:
                        contentEncoding: base64
                        type: string
                      metadata:
                        additionalProperties:
                          oneOf:
                          - type: 'null'
                          - type: boolean
                          - type: number
                          - type: string
                          - type: array
                          - type: object
                        type: object
                      mime_type:
                        type: string
                      uuid:
                        type: string
                    required:
                    - uuid
                    - content
                    - metadata
                    - mime_type
                    type: object
                  type: array
                scores:
                  items:
                    type: number
                  type: array
                step_type:
                  default: memory_retrieval
                  enum:
                  - model_inference
                  - tool_execution
                  - safety_filtering
                  - memory_retrieval
                  title: The type of execution step.
                  type: string
                uuid:
                  type: string
              required:
              - step_type
              - uuid
              - documents
              - scores
              type: object
          type: array
        user_messages:
          items:
            $ref: '#/components/schemas/Message'
          type: array
      required:
      - user_messages
      - steps
      - response_message
      title: A single turn in an interaction with an Agentic System.
      type: object
    Attachment:
      additionalProperties: false
      properties:
        mime_type:
          type: string
        url:
          $ref: '#/components/schemas/URL'
      required:
      - url
      - mime_type
      title: Attachments are used to refer to external resources, such as images,
        videos, audio, etc.
      type: object
    BatchChatCompletionRequest:
      additionalProperties: false
      properties:
        available_tools:
          items:
            additionalProperties: false
            properties:
              input_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              output_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              parameters:
                additionalProperties:
                  oneOf:
                  - type: 'null'
                  - type: boolean
                  - type: number
                  - type: string
                  - type: array
                  - type: object
                type: object
              tool_name:
                oneOf:
                - enum:
                  - web_search
                  - math
                  - image_gen
                  - code_interpreter
                  type: string
                - type: string
            required:
            - tool_name
            - input_shields
            - output_shields
            type: object
          type: array
        batch_dialogs:
          items:
            $ref: '#/components/schemas/Dialog'
          type: array
        logprobs:
          default: false
          type: boolean
        max_tokens:
          default: 0
          type: integer
        model:
          enum:
          - llama3_8b_chat
          - llama3_70b_chat
          type: string
        sampling_params:
          additionalProperties: false
          properties:
            strategy:
              default: greedy
              type: string
            temperature:
              default: 0.0
              type: number
            top_k:
              default: 0
              type: integer
            top_p:
              default: 0.95
              type: number
          required:
          - temperature
          - strategy
          - top_p
          - top_k
          type: object
      required:
      - model
      - batch_dialogs
      - sampling_params
      - available_tools
      - max_tokens
      - logprobs
      type: object
    BatchCompletionRequest:
      additionalProperties: false
      properties:
        content_batch:
          items:
            oneOf:
            - type: string
            - $ref: '#/components/schemas/Attachment'
            - items:
                oneOf:
                - type: string
                - $ref: '#/components/schemas/Attachment'
              type: array
          type: array
        logprobs:
          default: false
          type: boolean
        max_tokens:
          default: 0
          type: integer
        model:
          enum:
          - llama3_8b
          - llama3_70b
          type: string
        sampling_params:
          additionalProperties: false
          properties:
            strategy:
              default: greedy
              type: string
            temperature:
              default: 0.0
              type: number
            top_k:
              default: 0
              type: integer
            top_p:
              default: 0.95
              type: number
          required:
          - temperature
          - strategy
          - top_p
          - top_k
          type: object
      required:
      - model
      - content_batch
      - sampling_params
      - max_tokens
      - logprobs
      type: object
    ChatCompletionRequest:
      additionalProperties: false
      properties:
        available_tools:
          items:
            additionalProperties: false
            properties:
              input_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              output_shields:
                items:
                  $ref: '#/components/schemas/ShieldConfig'
                type: array
              parameters:
                additionalProperties:
                  oneOf:
                  - type: 'null'
                  - type: boolean
                  - type: number
                  - type: string
                  - type: array
                  - type: object
                type: object
              tool_name:
                oneOf:
                - enum:
                  - web_search
                  - math
                  - image_gen
                  - code_interpreter
                  type: string
                - type: string
            required:
            - tool_name
            - input_shields
            - output_shields
            type: object
          type: array
        dialog:
          $ref: '#/components/schemas/Dialog'
        logprobs:
          default: false
          type: boolean
        max_tokens:
          default: 0
          type: integer
        model:
          enum:
