// Decart Action-to-Video Service - gRPC API (A2V)
//
// Action-to-video inference: the model has no video input. Each Infer call
// consumes 4 driving actions ({throttle, steering}) and returns 4 generated
// video frames. The server runs the physics (DrivingActionController) and
// produces the 4D AdaLN action consumed by the WM model.
//
// Lifecycle:
// 1. Initialize - Allocate session, negotiate output frame format
// 2. Prompt - Set the text prompt (scene/style)
// 3. Infer (loop) - Send 4 actions, receive 4 frames
// 4. Finish - Release resources, get statistics
//
// Session Management:
// - session_id is a token, not tied to TCP connection
// - If gRPC connection drops, reconnect and use the same session_id
//
syntax = "proto3";
package decart.inference.a2v;

option go_package = "github.com/decart/api/inference/a2v";

import "common.proto";

service A2VService {
  rpc Initialize(InitializeRequest) returns (InitializeResponse);
  rpc Prompt(PromptRequest) returns (PromptResponse);
  rpc Infer(InferRequest) returns (InferResponse);
  rpc Finish(FinishRequest) returns (FinishResponse);
}

message InitializeRequest {
  string client_sdk_version = 1;
  // Decart API key. Required; ERROR_CODE_INVALID_API_KEY.
  string api_key = 2;
  repeated decart.inference.common.FrameFormat accepted_output_formats = 3;
}

message InitializeResponse {
  optional decart.inference.common.Error error = 1;
  string session_id = 2;
  decart.inference.common.FrameFormat output_format = 3;
  repeated StreamInfo streams = 4;
}

enum SensorType {
  SENSOR_TYPE_UNSPECIFIED = 0;
  SENSOR_TYPE_RGB = 1;
  SENSOR_TYPE_LIDAR = 2;
  SENSOR_TYPE_RADAR = 3;
}

message StreamInfo {
  string name = 1;
  SensorType type = 2;
  uint32 height = 3;
  uint32 width = 4;
}

message PromptRequest {
  string session_id = 1;
  string prompt = 2;
}

message PromptResponse {
  optional decart.inference.common.Error error = 1;
}

message Action {
  float throttle = 1;
  float steering = 2;
}

message InferRequest {
  string session_id = 1;
  uint64 sequence_num = 2;
  repeated Action actions = 3;
}

message OutputStream {
  string name = 1;
  repeated bytes frames = 2;
}

message InferResponse {
  uint64 sequence_num = 1;
  optional decart.inference.common.Error error = 2;
  repeated OutputStream streams = 3;
}

message FinishRequest {
  string session_id = 1;
}

message FinishResponse {
  optional decart.inference.common.Error error = 1;
}