import * as z from "zod";
import { Id } from "./ids.js";

/** What a media job does. Generation runs on the configured image/video backend, edits run locally. */
export const MediaOperationSchema = z.enum([
  "generate_image",
  "generate_video",
  "edit_image",
  "edit_video",
  "export_video",
  "transcribe",
]);
export type MediaOperation = z.infer<typeof MediaOperationSchema>;

/** `waiting` means the job is queued behind another GPU job or for free GPU memory. */
export const MediaJobStatusSchema = z.enum([
  "queued",
  "waiting",
  "running",
  "done",
  "failed",
  "cancelled",
]);
export type MediaJobStatus = z.infer<typeof MediaJobStatusSchema>;

export const MEDIA_JOB_ACTIVE_STATUSES = [
  "queued",
  "waiting",
  "running",
] as const satisfies readonly MediaJobStatus[];

export const MediaOutputKindSchema = z.enum(["image", "video", "audio", "file"]);
export type MediaOutputKind = z.infer<typeof MediaOutputKindSchema>;

/**
 * Output formats for video cuts and exports. Width and height are the encoded frame; every
 * preset encodes H.264 + AAC in MP4 so the result plays everywhere it is posted.
 */
export const MEDIA_PRESETS = {
  vertical: { aspect: "9:16", width: 1080, height: 1920 },
  landscape: { aspect: "16:9", width: 1920, height: 1080 },
  square: { aspect: "1:1", width: 1080, height: 1080 },
  portrait: { aspect: "4:5", width: 1080, height: 1350 },
  classic: { aspect: "4:3", width: 1440, height: 1080 },
  cinema: { aspect: "21:9", width: 2560, height: 1080 },
} as const;
export const MediaPresetIdSchema = z.enum([
  "vertical",
  "landscape",
  "square",
  "portrait",
  "classic",
  "cinema",
]);
export type MediaPresetId = z.infer<typeof MediaPresetIdSchema>;
export type MediaPreset = (typeof MEDIA_PRESETS)[MediaPresetId];

/** How a frame of another aspect ratio fills a preset. */
export const MediaFitSchema = z.enum(["crop", "pad", "blur"]);
export type MediaFit = z.infer<typeof MediaFitSchema>;

export const MediaOutputSchema = z.object({
  artifactId: Id,
  kind: MediaOutputKindSchema,
  mimeType: z.string(),
  name: z.string(),
  size: z.number().int().nonnegative(),
  width: z.number().int().positive().optional(),
  height: z.number().int().positive().optional(),
  durationMs: z.number().int().nonnegative().optional(),
  posterArtifactId: Id.optional(),
  /** The artifact this output was made from, for version history. */
  parentArtifactId: Id.optional(),
  preset: MediaPresetIdSchema.optional(),
});
export type MediaOutput = z.infer<typeof MediaOutputSchema>;

/** The live media card a job keeps up to date in the conversation. */
export const MediaJobBlockSchema = z.object({
  kind: z.literal("media"),
  jobId: Id,
  operation: MediaOperationSchema,
  status: MediaJobStatusSchema,
  /** Short label: the prompt, or what the edit does. */
  title: z.string(),
  progress: z.number().min(0).max(1).optional(),
  error: z.string().optional(),
  outputs: z.array(MediaOutputSchema),
});
export type MediaJobBlock = z.infer<typeof MediaJobBlockSchema>;

export const VideoBlockSchema = z.object({
  kind: z.literal("video"),
  artifactId: Id,
  mimeType: z.string(),
  name: z.string(),
  size: z.number().int().nonnegative(),
  durationMs: z.number().int().nonnegative().optional(),
  width: z.number().int().positive().optional(),
  height: z.number().int().positive().optional(),
  posterArtifactId: Id.optional(),
});
export type VideoBlock = z.infer<typeof VideoBlockSchema>;

export const AudioBlockSchema = z.object({
  kind: z.literal("audio"),
  artifactId: Id,
  mimeType: z.string(),
  name: z.string(),
  size: z.number().int().nonnegative(),
  durationMs: z.number().int().nonnegative().optional(),
});
export type AudioBlock = z.infer<typeof AudioBlockSchema>;