          - llama3_8b_chat
          - llama3_70b_chat
          type: string
        sampling_params:
          additionalProperties: false
          properties:
            strategy:
              default: greedy
              type: string
            temperature:
              default: 0.0
              type: number
            top_k:
              default: 0
              type: integer
            top_p:
              default: 0.95
              type: number
          required:
          - temperature
          - strategy
          - top_p
          - top_k
          type: object
        stream:
          default: false
          type: boolean
      required:
      - model
      - dialog
      - sampling_params
      - available_tools
      - max_tokens
      - stream
      - logprobs
      type: object
    ChatCompletionResponse:
      additionalProperties: false
      properties:
        content:
          oneOf:
          - type: string
          - $ref: '#/components/schemas/Attachment'
          - items:
              oneOf:
              - type: string
              - $ref: '#/components/schemas/Attachment'
            type: array
        logprobs:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        stop_reason:
          enum:
          - not_stopped
          - finished_ok
          - max_tokens
          title: Stop reasons are used to indicate why the model stopped generating
            text.
          type: string
        tool_calls:
          items:
            additionalProperties: false
            properties:
              arguments:
                additionalProperties:
                  oneOf:
                  - type: 'null'
                  - type: boolean
                  - type: number
                  - type: string
                  - type: array
                  - type: object
                type: object
              tool_name:
                type: string
            required:
            - tool_name
            - arguments
            title: A tool call is a request to a tool.
            type: object
          type: array
      required:
      - content
      - tool_calls
      title: Normal chat completion response.
      type: object
    ChatCompletionResponseStreamChunk:
      additionalProperties: false
      properties:
        stop_reason:
          enum:
          - not_stopped
          - finished_ok
          - max_tokens
          title: Stop reasons are used to indicate why the model stopped generating
            text.
          type: string
        text_delta:
          type: string
        tool_call:
          additionalProperties: false
          properties:
            arguments:
              additionalProperties:
                oneOf:
                - type: 'null'
                - type: boolean
                - type: number
                - type: string
                - type: array
                - type: object
              type: object
            tool_name:
              type: string
          required:
          - tool_name
          - arguments
          title: A tool call is a request to a tool.
          type: object
      required:
      - text_delta
      title: Streamed chat completion response. The actual response is a series of
        such objects.
      type: object
    CompletionRequest:
      additionalProperties: false
      properties:
        content:
          oneOf:
          - type: string
          - $ref: '#/components/schemas/Attachment'
          - items:
              oneOf:
              - type: string
              - $ref: '#/components/schemas/Attachment'
            type: array
        logprobs:
          default: false
          type: boolean
        max_tokens:
          default: 0
          type: integer
        model:
          enum:
          - llama3_8b
          - llama3_70b
          type: string
        sampling_params:
          additionalProperties: false
          properties:
            strategy:
              default: greedy
              type: string
            temperature:
              default: 0.0
              type: number
            top_k:
              default: 0
              type: integer
            top_p:
              default: 0.95
              type: number
          required:
          - temperature
          - strategy
          - top_p
          - top_k
          type: object
        stream:
          default: false
          type: boolean
      required:
      - content
      - model
      - sampling_params
      - max_tokens
      - stream
      - logprobs
      type: object
    CompletionResponse:
      additionalProperties: false
      properties:
        content:
          oneOf:
          - type: string
          - $ref: '#/components/schemas/Attachment'
          - items:
              oneOf:
              - type: string
              - $ref: '#/components/schemas/Attachment'
            type: array
        logprobs:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        stop_reason:
          enum:
          - not_stopped
          - finished_ok
          - max_tokens
          title: Stop reasons are used to indicate why the model stopped generating
            text.
          type: string
      required:
      - content
      title: Normal completion response.
      type: object
    CompletionResponseStreamChunk:
      additionalProperties: false
      properties:
        logprobs:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        stop_reason:
          enum:
          - not_stopped
          - finished_ok
          - max_tokens
          title: Stop reasons are used to indicate why the model stopped generating
            text.
          type: string
        text_delta:
          type: string
      required:
      - text_delta
      title: streamed completion response.
      type: object
    CreateDatasetRequest:
      additionalProperties: false
      properties:
        dataset:
          $ref: '#/components/schemas/Dataset'
        uuid:
          type: string
      required:
      - uuid
      - dataset
      title: Request to create a dataset.
      type: object
    DPOAlignmentConfig:
      additionalProperties: false
      properties:
        epsilon:
          type: number
        gamma:
          type: number
        reward_clip:
          type: number
        reward_scale:
          type: number
      required:
      - reward_scale
      - reward_clip
      - epsilon
      - gamma
      type: object
    Dataset:
      additionalProperties: false
      properties:
        columns:
          additionalProperties:
            enum:
            - dialog
            - text
            - media
            - number
            - json
            type: string
          type: object
        content_url:
          $ref: '#/components/schemas/URL'
