generated: '2026-08-26'
method: derived
source: openapi/multiverse-computing-compactifai-openapi.yml
also_searched:
- https://docs.compactif.ai/models/
- https://docs.compactif.ai/api_reference/
summary:
  entities: 6
  relationships: 10
  note: 'The public data model is small and job-shaped. Of 13 operations only 5 return a
    named schema; the rest declare an empty 200 body, so the entity graph below is derived
    from the schemas the contract actually names plus the id-reference fields inside them.'
entities:
- name: Model
  id_field: id
  id_format: 'Slug, e.g. hypernova-60b, cai-mistral-small-3-1-slim, quasar-438b'
  operations:
  - get_model_info_list_v1_models_get
  - get_model_info_v1_models__model__get
  schema: null
  schema_note: 'Both model operations declare an empty 200 schema ({}) in the OpenAPI, so
    the model object shape is undocumented in the contract. The human catalog at
    https://docs.compactif.ai/models/ lists 12 model IDs with original architecture.'
  referenced_by:
  - ChatCompletionRequest.model
  - CompletionRequest.model
  - ResponsesRequest.model
  - Batch.model
- name: ChatCompletion
  id_field: id
  operations:
  - chat_completions_v1_chat_completions_post
  schema: null
  schema_note: 'POST /v1/chat/completions declares an empty 200 schema. The REQUEST is
    fully typed (ChatCompletionRequest, 18 properties) but the response is not — the
    flagship operation of the API has no documented response body.'
- name: Completion
  id_field: id
  operations:
  - complete_prompt_v1_completions_post
  schema: CompletionResponse
  fields:
  - id
  - choices
  - created
  - model
  - object
  - system_fingerprint
  - usage
- name: Response
  id_field: id
  operations:
  - post_responses_v1_responses_post
  - get_response_v1_responses__response_id__get
  schema: ResponsesResult
  key_fields:
  - id
  - model
  - status
  - output
  - usage
  - previous_response_id
  - conversation
  - metadata
  - completed_at
  persistence: 'Only responses created with store:true are retrievable, and only by the
    user who created them (per the operation description).'
- name: Batch
  id_field: id
  id_prefix: batch_
  operations:
  - create_batch_v1_batches_post
  - get_batches_v1_batches_get
  - get_batch_v1_batches__batch_id__get
  - cancel_batch_v1_batches__batch_id__cancel_post
  schema: Batch
  key_fields:
  - id
  - status
  - endpoint
  - completion_window
  - input_file_id
  - output_file_id
  - error_file_id
  - request_counts
  - usage
  - metadata
  states:
  - validating
  - in_progress
  - finalizing
  - completed
  - failed
  - expired
  - cancelling
  - cancelled
  timestamps:
  - created_at
  - in_progress_at
  - finalizing_at
  - completed_at
  - failed_at
  - expired_at
  - expires_at
  - cancelling_at
  - cancelled_at
  list_envelope: BatchList {object, data, has_more, first_id, last_id}
- name: File
  id_field: id
  operations:
  - post_files_v1_files_post
  - get_file_content_v1_files__file_id__content_get
  schema: FileObject
  key_fields:
  - id
  - object
  - bytes
  - created_at
  - filename
  - purpose
  - owner_id
  - s3_uri
  note: 'FileObject.id is documented as the identifier "to use with /v1/responses"; purpose
    is free text (e.g. user_data, assistants). FileObject exposes an s3_uri (S3Uri schema),
    leaking the storage backend into the public contract.'
relationships:
- from: Batch
  to: File
  type: belongs_to
  via: input_file_id
  note: The JSONL of requests the batch runs.
- from: Batch
  to: File
  type: has_one
  via: output_file_id
  cardinality: 0..1
- from: Batch
  to: File
  type: has_one
  via: error_file_id
  cardinality: 0..1
- from: Batch
  to: Model
  type: belongs_to
  via: model
- from: BatchList
  to: Batch
  type: has_many
  via: data
- from: Response
  to: Model
  type: belongs_to
  via: model
- from: Response
  to: Response
  type: belongs_to
  via: previous_response_id
  note: Self-referencing conversation chain.
- from: Response
  to: File
  type: belongs_to
  via: ResponsesRequest.file_id
  cardinality: 0..1
- from: ChatCompletion
  to: Model
  type: belongs_to
  via: model
- from: Completion
  to: Model
  type: belongs_to
  via: model
usage_accounting:
  entity: CompletionUsage / BatchUsage / ResponseUsage
  fields:
  - prompt_tokens / input_tokens
  - completion_tokens / output_tokens
  - total_tokens
  - PromptTokensDetails
  - CompletionTokensDetails
  note: Token counts are the billing unit; see plans/multiverse-computing-plans-pricing.yml.
gaps:
- 'No account, organization, API key or usage entity is exposed. Usage and billing moved to
  the Dashboard when POST /v1/usage/completions was removed on 2026-03-24, so there is no
  programmatic way to read spend.'
- 'No DELETE exists for File or Response, so nothing in the model can be removed through
  the API.'
- 'Five operations (chat completions, models list, model info, audio transcriptions, file
  content JSON) declare an empty 200 schema, so a code generator produces untyped
  responses for the two most-called endpoints in the product.'