API Evangelist enhancements for the Eden AI API V3

overlay: 1.0.0
info:
  title: API Evangelist enhancements for the Eden AI API V3
  version: 1.0.0
extends: ../openapi/_original/eden-ai-v3-openapi.json
x-provenance:
  generated: '2026-09-06'
  method: generated
  source: >-
    Derived from the harvested first-party spec at https://api.edenai.run/v3/docs/openapi.json plus
    Eden AI's published documentation. Records API Evangelist's additions ONLY — the original spec
    is never mutated. Nothing here is asserted as Eden AI's own text.
  note: >-
    The upstream spec is FastAPI-generated and carries no tags[], no servers description, no
    security block at the document level, and only 200/422 responses. These actions add the
    navigation and runtime semantics an agent needs, every one of them traceable to a documented
    Eden AI statement recorded elsewhere in this repository.
actions:
  - target: $.info
    description: Record the surfaces and the harvest provenance alongside the contract.
    update:
      x-apis-io-harvest:
        harvested: '2026-09-06'
        url: https://api.edenai.run/v3/docs/openapi.json
        http_status: 200
        indexed_by_provider_at: https://www.edenai.co/docs/llms.txt
      x-companion-surfaces:
        mcp: https://mcp.edenai.run/mcp
        webhooks: asyncapi/eden-ai-webhooks.yml
        legacy_v2_slices: 'https://api.edenai.run/v2/info/splitted-schema/{slice}/openapi.json'
  - target: $.servers
    description: Add the documented EU regional endpoint, which the upstream spec omits.
    update:
      - url: https://api.eu.edenai.run
        description: >-
          EU endpoint — routes exclusively through providers and models cleared for European
          processing. Same key, same request and response shapes.
        x-source: https://www.edenai.co/docs/v3/data-governance/eu-endpoint
  - target: $
    description: Apply the bearer scheme at the document level; upstream declares AuthBearer per operation but sets no root security.
    update:
      security:
        - AuthBearer: []
      tags:
        - {name: LLM, description: 'OpenAI-compatible chat, responses, embeddings, moderation.'}
        - {name: Anthropic, description: 'Anthropic Messages-compatible pass-through.'}
        - {name: Images, description: OpenAI-compatible image generation and edits.}
        - {name: Audio, description: OpenAI-compatible transcription and speech.}
        - {name: Universal AI, description: 'Expert models — one model-routed endpoint, sync and async.'}
        - {name: Collections, description: Provider-side collections and their items.}
        - {name: Files, description: Persistent file storage referenced by id from any file parameter.}
        - {name: Info, description: Feature, subfeature and model catalog.}
  - target: $.paths['/v3/universal-ai/async'].post
    description: Record the webhook callback contract, which the upstream spec references only as a request field.
    update:
      x-callbacks:
        async_job_completed:
          transport: 'HTTPS POST to webhook_receiver'
          signature: 'X-Edenai-Signature (RSA PKCS1 v1.5 over the SHA-256 hex digest of canonical JSON)'
          catalog: asyncapi/eden-ai-webhooks.yml
          docs: https://www.edenai.co/docs/v3/expert-models/webhooks
  - target: $.paths['/v3/chat/completions'].post
    description: Flag the undeclared runtime failure modes an agent must handle on this operation.
    update:
      x-runtime-semantics:
        undeclared_statuses: [402, 429, 500]
        error_in_2xx: >-
          When the primary model and every provider in `fallbacks` fail, the failure is reported in
          the body of a 2xx. Inspect `error` even on success.
        idempotency: none
        catalog: errors/eden-ai-problem-types.yml
  - target: $.paths['/v3/universal-ai'].post
    description: Same runtime hazards on the expert-model endpoint.
    update:
      x-runtime-semantics:
        undeclared_statuses: [400, 402, 403, 404, 429, 500]
        model_format: 'feature/subfeature/provider'
        idempotency: none
        catalog: errors/eden-ai-problem-types.yml