        metadata:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
      required:
      - columns
      - content_url
      - metadata
      title: Dataset to be used for training or evaluating language models.
      type: object
    Dialog:
      additionalProperties: false
      properties:
        message:
          $ref: '#/components/schemas/Message'
        message_history:
          items:
            $ref: '#/components/schemas/Message'
          type: array
      required:
      - message
      - message_history
      type: object
    DoraFinetuningConfig:
      additionalProperties: false
      properties:
        alpha:
          type: integer
        apply_lora_to_mlp:
          type: boolean
        apply_lora_to_output:
          type: boolean
        lora_attn_modules:
          items:
            type: string
          type: array
        rank:
          type: integer
      required:
      - lora_attn_modules
      - apply_lora_to_mlp
      - apply_lora_to_output
      - rank
      - alpha
      type: object
    KScoredPromptGenerations:
      additionalProperties: false
      properties:
        k_scored_generations:
          items:
            $ref: '#/components/schemas/ScoredMessage'
          type: array
        prompt:
          $ref: '#/components/schemas/Message'
      required:
      - prompt
      - k_scored_generations
      type: object
    LoraFinetuningConfig:
      additionalProperties: false
      properties:
        alpha:
          type: integer
        apply_lora_to_mlp:
          type: boolean
        apply_lora_to_output:
          type: boolean
        lora_attn_modules:
          items:
            type: string
          type: array
        rank:
          type: integer
      required:
      - lora_attn_modules
      - apply_lora_to_mlp
      - apply_lora_to_output
      - rank
      - alpha
      type: object
    MemoryBank:
      additionalProperties: false
      properties:
        name:
          type: string
        uuid:
          type: string
      required:
      - uuid
      - name
      type: object
    MemoryBankDocument:
      additionalProperties: false
      properties:
        content:
          contentEncoding: base64
          type: string
        metadata:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        mime_type:
          type: string
        uuid:
          type: string
      required:
      - uuid
      - content
      - metadata
      - mime_type
      type: object
    Message:
      additionalProperties: false
      properties:
        content:
          oneOf:
          - type: string
          - $ref: '#/components/schemas/Attachment'
          - items:
              oneOf:
              - type: string
              - $ref: '#/components/schemas/Attachment'
            type: array
        role:
          enum:
          - system
          - user
          - assistant
          - tool
          type: string
        tool_calls:
          items:
            additionalProperties: false
            properties:
              arguments:
                additionalProperties:
                  oneOf:
                  - type: 'null'
                  - type: boolean
                  - type: number
                  - type: string
                  - type: array
                  - type: object
                type: object
              tool_name:
                type: string
            required:
            - tool_name
            - arguments
            title: A tool call is a request to a tool.
            type: object
          type: array
        tool_responses:
          items:
            additionalProperties: false
            properties:
              content:
                oneOf:
                - type: string
                - $ref: '#/components/schemas/Attachment'
                - items:
                    oneOf:
                    - type: string
                    - $ref: '#/components/schemas/Attachment'
                  type: array
              tool_name:
                type: string
            required:
            - tool_name
            - content
            type: object
          type: array
      required:
      - role
      - content
      - tool_calls
      - tool_responses
      type: object
    OptimizerConfig:
      additionalProperties: false
      properties:
        lr:
          type: number
        lr_min:
          type: number
        optimizer_type:
          enum:
          - adam
          - adamw
          - sgd
          type: string
        weight_decay:
          type: number
      required:
      - optimizer_type
      - lr
      - lr_min
      - weight_decay
      type: object
    PostTrainingJobArtifactsResponse:
      additionalProperties: false
      properties:
        checkpoints:
          items:
            additionalProperties: false
            properties:
              iters:
                type: integer
              path:
                $ref: '#/components/schemas/URL'
            required:
            - iters
            - path
            type: object
          type: array
        job_uuid:
          type: string
      required:
      - job_uuid
      - checkpoints
      title: Artifacts of a finetuning job.
      type: object
    PostTrainingJobLogStream:
      additionalProperties: false
      properties:
        job_uuid:
          type: string
        log_lines:
          items:
            type: string
          type: array
      required:
      - job_uuid
      - log_lines
      title: Stream of logs from a finetuning job.
      type: object
    PostTrainingJobStatusResponse:
      additionalProperties: false
      properties:
        checkpoints:
          items:
            additionalProperties: false
            properties:
              iters:
                type: integer
              path:
                $ref: '#/components/schemas/URL'
            required:
            - iters
            - path
            type: object
          type: array
        completed_at:
          format: date-time
          type: string
        job_uuid:
          type: string
        resources_allocated:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        scheduled_at:
          format: date-time
          type: string
        started_at:
          format: date-time
          type: string
        status:
          enum:
          - running
          - completed
          - failed
          - scheduled
          type: string
      required:
      - job_uuid
      - status
      - checkpoints
      title: Status of a finetuning job.
      type: object
    PostTrainingRLHFRequest:
      additionalProperties: false
      properties:
        algorithm:
          enum:
          - dpo
          type: string
        algorithm_config:
