gRPC Connector¶

Strake supports querying typed microservices directly over gRPC. It dynamically encodes JSON request payloads into binary Protobuf payloads, sends the request over dynamic HTTP/2 channels, and decodes the stream directly into Arrow RecordBatch formats.

1. Schema Discovery Modes¶

To map binary Protobuf data to Arrow records, Strake requires schema definitions. It supports two modes:

Active gRPC Reflection¶

Strake queries the active server reflection endpoint to discover service types and structure.

Serialized FileDescriptorSet¶

To avoid runtime reflection overhead or support offline catalog building, you can provide a path to a precompiled Protobuf descriptor binary file (descriptor.bin):

# Generate descriptor using protoc
protoc --include_imports --descriptor_set_out=user_service_desc.bin user_service.proto

2. Configuration Parameters¶

The gRPC connector is registered with type: grpc:

Parameter	Type	Required	Description
`url`	string	Yes	Active gRPC server endpoint (e.g. `http://localhost:50051`).
`service`	string	Yes	Fully qualified service name (e.g. `package.Service`).
`method`	string	Yes	RPC method to invoke.
`descriptor_set`	string	No	Path to the local precompiled FileDescriptorSet binary file.
`request_body`	string	No	JSON string body to serve as the default RPC request payload.

3. Configuration Snippet¶

Add the following block to your sources.yaml to register a gRPC microservice connector:

sources:
  - name: user_service_grpc
    type: grpc
    config:
      url: "http://user-service.internal:50051"
      service: "my.company.UserService"
      method: "GetActiveUsers"
      descriptor_set: "/workspaces/rust-postgres/data/user_service_desc.bin"
      request_body: '{"status": "ACTIVE"}'
      columns:
        - name: "user_id"
          type: "Int64"
        - name: "email"
          type: "Utf8"