kdeps-components-vector_store

version: 1.0.0
requires:
    schemas/v1: ""
    tools/vector_store: ""
files:
    component.yaml: |
        apiVersion: kdeps.io/v1
        kind: Component
        metadata:
          name: vector_store
          description: Run one vector store operation through the kdeps API
          version: "1.0.0"
        interface:
          inputs:
            - name: operation
              type: string
              required: true
              description: "add_documents or similarity_search"
            - name: collection
              type: string
              required: true
              description: Collection, index, or table name in the store
            - name: provider
              type: string
              description: "Vector store backend: qdrant (default), chroma, pinecone, azureaisearch, opensearch, elasticsearch, weaviate, pgvector, postgres, alloydb, cloudsql, mariadb, dolt, mysql, mongodb, redis, bedrock"
            - name: url
              type: string
              description: Endpoint or DSN for the store
            - name: apiKey
              type: string
              description: "Auth token. For mongodb the database name (default kdeps); for opensearch/elasticsearch user:pass"
            - name: documents
              type: array
              description: Documents to upsert for add_documents, each an object with content and optional metadata
            - name: query
              type: string
              description: Natural language search query (operation similarity_search)
            - name: topK
              type: number
              default: 0
              description: Number of results to return for similarity_search; 0 uses the default of 5
            - name: embedModel
              type: string
              description: Embedding model used to vectorize documents and queries, for example text-embedding-3-small
            - name: embedBackend
              type: string
              description: "Embedding provider: openai, ollama, google, or system (follow KDEPS_DEFAULT_BACKEND)"
            - name: embedBaseURL
              type: string
              description: Custom base URL for an OpenAI-compatible embedding backend
          # body is the handler text. The HTTP envelope stays inside the call.
          output: "{{ output('call').body }}"
        resources:
          - actionId: call
            name: Vector store operation
            httpClient:
              method: POST
              # In-process host. The HTTP executor calls the handler without dialing.
              url: http://kdeps.api/apis/tools.kdeps.io/v1/tools/vector_store/call
              headers:
                Content-Type: application/json
              failOnError: true
              timeout: 120s  # hard stop for the call; the store and embedding backend are remote
              data:
                operation: "{{ get('vector_store.operation') }}"
                collection: "{{ get('vector_store.collection') }}"
                provider: "{{ get('vector_store.provider') }}"
                url: "{{ get('vector_store.url') }}"
                apiKey: "{{ get('vector_store.apiKey') }}"
                documents: "{{ get('vector_store.documents') }}"
                query: "{{ get('vector_store.query') }}"
                topK: "{{ get('vector_store.topK') }}"
                embedModel: "{{ get('vector_store.embedModel') }}"
                embedBackend: "{{ get('vector_store.embedBackend') }}"
                embedBaseURL: "{{ get('vector_store.embedBaseURL') }}"