naftiko: 1.0.0-alpha2
info:
  label: Fastly AI Accelerator — Chat Completions
  description: Fastly AI Accelerator semantic-caching proxy for LLM chat completions. Provides OpenAI- and Google Gemini-compatible endpoints served from the Fastly edge, returning cached responses for semantically similar prompts.
  tags:
  - Fastly
  - AI Accelerator
  - AI
  - LLM
  - Semantic Caching
  created: '2026-05-22'
  modified: '2026-05-22'
binds:
- namespace: env
  keys:
    FASTLY_API_KEY: FASTLY_API_KEY
capability:
  consumes:
  - type: http
    namespace: ai-accelerator
    baseUri: https://api.fastly.ai
    description: Fastly AI Accelerator semantic-caching proxy for LLM chat completions.
    resources:
    - name: openai-v1-chat-completions
      path: /openai/v1/chat/completions
      operations:
      - name: createopenaichatcompletion
        method: POST
        description: Create chat completion via OpenAI (semantically cached)
        outputRawFormat: json
        outputParameters:
        - name: result
          type: object
          value: $.
    - name: gemini-v1-models-model-generate-content
      path: /gemini/v1/models/{model}:generateContent
      operations:
      - name: generategeminicontent
        method: POST
        description: Generate content via Google Gemini (semantically cached)
        outputRawFormat: json
        outputParameters:
        - name: result
          type: object
          value: $.
    - name: openai-v1-embeddings
      path: /openai/v1/embeddings
      operations:
      - name: createembeddings
        method: POST
        description: Create embeddings via OpenAI
        outputRawFormat: json
        outputParameters:
        - name: result
          type: object
          value: $.
    authentication:
      type: apikey
      key: Fastly-Key
      value: '{{env.FASTLY_API_KEY}}'
      placement: header
  exposes:
  - type: rest
    namespace: ai-accelerator-rest
    port: 8080
    description: REST adapter for Fastly AI Accelerator.
    resources:
    - path: /v1/openai/v1/chat/completions
      name: openai-v1-chat-completions
      description: REST surface for OpenAI-compatible chat completions through AI Accelerator.
      operations:
      - method: POST
        name: createopenaichatcompletion
        description: Create chat completion via OpenAI
        call: ai-accelerator.createopenaichatcompletion
        outputParameters:
        - type: object
          mapping: $.
    - path: /v1/gemini/v1/models/{model}/generate-content
      name: gemini-v1-models-model-generate-content
      description: REST surface for Google Gemini generateContent through AI Accelerator.
      operations:
      - method: POST
        name: generategeminicontent
        description: Generate content via Google Gemini
        call: ai-accelerator.generategeminicontent
        outputParameters:
        - type: object
          mapping: $.
    - path: /v1/openai/v1/embeddings
      name: openai-v1-embeddings
      description: REST surface for OpenAI embeddings through AI Accelerator.
      operations:
      - method: POST
        name: createembeddings
        description: Create embeddings
        call: ai-accelerator.createembeddings
        outputParameters:
        - type: object
          mapping: $.
  - type: mcp
    namespace: ai-accelerator-mcp
    port: 9090
    transport: http
    description: MCP adapter for Fastly AI Accelerator.
    tools:
    - name: openai-chat-completion
      description: Create OpenAI-compatible chat completion via Fastly AI Accelerator semantic cache
      hints:
        readOnly: false
        destructive: false
        idempotent: false
      call: ai-accelerator.createopenaichatcompletion
      outputParameters:
      - type: object
        mapping: $.
    - name: gemini-generate-content
      description: Generate Google Gemini content via Fastly AI Accelerator
      hints:
        readOnly: false
        destructive: false
        idempotent: false
      call: ai-accelerator.generategeminicontent
      outputParameters:
      - type: object
        mapping: $.
    - name: openai-embeddings
      description: Create OpenAI embeddings via Fastly AI Accelerator
      hints:
        readOnly: false
        destructive: false
        idempotent: true
      call: ai-accelerator.createembeddings
      outputParameters:
      - type: object
        mapping: $.