@dickpy/dsh-imagegen 1.4.0 → 1.5.1

package/src/protocol.ts

@@ -1,338 +1,380 @@
-/**
- * Wire contract shared by the host and client halves of dsh-imagegen: the
- * settings namespace, the route paths, and the generate payload/result shapes.
- * Pure types + constants 鈥?safe for the client bundle to inline.
- */
-
-/** Settings namespace this plugin owns (host settings seam + bridge). */
-export const IMAGEGEN_SETTINGS_NAMESPACE = 'dsh-imagegen'
-
-/** Published package version shared by the host updater and the client UI. */
-export const PLUGIN_VERSION = '1.3.1'
-
-/** Same-origin route family (loopback-only, mirroring the dsh-ssh fence). */
-export const SETTINGS_API = {
-  describe: '/api/dsh-imagegen/settings/describe',
-  mutate: '/api/dsh-imagegen/settings/mutate',
-} as const
-
-/** The image-generation proxy route. */
-export const GENERATE_API = '/api/dsh-imagegen/generate'
-
-/** Host-mediated OpenAI-compatible prompt enhancement endpoints. */
-export const PROMPT_ENHANCE_API = {
-  models: '/api/dsh-imagegen/prompt-enhance/models',
-  enhance: '/api/dsh-imagegen/prompt-enhance',
-} as const
-
-/** Host-mediated candidate discovery for the configured image API. */
-export const IMAGE_MODEL_API = {
-  models: '/api/dsh-imagegen/image-models',
-} as const
-
-/** Host-served built-in provider catalog (channels the user can instantiate). */
-export const PRESETS_API = '/api/dsh-imagegen/presets' as const
-
-/** Loopback-only image reader for Agent tool-result previews. */
-export const AGENT_IMAGE_API = '/api/dsh-imagegen/agent-image' as const
-
-/**
- * Host-computed per-channel usage counters (generation-count badges in the
- * settings card): entries are tallied from the persisted history and gallery
- * by channel + model alias.
- */
-export const USAGE_API = '/api/dsh-imagegen/usage' as const
-
-/** Host-resident generation queue endpoints. */
-export const TASK_API = {
-  submit: '/api/dsh-imagegen/tasks/submit',
-  list: '/api/dsh-imagegen/tasks/list',
-  cancel: '/api/dsh-imagegen/tasks/cancel',
-  retry: '/api/dsh-imagegen/tasks/retry',
-} as const
-
-/** Host-mediated GitHub Release update routes. */
-export const UPDATE_API = {
-  check: '/api/dsh-imagegen/update/check',
-  apply: '/api/dsh-imagegen/update/apply',
-} as const
-
-/**
- * Same-origin route family for the host-persisted generation history. Images
- * live as files under ~/.dsh/dsh-imagegen/images/ and are served back through
- * the `image` prefix route, so list responses carry metadata only (never
- * base64) and the browser loads thumbnails/previews lazily.
- */
-export const HISTORY_API = {
-  list: '/api/dsh-imagegen/history/list',
-  append: '/api/dsh-imagegen/history/append',
-  remove: '/api/dsh-imagegen/history/remove',
-  clear: '/api/dsh-imagegen/history/clear',
-  image: '/api/dsh-imagegen/history/image',
-} as const
-
-/**
- * Same-origin route family for the user-curated gallery (favorites). Entries
- * reuse the history wire shape and persist under ~/.dsh/dsh-imagegen/gallery/;
- * unlike history there is no size cap 鈥?the user adds images on purpose.
- */
-export const GALLERY_API = {
-  list: '/api/dsh-imagegen/gallery/list',
-  append: '/api/dsh-imagegen/gallery/append',
-  remove: '/api/dsh-imagegen/gallery/remove',
-  clear: '/api/dsh-imagegen/gallery/clear',
-  tags: '/api/dsh-imagegen/gallery/tags',
-  image: '/api/dsh-imagegen/gallery/image',
-} as const
-
-/** Maximum number of history entries retained host-side (oldest evicted). */
-export const HISTORY_MAX = 50
-
-/**
- * Same-origin route family for the bundled prompt-template library
- * (awesome-gpt-image-2 mirror). The case list ships inside the package and is
- * served by the host; reference images are proxied through the `image` prefix
- * route and cached on disk so repeated views never hit the network again.
