Lamini Inference API

openapi: 3.0.1
info:
  title: Lamini Platform Classify Inference API
  description: REST API for the Lamini enterprise LLM platform covering inference (completions), fine-tuning and Memory Tuning jobs, classification, and embeddings over open base and tuned models. All requests are authenticated with a Bearer API key and served from https://api.lamini.ai. Endpoints and request fields are derived from the official Lamini Python client (github.com/lamini-ai/lamini) and the Lamini REST API documentation.
  termsOfService: https://www.lamini.ai/terms
  contact:
    name: Lamini Support
    url: https://www.lamini.ai
  version: '1.0'
servers:
- url: https://api.lamini.ai
security:
- bearerAuth: []
tags:
- name: Inference
  description: Text completion and streaming generation endpoints.
paths:
  /v1/completions:
    post:
      operationId: createCompletion
      tags:
      - Inference
      summary: Generate a completion
      description: Generate a text completion from a base or tuned model. Supports a typed output schema via output_type for structured generation.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CompletionRequest'
      responses:
        '200':
          description: A generated completion.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/CompletionResponse'
        '401':
          description: Missing or invalid API key.
        '429':
          description: Rate limit exceeded.
  /v3/streaming_completions:
    post:
      operationId: createStreamingCompletion
      tags:
      - Inference
      summary: Generate a streaming completion
      description: Generate a completion as an incremental stream of token chunks for the provided prompt and model.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CompletionRequest'
      responses:
        '200':
          description: A stream of completion chunks.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/CompletionResponse'
components:
  schemas:
    CompletionRequest:
      type: object
      required:
      - prompt
      - model_name
      properties:
        prompt:
          oneOf:
          - type: string
          - type: array
            items:
              type: string
          description: One or more input prompts.
        model_name:
          type: string
          description: Base or tuned model identifier to generate from.
        output_type:
          type: object
          additionalProperties: true
          description: Optional typed output schema for structured generation.
        max_tokens:
          type: integer
          nullable: true
        max_new_tokens:
          type: integer
          nullable: true
        cache_id:
          type: string
          nullable: true
    CompletionResponse:
      type: object
      properties:
        output:
          oneOf:
          - type: string
          - type: object
            additionalProperties: true
          description: Generated text or structured output.
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: 'Lamini platform API key passed as Authorization: Bearer <API_KEY>. Requests may also include a Lamini-Version header.'