Lamini Inference Completions API

openapi: 3.0.1
info:
  title: Lamini Platform 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.
  - name: Fine-Tuning
    description: Tuning and Memory Tuning job submission and management.
  - name: Classify
    description: LLM classifier classification and prediction endpoints.
  - name: Embeddings
    description: Text embedding generation.
  - name: Platform
    description: Version and platform metadata 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'
  /v1/embedding:
    post:
      operationId: createEmbedding
      tags:
        - Embeddings
      summary: Generate embeddings
      description: >-
        Encode one or more text prompts into embedding vectors for similarity
        search, retrieval, and indexing.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/EmbeddingRequest'
      responses:
        '200':
          description: Generated embedding vectors.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/EmbeddingResponse'
  /v1/train:
    post:
      operationId: createTrainingJob
      tags:
        - Fine-Tuning
      summary: Submit a tuning job
      description: >-
        Submit a fine-tuning or Memory Tuning job for a base model against an
        uploaded dataset. The train_type field selects the tuning method (for
        example full fine-tuning or memory_rag).
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TrainRequest'
      responses:
        '200':
          description: The submitted training job.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TrainJob'
  /v1/train/jobs:
    get:
      operationId: listTrainingJobs
      tags:
        - Fine-Tuning
      summary: List tuning jobs
      description: List tuning jobs for the authenticated account.
      responses:
        '200':
          description: A list of tuning jobs.
          content:
            application/json:
              schema:
                type: array
                items:
                  $ref: '#/components/schemas/TrainJob'
  /v1/train/jobs/{job_id}:
    get:
      operationId: getTrainingJob
      tags:
        - Fine-Tuning
      summary: Get a tuning job
      description: Retrieve the status and metadata of a single tuning job.
      parameters:
        - name: job_id
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: The requested tuning job.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TrainJob'
  /v1/train/jobs/{job_id}/cancel:
    post:
      operationId: cancelTrainingJob
      tags:
        - Fine-Tuning
      summary: Cancel a tuning job
      description: Cancel a running tuning job by its job id.
      parameters:
        - name: job_id
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: The cancelled tuning job.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TrainJob'
  /v1/train/jobs/{job_id}/resume:
    post:
      operationId: resumeTrainingJob
      tags:
        - Fine-Tuning
      summary: Resume a tuning job
      description: Resume a previously paused or cancelled tuning job.
      parameters:
        - name: job_id
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: The resumed tuning job.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TrainJob'
  /v1/train/jobs/cancel:
    post:
      operationId: cancelAllTrainingJobs
      tags:
        - Fine-Tuning
      summary: Cancel all tuning jobs
      description: Cancel all running tuning jobs for the authenticated account.
      responses:
        '200':
          description: Cancellation result.
  /v1/classifier/{model_id}/classification:
    post:
      operationId: classify
      tags:
        - Classify
      summary: Classify text
      description: >-
        Classify one or more text prompts with a trained classifier model,
        returning scored class labels.
      parameters:
        - name: model_id
          in: path
          required: true
          schema:
            type: string
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ClassifyRequest'
      responses:
        '200':
          description: Classification scores.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ClassifyResponse'
  /v1/classifier/{model_id}/prediction:
    post:
      operationId: predict
      tags:
        - Classify
      summary: Predict a class
      description: >-
        Return the single predicted class for one or more text prompts using a
        trained classifier model.
      parameters:
        - name: model_id
          in: path
          required: true
          schema:
            type: string
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ClassifyRequest'
      responses:
        '200':
          description: Predicted class label(s).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ClassifyResponse'
  /v1/version:
    get:
      operationId: getVersion
      tags:
        - Platform
      summary: Get platform version
      description: >-
        Return server and client version metadata and the list of available
        API endpoints and their versions.
      responses:
        '200':
          description: Version and endpoint metadata.
components:
  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.
  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.
    EmbeddingRequest:
      type: object
      required:
        - prompt
      properties:
        prompt:
          oneOf:
            - type: string
            - type: array
              items:
                type: string
          description: One or more texts to embed.
    EmbeddingResponse:
      type: object
      properties:
        embedding:
          type: array
          items:
            type: array
            items:
              type: number
          description: One embedding vector per input prompt.
    TrainRequest:
      type: object
      required:
        - model_name
        - dataset_id
      properties:
        model_name:
          type: string
          description: Base model to tune.
        dataset_id:
          type: string
          description: Identifier of an uploaded dataset.
        upload_file_path:
          type: string
          nullable: true
        finetune_args:
          type: object
          additionalProperties: true
          nullable: true
        gpu_config:
          type: object
          additionalProperties: true
          nullable: true
        is_public:
          type: boolean
          nullable: true
        custom_model_name:
          type: string
          nullable: true
        train_type:
          type: string
          nullable: true
          description: >-
            Tuning method, for example full fine-tuning, memory_rag, or
            mome_mini (Memory Tuning).
    TrainJob:
      type: object
      properties:
        job_id:
          type: string
        dataset_id:
          type: string
        status:
          type: string
          description: Current status of the tuning job.
        model_name:
          type: string
    ClassifyRequest:
      type: object
      required:
        - prompt
      properties:
        prompt:
          oneOf:
            - type: string
            - type: array
              items:
                type: string
        top_n:
          type: integer
          nullable: true
          description: Return only the top N classes.
        threshold:
          type: number
          nullable: true
          description: Confidence threshold for a prediction.
    ClassifyResponse:
      type: object
      properties:
        classification:
          type: array
          items:
            type: object
            additionalProperties: true
          description: Scored class labels per prompt.