- */
-export const TEMPLATES_API = {
-  list: '/api/dsh-imagegen/templates/list',
-  refresh: '/api/dsh-imagegen/templates/refresh',
-  image: '/api/dsh-imagegen/templates/image',
-} as const
-
-/** One prompt-library case as the browser consumes it. */
-export interface TemplateCase {
-  /** Upstream case number (stable across refreshes). */
-  id: number
-  /** Short case title. */
-  title: string
-  /** Full reusable prompt text. */
-  prompt: string
-  /** English category name (grouping key). */
-  category: string
-  /** Chinese category display name. */
-  categoryZh: string
-  /** Style tags. */
-  styles: string[]
-  /** Scene tags. */
-  scenes: string[]
-  /** Original author handle, e.g. @vista8. */
-  sourceLabel: string
-  /** Original author link. */
-  sourceUrl: string
-  /** awesome-gpt-image-2 repo anchor link. */
-  githubUrl: string
-  /** Reference-image file name served through the image route ('' when none). */
-  image: string
-  /** Whether the source gallery featured the case. */
-  featured: boolean
-}
-
-/** Template-library list payload. */
-export interface TemplateListResult {
-  cases: TemplateCase[]
-  total: number
-  /** Where the served list came from. */
-  origin: 'bundled' | 'refreshed'
-  /** Upstream repository the library mirrors. */
-  repository: string
-  /** ISO time of the last successful refresh / bundle snapshot. */
-  fetchedAt: string
-}
-
-/** Template-library refresh outcome. */
-export interface TemplateRefreshResult {
-  total: number
-  fetchedAt: string
-}
-
-/** Generation modes. */
-export type GenerateMode = 'text' | 'edit'
-
-/** A client → host generate request (what the panel collects). */
-export interface GenerateRequest {
-  /** text-to-image (images/generations) or image-to-image (images/edits). */
-  mode: GenerateMode
-  /**
-   * User-facing model name (an alias from the channel's model catalog). The
-   * host maps it onto the configured channel and fills `upstream` with the
-   * real id before the engine sees it.
-   */
-  model: string
-  /** The prompt. Upstream providers may impose their own length limits. */
-  prompt: string
-  /** Canvas size as an aspect ratio: 'auto' or e.g. '1:1' / '16:9' / '21:9'.
-   *  The host maps it onto each model's own vocabulary (aspect_ratio for Grok
-   *  and Nano Banana, resolution-tier size for Seedream, the closest pixel size for
-   *  OpenAI-compatible endpoints). */
-  size: string
-  /** Clarity tier: 'auto' | '1k' | '2k' | '4k'. The host maps it onto the
-   *  model's own vocabulary (resolution for Grok, image_size for Nano Banana,
-   *  and size for Seedream,
-   *  Nano Banana, quality for OpenAI). */
-  quality: string
-  /** Number of images, 1-4. */
-  n: number
-  /**
-   * Passthrough detail parameter: '' (omit), 'standard', or 'high'. Some
-   * gpt-image-2 gateways expose it; official OpenAI endpoints reject unknown
-   * parameters, so the UI defaults to '' (omit).
-   */
-  detail: string
-  /** Reference image as a data URL (edit mode only). */
-  image?: string
-  /** Original reference-image name, retained in the history entry. */
-  refName?: string
-  /** Channel this request targets (the host falls back to the default when
-   *  absent, and re-routes by model alias when the alias lives elsewhere). */
-  channelId?: string
-  /** Channel display name snapshot, kept on the history entry (host-filled). */
-  channel?: string
-  /** Upstream model id actually sent to the gateway (host-filled from the
-   *  alias mapping; defaults to `model` when absent). */
-  upstream?: string
-  /** Stable client-created id shared by the tasks in one comparison run. */
-  comparisonId?: string
-  /** All model aliases selected for one comparison run. */
-  comparisonModels?: string[]
-}
-
-/** One generated image, normalized host-side to base64 so the browser never
- *  has to fetch the upstream (no CORS, no key exposure). */
-export interface GeneratedImage {
-  /** Raw base64 payload (no data: prefix). */
-  b64: string
-  /** MIME type of the payload, e.g. image/png. */
-  mime: string
-  /** Upstream revised prompt, when provided. */
-  revisedPrompt?: string
-}
-
-/** Successful generate outcome. */
-export interface GenerateResult {
-  images: GeneratedImage[]
-  /** Updated host-persisted history, when returned by the generate route. */
-  history?: HistoryEntry[]
-  /** Persistence failure after images were successfully generated. */
-  historyError?: string
-}
-
-/**
- * One model mapping in a channel's catalog: the display alias the user, the
- * panel, and the Agent see, and the upstream model id actually sent to the
- * gateway. The alias defaults to the upstream id but can be renamed freely.
