version: v2
kind: Spicepod
name: vercel-ai-sdk

runtime:
  dataset_load_parallelism: 1

datasets:
  - from: github:github.com/vercel/ai/files/main
    name: vercel_ai_sdk.docs
    description: Vercel AI SDK documentation from https://sdk.vercel.ai/docs
    params:
      file_format: md
      github_client_id: ${secrets:GITHUB_CLIENT_ID}
      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
      include: "content/docs/**/*.mdx"
    acceleration:
      enabled: true
      refresh_check_interval: 4h
      refresh_jitter_enabled: true
      refresh_jitter_max: 15m
    columns:
      - name: content
        embeddings:
          - from: openai_embeddings
            row_id:
              - path
            chunking:
              enabled: true
              target_chunk_size: 512
              overlap_size: 128
              trim_whitespace: true

  - from: github:github.com/vercel/ai/issues
    name: vercel_ai_sdk.issues
    description: Vercel AI SDK GitHub issues from github.com/vercel/ai
    time_column: created_at
    params:
      file_format: md
      github_client_id: ${secrets:GITHUB_CLIENT_ID}
      github_private_key: ${secrets:GITHUB_PRIVATE_KEY}
      github_installation_id: ${secrets:GITHUB_INSTALLATION_ID}
    acceleration:
      enabled: true
      refresh_mode: full
      refresh_check_interval: 4h
      refresh_jitter_enabled: true
      refresh_jitter_max: 15m
      refresh_sql: "SELECT * FROM vercel_ai_sdk.issues LIMIT 100"
    # embeddings:
    #   - column: title
    #     column_pk:
    #       - id
    #     use: openai_embeddings

embeddings:
  - from: openai
    name: openai_embeddings
    params:
      openai_api_key: ${ secrets:OPENAI_API_KEY }