          $ref: '#/components/schemas/DPOAlignmentConfig'
        dataset:
          $ref: '#/components/schemas/Dataset'
        finetuned_model:
          $ref: '#/components/schemas/URL'
        hyperparam_search_config:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        job_uuid:
          type: string
        logger_config:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        optimizer_config:
          $ref: '#/components/schemas/OptimizerConfig'
        training_config:
          $ref: '#/components/schemas/TrainingConfig'
        validation_dataset:
          $ref: '#/components/schemas/Dataset'
      required:
      - job_uuid
      - finetuned_model
      - dataset
      - validation_dataset
      - algorithm
      - algorithm_config
      - optimizer_config
      - training_config
      - hyperparam_search_config
      - logger_config
      title: Request to finetune a model.
      type: object
    PostTrainingSFTRequest:
      additionalProperties: false
      properties:
        algorithm:
          enum:
          - full
          - lora
          - qlora
          - dora
          type: string
        algorithm_config:
          oneOf:
          - $ref: '#/components/schemas/LoraFinetuningConfig'
          - $ref: '#/components/schemas/QLoraFinetuningConfig'
          - $ref: '#/components/schemas/DoraFinetuningConfig'
        dataset:
          $ref: '#/components/schemas/Dataset'
        hyperparam_search_config:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        job_uuid:
          type: string
        logger_config:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        model:
          enum:
          - llama3_8b
          - llama3_70b
          type: string
        optimizer_config:
          $ref: '#/components/schemas/OptimizerConfig'
        training_config:
          $ref: '#/components/schemas/TrainingConfig'
        validation_dataset:
          $ref: '#/components/schemas/Dataset'
      required:
      - job_uuid
      - model
      - dataset
      - validation_dataset
      - algorithm
      - algorithm_config
      - optimizer_config
      - training_config
      - hyperparam_search_config
      - logger_config
      title: Request to finetune a model.
      type: object
    QLoraFinetuningConfig:
      additionalProperties: false
      properties:
        alpha:
          type: integer
        apply_lora_to_mlp:
          type: boolean
        apply_lora_to_output:
          type: boolean
        lora_attn_modules:
          items:
            type: string
          type: array
        rank:
          type: integer
      required:
      - lora_attn_modules
      - apply_lora_to_mlp
      - apply_lora_to_output
      - rank
      - alpha
      type: object
    RewardScoringRequest:
      additionalProperties: false
      properties:
        model:
          enum:
          - llama3_405b_reward
          type: string
        prompt_generations:
          items:
            additionalProperties: false
            properties:
              dialog:
                $ref: '#/components/schemas/Dialog'
              k_generations:
                items:
                  $ref: '#/components/schemas/Message'
                type: array
            required:
            - dialog
            - k_generations
            type: object
          type: array
      required:
      - prompt_generations
      - model
      title: Request to score a reward function. A list of prompts and a list of responses
        per prompt.
      type: object
    RewardScoringResponse:
      additionalProperties: false
      properties:
        scored_generations:
          items:
            $ref: '#/components/schemas/KScoredPromptGenerations'
          type: array
      required:
      - scored_generations
      title: Response from the reward scoring. Batch of (prompt, response, score)
        tuples that pass the threshold.
      type: object
    ScoredMessage:
      additionalProperties: false
      properties:
        message:
          $ref: '#/components/schemas/Message'
        score:
          type: number
      required:
      - message
      - score
      type: object
    ShieldConfig:
      additionalProperties: false
      properties:
        params:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        shield_type:
          enum:
          - llama_guard
          - prompt_guard
          - code_guard
          title: The type of safety shield.
          type: string
      required:
      - shield_type
      - params
      type: object
    SyntheticDataGenerationRequest:
      additionalProperties: false
      properties:
        filtering_function:
          default: none
          enum:
          - none
          - random
          - top_k
          - top_p
          - top_k_top_p
          - sigmoid
          title: The type of filtering function.
          type: string
        prompts:
          items:
            $ref: '#/components/schemas/Message'
          type: array
        reward_scoring:
          type: object
      required:
      - prompts
      - filtering_function
      title: Request to generate synthetic data. A small batch of prompts and a filtering
        function
      type: object
    SyntheticDataGenerationResponse:
      additionalProperties: false
      properties:
        statistics:
          additionalProperties:
            oneOf:
            - type: 'null'
            - type: boolean
            - type: number
            - type: string
            - type: array
            - type: object
          type: object
        synthetic_data:
          items:
            $ref: '#/components/schemas/KScoredPromptGenerations'
          type: array
      required:
      - synthetic_data
      title: Response from the synthetic data generation. Batch of (prompt, response,
        score) tuples that pass the threshold.
      type: object
    TrainingConfig:
      additionalProperties: false
      properties:
        batch_size:
          type: integer
        enable_activation_checkpointing:
          type: boolean
        fsdp_cpu_offload:
          type: boolean
        memory_efficient_fsdp_wrap:
          type: boolean
        n_epochs:
          type: integer
        n_iters:
          type: integer
        shuffle:
          type: boolean
      required:
      - n_epochs
      - batch_size
      - shuffle
      - n_iters
      - enable_activation_checkpointing
      - memory_efficient_fsdp_wrap
      - fsdp_cpu_offload
      type: object
    URL:
      format: uri
      pattern: ^(https?://|file://|data:)
      type: string
info:
  description: "Meta has built out a fairly sophisticated platform internally to post\
    \ train, evaluate, and \n                serve Llama models to support Meta’s\