- */
-export interface ModelMapping {
-  /** User-facing model name (defaults to the upstream id). */
-  alias: string
-  /** Upstream model id sent to the gateway. */
-  id: string
-}
-
-/**
- * One configured image channel (provider). Secrets never live here — the API
- * key is stored at `channelSecrets.<channelId>` in the settings document so
- * whole-array writes can never clobber keys the user did not re-enter.
- */
-export interface ChannelConfig {
-  /** Stable channel id (the channelSecrets dict is keyed by it). */
-  id: string
-  /** Preset provider id this channel was created from ('' = custom). */
-  preset: string
-  /** Display name shown in the list, the panel, and Agent guidance. */
-  name: string
-  /** OpenAI-compatible base URL. */
-  apiUrl: string
-  /** The channel's model catalog (alias → upstream id). */
-  models: ModelMapping[]
-}
-
-/** One built-in provider as the settings card consumes it. */
-export interface PresetProviderView {
-  id: string
-  name: string
-  apiUrl: string
-  hint: string
-  models: ModelMapping[]
-}
-
-export type GenerationTaskStatus = 'queued' | 'running' | 'completed' | 'failed' | 'cancelled'
-
-export interface GenerationTask {
-  id: string
-  request: GenerateRequest
-  status: GenerationTaskStatus
-  createdAt: number
-  startedAt?: number
-  finishedAt?: number
-  result?: GenerateResult
-  error?: string
-}
-
-/** GitHub Release update information shown by the client. */
-export interface UpdateInfo {
-  currentVersion: string
-  latestVersion: string
-  updateAvailable: boolean
-  releaseUrl: string
-  publishedAt?: string
-}
-
-/** One history image reference as the browser consumes it (a served URL). */
-export interface HistoryImageRef {
-  /** Same-origin URL: `${HISTORY_API.image}/<file>`. */
-  url: string
-  /** MIME type, e.g. image/png. */
-  mime: string
-  /** Upstream revised prompt, when provided. */
-  revisedPrompt?: string
-}
-
-/** A saved generation as the browser consumes it (metadata + served images). */
-export interface HistoryEntry {
-  id: string
-  createdAt: number
-  mode: GenerateMode
-  model: string
-  prompt: string
-  size: string
-  quality: string
-  detail: string
-  n: number
-  images: HistoryImageRef[]
-  /** Reference-image filename (edit mode), kept for display only. */
-  refName?: string
-  /** User-managed gallery labels (unused by history entries). */
-  tags?: string[]
-  /** Channel id snapshot (usage counters key by it for new entries). */
-  channelId?: string
-  /** Channel display name snapshot (survives channel deletion). */
-  channel?: string
-  /** Stable id shared by the history entries in one comparison run. */
-  comparisonId?: string
-  /** Model aliases included in the comparison run. */
-  comparisonModels?: string[]
-}
-
-/** A history entry the client submits for persistence (images still carry base64). */
-export interface HistoryEntryInput {
-  id: string
-  createdAt: number
-  mode: GenerateMode
-  model: string
-  prompt: string
-  size: string
-  quality: string
-  detail: string
-  n: number
-  images: GeneratedImage[]
-  refName?: string
-  /** Channel id snapshot, tallied by the usage endpoint. */
-  channelId?: string
-  /** Channel display name snapshot (survives channel deletion). */
-  channel?: string
-  /** Stable id shared by the history entries in one comparison run. */
-  comparisonId?: string
-  /** Model aliases included in the comparison run. */
-  comparisonModels?: string[]
-}
+/**
+ * Wire contract shared by the host and client halves of dsh-imagegen: the
+ * settings namespace, the route paths, and the generate payload/result shapes.
+ * Pure types + constants 鈥?safe for the client bundle to inline.
