openapi: 3.1.0
info:
  title: EvolutionaryScale Forge ESM3 Embeddings API
  description: 'Hosted inference API for the ESM3 multimodal protein language model from EvolutionaryScale. ESM3 reasons jointly across sequence, structure, and function tracks. The Forge API exposes generate, batch_generate, encode, decode, forward_and_sample, and logits operations across small (1.4B), medium (7B), and large (98B) parameter checkpoints. All requests authenticate with a Forge API token obtained from forge.evolutionaryscale.ai. This OpenAPI is reconstructed from the open-source `esm` Python SDK (Biohub/esm) which is the canonical client for the Forge service.

    '
  version: 2024-08-01
  contact:
    name: EvolutionaryScale Forge
    url: https://forge.evolutionaryscale.ai
  license:
    name: Cambrian Inference Clickthrough License Agreement
    url: https://www.evolutionaryscale.ai/policies/cambrian-inference-clickthrough-license-agreement
servers:
- url: https://forge.evolutionaryscale.ai
  description: EvolutionaryScale Forge Production
security:
- BearerAuth: []
tags:
- name: Embeddings
  description: Sequence-only tokenization and representation extraction.
paths:
  /api/v1/esmc/encode:
    post:
      summary: Encode Protein Sequence With ESM C
      description: 'Tokenize an `ESMProtein` (sequence track only) into an `ESMProteinTensor` ready for logits / embedding retrieval. ESM C operates over the sequence track exclusively.

        '
      operationId: esmcEncode
      tags:
      - Embeddings
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ESMCEncodeRequest'
            examples:
              SimpleEncode:
                $ref: '#/components/examples/SimpleEncodeExample'
      responses:
        '200':
          description: Sequence tensor for the supplied protein.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ESMProteinTensor'
        4XX:
          $ref: '#/components/responses/ErrorResponse'
components:
  schemas:
    ESMCEncodeRequest:
      type: object
      required:
      - model
      - protein
      properties:
        model:
          type: string
          example: esmc-300m-2024-12
          enum:
          - esmc-300m-2024-12
          - esmc-600m-2024-12
          - esmc-6b-2024-12
        protein:
          $ref: '#/components/schemas/ESMProtein'
    Error:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
            message:
              type: string
            code:
              type: string
    ESMProtein:
      type: object
      properties:
        sequence:
          type: string
          description: Amino acid sequence using one-letter codes.
          example: MKTAYIAKQRQISFVKAHFSRQLEERLGLIEVQ
    ESMProteinTensor:
      type: object
      properties:
        sequence:
          type: array
          items:
            type: integer
  examples:
    SimpleEncodeExample:
      summary: Encode a short protein sequence
      value:
        model: esmc-300m-2024-12
        protein:
          sequence: AAAAA
  responses:
    ErrorResponse:
      description: Error returned by the Forge API.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      description: Forge API token issued via forge.evolutionaryscale.ai.