    \ products. Given the newer capabilities of the llama models, \n             \
    \   the model development and model serving capabilities of the platform need\
    \ to be enhanced in \n                specific ways in order to best leverage\
    \ the models. For example, the inference platform needs \n                to support\
    \ code execution to take advantage of the built-in knowledge of tools of the model.\
    \ \n                The largest models are of high enough quality to be used to\
    \ generate synthetic data or be used \n                as reward models. There\
    \ are specific fine tuning and quantization techniques that we have found \n \
    \               result in the best performing Llama models. We would like to share\
    \ ways in which an LLM Ops \n                toolchain can be designed by leveraging\
    \ our learnings in getting Llama models to power Meta’s products.\n          \
    \      <br>\n                In addition, the Llama 3 models Meta will release\
    \ in July should not just be seen as a model, but \n                really as\
    \ a system starting the transition towards an entity capable of performing \"\
    agentic\" tasks \n                which require the ability to act as the central\
    \ planner and break a task down and perform multi-step \n                reasoning\
    \ and call tools for specific operations. In addition, there needs to be general\
    \ model-level \n                safety checks as well as task-specific safety\
    \ checks that are performed at a system level. \n                <br>\n      \
    \          We are defining the Llama Stack as a set of APIs and standards by synthesizing\
    \ our learnings while \n                working with Llama models. The APIs are\
    \ divided into the llama-toolchain-api and the llama-agentic-system-api. \n  \
    \              These APIs provide a coherent way for model developers to fine\
    \ tune and serve Llama models, and agentic app \n                developers to\
    \ leverage all the capabilities of the Llama models seamlessly. We would like\
    \ to work with the \n                ecosystem to enhance and simplify the API.\
    \ In addition, we will be releasing a plug-in architecture to allow \n       \
    \         creating distributions of the llama stack with different implementations.\n\
    \                <br>\n                This is the specification of the llama\
    \ stack that provides \n                a set of endpoints and their corresponding\
    \ interfaces that are tailored to \n                best leverage Llama Models.\
    \ The specification is still in draft and subject to change."
  title: '[DRAFT] Llama Stack Specification'
  version: 0.0.1
jsonSchemaDialect: https://json-schema.org/draft/2020-12/schema
openapi: 3.1.0
paths:
  /agentic_system/create:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/AgenticSystemCreateRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/AgenticSystemCreateResponse'
          description: OK
      tags:
      - AgenticSystem
  /agentic_system/delete:
    delete:
      parameters:
      - in: query
        name: agent_id
        required: true
        schema:
          type: string
      responses:
        '200':
          description: OK
      tags:
      - AgenticSystem
  /agentic_system/execute:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/AgenticSystemExecuteRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                oneOf:
                - $ref: '#/components/schemas/AgenticSystemExecuteResponse'
                - $ref: '#/components/schemas/AgenticSystemExecuteResponseStreamChunk'
          description: non-stream response from the agentic system. **OR** Streamed
            agent execution response.
      tags:
      - AgenticSystem
  /datasets/create:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CreateDatasetRequest'
        required: true
      responses:
        '200':
          description: OK
      tags:
      - Datasets
  /datasets/delete:
    delete:
      parameters:
      - in: query
        name: dataset_id
        required: true
        schema:
          type: string
      responses:
        '200':
          description: OK
      tags:
      - Datasets
  /datasets/get:
    get:
      parameters:
      - in: query
        name: dataset_id
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Dataset'
          description: OK
      tags:
      - Datasets
  /inference/batch_chat_completion:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/BatchChatCompletionRequest'
        required: true
      responses:
        '200':
          content:
            application/jsonl:
              schema:
                $ref: '#/components/schemas/ChatCompletionResponse'
          description: OK
      tags:
      - Inference
  /inference/batch_completion:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/BatchCompletionRequest'
        required: true
      responses:
        '200':
          content:
            application/jsonl:
              schema:
                $ref: '#/components/schemas/CompletionResponse'
          description: OK
      tags:
      - Inference
  /inference/chat_completion:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ChatCompletionRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                oneOf:
                - $ref: '#/components/schemas/ChatCompletionResponse'
                - $ref: '#/components/schemas/ChatCompletionResponseStreamChunk'
          description: Normal chat completion response. **OR** Streamed chat completion
            response. The actual response is a series of such objects.
      tags:
      - Inference
  /inference/completion:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CompletionRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                oneOf:
                - $ref: '#/components/schemas/CompletionResponse'
                - $ref: '#/components/schemas/CompletionResponseStreamChunk'
          description: Normal completion response. **OR** streamed completion response.
      tags:
      - Inference
  /memory_bank/delete:
    post:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      requestBody:
        content:
          application/json:
            schema:
              items:
                type: string
              type: array
        required: true
      responses:
        '200':
          content:
            application/jsonl:
              schema:
                type: string
          description: OK
      tags:
      - MemoryBanks
  /memory_bank/get:
    post:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      requestBody:
        content:
          application/json:
            schema:
              items:
                type: string
              type: array
        required: true
      responses:
        '200':
          content:
            application/jsonl:
              schema:
                $ref: '#/components/schemas/MemoryBankDocument'
          description: OK
      tags:
      - MemoryBanks
  /memory_bank/insert:
    post:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      requestBody:
        content:
          application/json:
            schema:
              items:
                additionalProperties: false
                properties:
                  content:
                    contentEncoding: base64
                    type: string
                  metadata:
                    additionalProperties:
                      oneOf:
                      - type: 'null'
                      - type: boolean
                      - type: number
                      - type: string
                      - type: array
                      - type: object
                    type: object
                  mime_type:
                    type: string
                  uuid:
                    type: string
                required:
                - uuid
                - content
                - metadata
                - mime_type
                type: object
              type: array
        required: true
      responses:
        '200':
          description: OK
      tags:
      - MemoryBanks
  /memory_bank/update:
    post:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      requestBody:
        content:
          application/json:
            schema:
              items:
                additionalProperties: false
                properties:
                  content:
                    contentEncoding: base64
                    type: string
                  metadata:
                    additionalProperties:
                      oneOf:
                      - type: 'null'
                      - type: boolean
                      - type: number
                      - type: string
                      - type: array
                      - type: object
                    type: object
                  mime_type:
                    type: string
                  uuid:
                    type: string
                required:
                - uuid
                - content
                - metadata
                - mime_type
                type: object
              type: array
        required: true
      responses:
        '200':
          description: OK
      tags:
      - MemoryBanks
  /memory_banks/create:
    post:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      - in: query
        name: bank_name
        required: true
        schema:
          type: string
      requestBody:
        content:
          application/json:
            schema:
              items:
                additionalProperties: false
                properties:
                  content:
                    contentEncoding: base64
                    type: string
                  metadata:
                    additionalProperties:
                      oneOf:
                      - type: 'null'
                      - type: boolean
                      - type: number
                      - type: string
                      - type: array
                      - type: object
                    type: object
                  mime_type:
                    type: string
                  uuid:
                    type: string
                required:
                - uuid
                - content
                - metadata
                - mime_type
                type: object
              type: array
        required: true
      responses:
        '200':
          description: OK
      tags:
      - MemoryBanks
  /memory_banks/drop:
    delete:
      parameters:
      - in: query
        name: bank_uuid
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                type: string
          description: OK
      tags:
      - MemoryBanks
  /memory_banks/get:
    get:
      parameters: []
      responses:
        '200':
          content:
            application/jsonl:
              schema:
                $ref: '#/components/schemas/MemoryBank'
          description: OK
      tags:
      - MemoryBanks
  /post_training/job/artifacts:
    get:
      parameters:
      - in: query
        name: job_uuid
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PostTrainingJobArtifactsResponse'
          description: OK
      tags:
      - PostTraining
  /post_training/job/logs:
    get:
      parameters:
      - in: query
        name: job_uuid
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PostTrainingJobLogStream'
          description: OK
      tags:
      - PostTraining
  /post_training/job/status:
    get:
      parameters:
      - in: query
        name: job_uuid
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PostTrainingJobStatusResponse'
          description: OK
      tags:
      - PostTraining
  /post_training/preference_optimize/:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PostTrainingRLHFRequest'
        required: true
      responses:
        '200':
          description: OK
      tags:
      - PostTraining
  /post_training/supervised_fine_tune/:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PostTrainingSFTRequest'
        required: true
      responses:
        '200':
          description: OK
      tags:
      - PostTraining
  /reward_scoring/score:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/RewardScoringRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/RewardScoringResponse'
          description: OK
      tags:
      - RewardScoring
  /synthetic_data_generation/generate:
    post:
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/SyntheticDataGenerationRequest'
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/SyntheticDataGenerationResponse'
          description: OK
      tags:
      - SyntheticDataGeneration
security:
- Default: []
servers:
- url: http://any-hosted-llama-stack.com
tags:
- name: SyntheticDataGeneration
- name: Inference
- name: MemoryBanks
- name: AgenticSystem
- name: Datasets
- name: RewardScoring
- name: PostTraining
- description: <SchemaDefinition schemaRef="#/components/schemas/ShieldConfig" />
  name: ShieldConfig
- description: <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemCreateRequest"
    />
  name: AgenticSystemCreateRequest
- description: <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemCreateResponse"
    />
  name: AgenticSystemCreateResponse
- description: <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemExecuteRequest"
    />
  name: AgenticSystemExecuteRequest
- description: 'A single turn in an interaction with an Agentic System.