+ */
+
+/** Settings namespace this plugin owns (host settings seam + bridge). */
+export const IMAGEGEN_SETTINGS_NAMESPACE = 'dsh-imagegen'
+
+/** Published package version shared by the host updater and the client UI. */
+export const PLUGIN_VERSION = '1.5.1'
+
+/** Same-origin route family (loopback-only, mirroring the dsh-ssh fence). */
+export const SETTINGS_API = {
+  describe: '/api/dsh-imagegen/settings/describe',
+  mutate: '/api/dsh-imagegen/settings/mutate',
+} as const
+
+/** The image-generation proxy route. */
+export const GENERATE_API = '/api/dsh-imagegen/generate'
+
+/** Host-mediated OpenAI-compatible prompt enhancement endpoints. */
+export const PROMPT_ENHANCE_API = {
+  models: '/api/dsh-imagegen/prompt-enhance/models',
+  enhance: '/api/dsh-imagegen/prompt-enhance',
+} as const
+
+/** Host-mediated candidate discovery for the configured image API. */
+export const IMAGE_MODEL_API = {
+  models: '/api/dsh-imagegen/image-models',
+} as const
+
+/** Host-served built-in provider catalog (channels the user can instantiate). */
+export const PRESETS_API = '/api/dsh-imagegen/presets' as const
+
+/** Loopback-only image reader for Agent tool-result previews. */
+export const AGENT_IMAGE_API = '/api/dsh-imagegen/agent-image' as const
+
+/** Store the current composer image for the direct edit_image command. */
+export const CONVERSATION_IMAGE_API = '/api/dsh-imagegen/conversation-image' as const
+
+/**
+ * Host-computed per-channel usage counters (generation-count badges in the
+ * settings card): entries are tallied from the persisted history and gallery
+ * by channel + model alias.
+ */
+export const USAGE_API = '/api/dsh-imagegen/usage' as const
+
+/** Host-resident generation queue endpoints. */
+export const TASK_API = {
+  submit: '/api/dsh-imagegen/tasks/submit',
+  list: '/api/dsh-imagegen/tasks/list',
+  cancel: '/api/dsh-imagegen/tasks/cancel',
+  retry: '/api/dsh-imagegen/tasks/retry',
+} as const
+
+/** Host-mediated GitHub Release update routes. */
+export const UPDATE_API = {
+  check: '/api/dsh-imagegen/update/check',
+  apply: '/api/dsh-imagegen/update/apply',
+} as const
+
+/**
+ * Same-origin route family for the host-persisted generation history. Images
+ * live as files under ~/.dsh/dsh-imagegen/images/ and are served back through
+ * the `image` prefix route, so list responses carry metadata only (never
+ * base64) and the browser loads thumbnails/previews lazily.
+ */
+export const HISTORY_API = {
+  list: '/api/dsh-imagegen/history/list',
+  append: '/api/dsh-imagegen/history/append',
+  remove: '/api/dsh-imagegen/history/remove',
+  clear: '/api/dsh-imagegen/history/clear',
+  image: '/api/dsh-imagegen/history/image',
+} as const
+
+/**
+ * Same-origin route family for the user-curated gallery (favorites). Entries
+ * reuse the history wire shape and persist under ~/.dsh/dsh-imagegen/gallery/;
+ * unlike history there is no size cap 鈥?the user adds images on purpose.
+ */
+export const GALLERY_API = {
+  list: '/api/dsh-imagegen/gallery/list',
+  append: '/api/dsh-imagegen/gallery/append',
+  remove: '/api/dsh-imagegen/gallery/remove',
+  clear: '/api/dsh-imagegen/gallery/clear',
+  tags: '/api/dsh-imagegen/gallery/tags',
+  image: '/api/dsh-imagegen/gallery/image',
+} as const
+
+/** Maximum number of history entries retained host-side (oldest evicted). */
+export const HISTORY_MAX = 50
+
+/**
+ * Same-origin route family for the bundled prompt-template library
+ * (awesome-gpt-image-2 mirror). The case list ships inside the package and is
+ * served by the host; reference images are proxied through the `image` prefix
+ * route and cached on disk so repeated views never hit the network again.
