// Copyright 2024 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

// This file is copied from
// https://source.chromium.org/chromiumos/chromiumos/codesearch/+/main:src/platform2/odml/mojom/mantis_processor.mojom
// Doc: go/mantis-bl-dd
module mantis.mojom;

// A specific error Mantis encountered when processing the request.
[Stable, Extensible]
enum MantisError {
  [Default] kUnknownError,
  [MinVersion=1] kProcessorNotInitialized,
  [MinVersion=1] kInputError,
  [MinVersion=1] kProcessFailed,
  [MinVersion=2] kMissingSegmenter,
  [MinVersion=3] kInputSafetyError,
  [MinVersion=3] kOutputSafetyError,
  [MinVersion=3] kPromptSafetyError,
};

// The result of image after processed by Mantis, or an error if any.
[Stable]
union MantisResult {
  MantisError error;
  array<uint8> result_image;
};

// This enum type identifies the result of Trust & Safety checking.
[Stable, Extensible]
enum SafetyClassifierVerdict {
  kPass,
  [Default] kFail,
  [MinVersion=3] kFailedText,
  [MinVersion=3] kFailedImage,
  [MinVersion=4] kServiceNotAvailable,
  [MinVersion=5] kNoInternetConnection,
};

// Struct that is used to represent a single point (x, y) in a
// stroke for use in e.g. handwriting recognition. (x, y) are coordinates
// in world space, with units in Density-Independent Pixels (DIP).
// The event fields are defined in:
// http://google3/third_party/sketchology/proto/scene_change.proto;l=123;rcl=693325121
[Stable]
struct TouchPoint {
  float x@0;
  float y@1;
};

// This enum type identifies the segmentation mode.
[Stable, Extensible]
enum SegmentationMode {
  [Default] kScribble,
  kLasso,
};

// Interface for processing images. The processor should already be initialized
// by `MantisService`.
//
// This interface is served by `odmld` process in CrOS to be used by image
// editing UI (e.g. Media App WebUI).
[Stable]
interface MantisProcessor {
  // Inpaints the image based on the mask and seed. Pass the same `seed` across
  // method calls to get identical result. The `image`, `mask`, and `result` are
  // byte arrays containing the encoded format of an image (e.g., PNG, JPEG).
  Inpainting@0(array<uint8> image, array<uint8> mask, uint32 seed)
      => (MantisResult result);

  // Fills the image generatively based on the text and seed. Pass the same
  // `seed` across method calls to get identical result. The `image`, `mask`,
  // and `result` are byte arrays containing the encoded format of an image
  // (e.g., PNG, JPEG).
  GenerativeFill@1(
      array<uint8> image, array<uint8> mask, uint32 seed, string prompt)
      => (MantisResult result);

  // Performs image segmentation on the image based on the prior selection.
  // The `image`, `prior`, and `result` are byte arrays containing the
  // encoded format of an image (e.g., PNG, JPEG).
  [MinVersion=1]
  Segmentation@2(array<uint8> image, array<uint8> prior)
      => (MantisResult result);

  // Classifies image for Trust & Safety checking.
  [MinVersion=2]
  ClassifyImageSafety@3(array<uint8> image)
      => (SafetyClassifierVerdict verdict);

  // Outpaints the image based on the mask and seed. Pass the same `seed` across
  // method calls to get identical result. The `image`, `mask`, and `result` are
  // byte arrays containing the encoded format of an image (e.g., PNG, JPEG).
  [MinVersion=6]
  Outpainting@4(array<uint8> image, array<uint8> mask, uint32 seed)
      => (MantisResult result);

  // Infers the segmentation mode (e.g., scribble or lasso) from a list of
  // touch events.
  [MinVersion=7]
  InferSegmentationMode@5(array<TouchPoint> gesture)
      => (SegmentationMode mode);
};