    <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemTurn" />'
  name: AgenticSystemTurn
- description: 'Attachments are used to refer to external resources, such as images,
    videos, audio, etc.


    <SchemaDefinition schemaRef="#/components/schemas/Attachment" />'
  name: Attachment
- description: <SchemaDefinition schemaRef="#/components/schemas/Message" />
  name: Message
- description: <SchemaDefinition schemaRef="#/components/schemas/URL" />
  name: URL
- description: 'non-stream response from the agentic system.


    <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemExecuteResponse"
    />'
  name: AgenticSystemExecuteResponse
- description: 'Streamed agent execution response.


    <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemExecuteResponseStreamChunk"
    />'
  name: AgenticSystemExecuteResponseStreamChunk
- description: 'Request to create a dataset.


    <SchemaDefinition schemaRef="#/components/schemas/CreateDatasetRequest" />'
  name: CreateDatasetRequest
- description: 'Dataset to be used for training or evaluating language models.


    <SchemaDefinition schemaRef="#/components/schemas/Dataset" />'
  name: Dataset
- description: <SchemaDefinition schemaRef="#/components/schemas/MemoryBank" />
  name: MemoryBank
- description: <SchemaDefinition schemaRef="#/components/schemas/MemoryBankDocument"
    />
  name: MemoryBankDocument
- description: 'Artifacts of a finetuning job.


    <SchemaDefinition schemaRef="#/components/schemas/PostTrainingJobArtifactsResponse"
    />'
  name: PostTrainingJobArtifactsResponse
- description: 'Status of a finetuning job.


    <SchemaDefinition schemaRef="#/components/schemas/PostTrainingJobStatusResponse"
    />'
  name: PostTrainingJobStatusResponse
- description: 'Stream of logs from a finetuning job.


    <SchemaDefinition schemaRef="#/components/schemas/PostTrainingJobLogStream" />'
  name: PostTrainingJobLogStream
- description: <SchemaDefinition schemaRef="#/components/schemas/BatchChatCompletionRequest"
    />
  name: BatchChatCompletionRequest
- description: <SchemaDefinition schemaRef="#/components/schemas/Dialog" />
  name: Dialog
- description: 'Normal chat completion response.


    <SchemaDefinition schemaRef="#/components/schemas/ChatCompletionResponse" />'
  name: ChatCompletionResponse
- description: <SchemaDefinition schemaRef="#/components/schemas/BatchCompletionRequest"
    />
  name: BatchCompletionRequest
- description: 'Normal completion response.


    <SchemaDefinition schemaRef="#/components/schemas/CompletionResponse" />'
  name: CompletionResponse
- description: <SchemaDefinition schemaRef="#/components/schemas/ChatCompletionRequest"
    />
  name: ChatCompletionRequest
- description: 'Streamed chat completion response. The actual response is a series
    of such objects.