+ */
+export const TEMPLATES_API = {
+  list: '/api/dsh-imagegen/templates/list',
+  refresh: '/api/dsh-imagegen/templates/refresh',
+  image: '/api/dsh-imagegen/templates/image',
+} as const
+
+/** One prompt-library case as the browser consumes it. */
+export interface TemplateCase {
+  /** Upstream case number (stable across refreshes). */
+  id: number
+  /** Short case title. */
+  title: string
+  /** Full reusable prompt text. */
+  prompt: string
+  /** English category name (grouping key). */
+  category: string
+  /** Chinese category display name. */
+  categoryZh: string
+  /** Style tags. */
+  styles: string[]
+  /** Scene tags. */
+  scenes: string[]
+  /** Original author handle, e.g. @vista8. */
+  sourceLabel: string
+  /** Original author link. */
+  sourceUrl: string
+  /** awesome-gpt-image-2 repo anchor link. */
+  githubUrl: string
+  /** Reference-image file name served through the image route ('' when none). */
+  image: string
+  /** Whether the source gallery featured the case. */
+  featured: boolean
+}
+
+/** Template-library list payload. */
+export interface TemplateListResult {
+  cases: TemplateCase[]
+  total: number
+  /** Where the served list came from. */
+  origin: 'bundled' | 'refreshed'
+  /** Upstream repository the library mirrors. */
+  repository: string
+  /** ISO time of the last successful refresh / bundle snapshot. */
+  fetchedAt: string
+}
+
+/** Template-library refresh outcome. */
+export interface TemplateRefreshResult {
+  total: number
+  fetchedAt: string
+}
+
+/** Generation modes. */
+export type GenerateMode = 'text' | 'edit'
+
+/** Metadata shared by the ecommerce product-set workflow. */
+export interface EcommerceTaskMeta {
+  workflow?: 'ecommerce'
+  projectId?: string
+  projectName?: string
+  slotKey?: string
+  slotLabel?: string
+}
+
+/** Role an uploaded product asset plays in the ecommerce workflow. 'none' is
+ *  only used as a slot selection meaning "generate without a reference". */
+export type EcommerceRefRole = 'none' | 'product' | 'packaging' | 'detail' | 'style'
+
+/** One planned image slot in a product set. */
+export interface ProductSetSlot {
+  key: string
+  label: string
+  description: string
+  count: number
+  enabled: boolean
+  /** Which uploaded asset role this slot uses as its edit reference. */
+  refRole?: EcommerceRefRole
+}
+
+/** A browser-local ecommerce product-set draft. */
+export interface ProductSetDraft {
+  projectId: string
+  projectName: string
+  category: string
+  platform: string
+  language: string
+  size: string
+  productName: string
+  sellingPoints: string
+  protectedFeatures: string
+  styleHint: string
+  slots: ProductSetSlot[]
+}
+
+/** A client → host generate request (what the panel collects). */
+export interface GenerateRequest extends EcommerceTaskMeta {
+  /** text-to-image (images/generations) or image-to-image (images/edits). */
+  mode: GenerateMode
+  /**
+   * User-facing model name (an alias from the channel's model catalog). The
+   * host maps it onto the configured channel and fills `upstream` with the
+   * real id before the engine sees it.
+   */
+  model: string
+  /** The prompt. Upstream providers may impose their own length limits. */
+  prompt: string
+  /** Canvas size as an aspect ratio: 'auto' or e.g. '1:1' / '16:9' / '21:9'.
+   *  The host maps it onto each model's own vocabulary (aspect_ratio for Grok
+   *  and Nano Banana, resolution-tier size for Seedream, the closest pixel size for
+   *  OpenAI-compatible endpoints). */
+  size: string
+  /** Clarity tier: 'auto' | '1k' | '2k' | '4k'. The host maps it onto the
+   *  model's own vocabulary (resolution for Grok, image_size for Nano Banana,
+   *  and size for Seedream,
+   *  Nano Banana, quality for OpenAI). */
+  quality: string
+  /** Number of images, 1-4. */
+  n: number
+  /**
+   * Passthrough detail parameter: '' (omit), 'standard', or 'high'. Some
+   * gpt-image-2 gateways expose it; official OpenAI endpoints reject unknown
+   * parameters, so the UI defaults to '' (omit).
+   */
+  detail: string
+  /** Reference image as a data URL (edit mode only). */
+  image?: string
+  /** Original reference-image name, retained in the history entry. */
+  refName?: string
+  /** Channel this request targets (the host falls back to the default when
+   *  absent, and re-routes by model alias when the alias lives elsewhere). */
+  channelId?: string
+  /** Channel display name snapshot, kept on the history entry (host-filled). */
+  channel?: string
+  /** Upstream model id actually sent to the gateway (host-filled from the
+   *  alias mapping; defaults to `model` when absent). */
+  upstream?: string
+  /** Stable client-created id shared by the tasks in one comparison run. */
+  comparisonId?: string
+  /** All model aliases selected for one comparison run. */
+  comparisonModels?: string[]
+}
+
+/** One generated image, normalized host-side to base64 so the browser never
+ *  has to fetch the upstream (no CORS, no key exposure). */
+export interface GeneratedImage {
+  /** Raw base64 payload (no data: prefix). */
+  b64: string
+  /** MIME type of the payload, e.g. image/png. */
+  mime: string
+  /** Upstream revised prompt, when provided. */
+  revisedPrompt?: string
+}
+
+/** Successful generate outcome. */
+export interface GenerateResult {
+  images: GeneratedImage[]
+  /** Updated host-persisted history, when returned by the generate route. */
+  history?: HistoryEntry[]
+  /** Persistence failure after images were successfully generated. */
+  historyError?: string
+}
+
+/**
+ * One model mapping in a channel's catalog: the display alias the user, the
+ * panel, and the Agent see, and the upstream model id actually sent to the
+ * gateway. The alias defaults to the upstream id but can be renamed freely.