    <SchemaDefinition schemaRef="#/components/schemas/ChatCompletionResponseStreamChunk"
    />'
  name: ChatCompletionResponseStreamChunk
- description: <SchemaDefinition schemaRef="#/components/schemas/CompletionRequest"
    />
  name: CompletionRequest
- description: 'streamed completion response.


    <SchemaDefinition schemaRef="#/components/schemas/CompletionResponseStreamChunk"
    />'
  name: CompletionResponseStreamChunk
- description: 'Request to generate synthetic data. A small batch of prompts and a
    filtering function


    <SchemaDefinition schemaRef="#/components/schemas/SyntheticDataGenerationRequest"
    />'
  name: SyntheticDataGenerationRequest
- description: <SchemaDefinition schemaRef="#/components/schemas/KScoredPromptGenerations"
    />
  name: KScoredPromptGenerations
- description: <SchemaDefinition schemaRef="#/components/schemas/ScoredMessage" />
  name: ScoredMessage
- description: 'Response from the synthetic data generation. Batch of (prompt, response,
    score) tuples that pass the threshold.


    <SchemaDefinition schemaRef="#/components/schemas/SyntheticDataGenerationResponse"
    />'
  name: SyntheticDataGenerationResponse
- description: <SchemaDefinition schemaRef="#/components/schemas/DPOAlignmentConfig"
    />
  name: DPOAlignmentConfig
- description: <SchemaDefinition schemaRef="#/components/schemas/OptimizerConfig"
    />
  name: OptimizerConfig
- description: 'Request to finetune a model.


    <SchemaDefinition schemaRef="#/components/schemas/PostTrainingRLHFRequest" />'
  name: PostTrainingRLHFRequest
- description: <SchemaDefinition schemaRef="#/components/schemas/TrainingConfig" />
  name: TrainingConfig
- description: 'Request to score a reward function. A list of prompts and a list of
    responses per prompt.


    <SchemaDefinition schemaRef="#/components/schemas/RewardScoringRequest" />'
  name: RewardScoringRequest
- description: 'Response from the reward scoring. Batch of (prompt, response, score)
    tuples that pass the threshold.


    <SchemaDefinition schemaRef="#/components/schemas/RewardScoringResponse" />'
  name: RewardScoringResponse
- description: <SchemaDefinition schemaRef="#/components/schemas/DoraFinetuningConfig"
    />
  name: DoraFinetuningConfig
- description: <SchemaDefinition schemaRef="#/components/schemas/LoraFinetuningConfig"
    />
  name: LoraFinetuningConfig
- description: 'Request to finetune a model.


    <SchemaDefinition schemaRef="#/components/schemas/PostTrainingSFTRequest" />'
  name: PostTrainingSFTRequest
- description: <SchemaDefinition schemaRef="#/components/schemas/QLoraFinetuningConfig"
    />
  name: QLoraFinetuningConfig
x-tagGroups:
- name: Operations
  tags:
  - AgenticSystem
  - Datasets
  - Inference
  - MemoryBanks
  - PostTraining
  - RewardScoring
  - SyntheticDataGeneration
- name: Types
  tags:
  - AgenticSystemCreateRequest
  - AgenticSystemCreateResponse
  - AgenticSystemExecuteRequest
  - AgenticSystemExecuteResponse
  - AgenticSystemExecuteResponseStreamChunk
  - AgenticSystemTurn
  - Attachment
  - BatchChatCompletionRequest
  - BatchCompletionRequest
  - ChatCompletionRequest
  - ChatCompletionResponse
  - ChatCompletionResponseStreamChunk
  - CompletionRequest
  - CompletionResponse
  - CompletionResponseStreamChunk
  - CreateDatasetRequest
  - DPOAlignmentConfig
  - Dataset
  - Dialog
  - DoraFinetuningConfig
  - KScoredPromptGenerations
  - LoraFinetuningConfig
  - MemoryBank
  - MemoryBankDocument
  - Message
  - OptimizerConfig
  - PostTrainingJobArtifactsResponse
  - PostTrainingJobLogStream
  - PostTrainingJobStatusResponse
  - PostTrainingRLHFRequest
  - PostTrainingSFTRequest
  - QLoraFinetuningConfig
  - RewardScoringRequest
  - RewardScoringResponse
  - ScoredMessage
  - ShieldConfig
  - SyntheticDataGenerationRequest
  - SyntheticDataGenerationResponse
  - TrainingConfig
  - URL