+ */
+export interface ModelMapping {
+  /** User-facing model name (defaults to the upstream id). */
+  alias: string
+  /** Upstream model id sent to the gateway. */
+  id: string
+}
+
+/**
+ * One configured image channel (provider). Secrets never live here — the API
+ * key is stored at `channelSecrets.<channelId>` in the settings document so
+ * whole-array writes can never clobber keys the user did not re-enter.
+ */
+export interface ChannelConfig {
+  /** Stable channel id (the channelSecrets dict is keyed by it). */
+  id: string
+  /** Preset provider id this channel was created from ('' = custom). */
+  preset: string
+  /** Display name shown in the list, the panel, and Agent guidance. */
+  name: string
+  /** OpenAI-compatible base URL. */
+  apiUrl: string
+  /** The channel's model catalog (alias → upstream id). */
+  models: ModelMapping[]
+}
+
+/** One built-in provider as the settings card consumes it. */
+export interface PresetProviderView {
+  id: string
+  name: string
+  apiUrl: string
+  hint: string
+  models: ModelMapping[]
+}
+
+export type GenerationTaskStatus = 'queued' | 'running' | 'completed' | 'failed' | 'cancelled'
+
+export interface GenerationTask extends EcommerceTaskMeta {
+  id: string
+  request: GenerateRequest
+  status: GenerationTaskStatus
+  createdAt: number
+  startedAt?: number
+  finishedAt?: number
+  result?: GenerateResult
+  error?: string
+}
+
+/** GitHub Release update information shown by the client. */
+export interface UpdateInfo {
+  currentVersion: string
+  latestVersion: string
+  updateAvailable: boolean
+  releaseUrl: string
+  publishedAt?: string
+}
+
+/** One history image reference as the browser consumes it (a served URL). */
+export interface HistoryImageRef {
+  /** Same-origin URL: `${HISTORY_API.image}/<file>`. */
+  url: string
+  /** MIME type, e.g. image/png. */
+  mime: string
+  /** Upstream revised prompt, when provided. */
+  revisedPrompt?: string
+}
+
+/** A saved generation as the browser consumes it (metadata + served images). */
+export interface HistoryEntry extends EcommerceTaskMeta {
+  id: string
+  createdAt: number
+  mode: GenerateMode
+  model: string
+  prompt: string
+  size: string
+  quality: string
+  detail: string
+  n: number
+  images: HistoryImageRef[]
+  /** Reference-image filename (edit mode), kept for display only. */
+  refName?: string
+  /** User-managed gallery labels (unused by history entries). */
+  tags?: string[]
+  /** Channel id snapshot (usage counters key by it for new entries). */
+  channelId?: string
+  /** Channel display name snapshot (survives channel deletion). */
+  channel?: string
+  /** Stable id shared by the history entries in one comparison run. */
+  comparisonId?: string
+  /** Model aliases included in the comparison run. */
+  comparisonModels?: string[]
+}
+
+/** A history entry the client submits for persistence (images still carry base64). */
+export interface HistoryEntryInput extends EcommerceTaskMeta {
+  id: string
+  createdAt: number
+  mode: GenerateMode
+  model: string
+  prompt: string
+  size: string
+  quality: string
+  detail: string
+  n: number
+  images: GeneratedImage[]
+  refName?: string
+  /** Channel id snapshot, tallied by the usage endpoint. */
+  channelId?: string
+  /** Channel display name snapshot (survives channel deletion). */
+  channel?: string
+  /** Stable id shared by the history entries in one comparison run. */
+  comparisonId?: string
+  /** Model aliases included in the comparison run. */
+  comparisonModels?: string[]
+}