Skip to content

The VS Code of video editing, built for agents.

TypeScript

Refactor generation configuration and type handling

konstantin-pauluse3af57b13671
diff --git a/apps/web/src/components/genai/action-bar.tsx b/apps/web/src/components/genai/action-bar.tsx
index 79f21405..467547c8 100644
--- a/apps/web/src/components/genai/action-bar.tsx
+++ b/apps/web/src/components/genai/action-bar.tsx
@@ -19,10 +19,10 @@ import { useGenerate } from "./use-generate";
import { useAutoCaptions } from "./use-auto-captions";
import { useMediaSelection } from "./selection";
import { useTransforms } from "./use-transforms";
-import { createDefaultConfig } from "./prompt-input";
+import { createDefaultConfig } from "./requests";
import { toast } from "somoto";
-import type { GenerationConfig } from "./schemas";
+import type { GenerationConfig } from "./types";
interface ActionBarProps {
openPromptInput?(config: GenerationConfig): void;
diff --git a/apps/web/src/components/genai/config.ts b/apps/web/src/components/genai/config.ts
index acc00c83..216461df 100644
--- a/apps/web/src/components/genai/config.ts
+++ b/apps/web/src/components/genai/config.ts
@@ -36,9 +36,6 @@ export const MODEL_MODES: Record<ModelId, PromptMode | "TEXT"> = {
export type PromptMode = "IMAGE" | "VIDEO" | "VOICE" | "AUDIO";
-/** The one model the voice mode speaks with; the UI picks the voice, not the model. */
-export const PROMPT_INPUT_VOICE_MODEL = "elevenlabs-v3" satisfies ModelId;
-
/** What an aspect ratio is worth in pixels, at 1080p. */
export const ASPECT_RATIO_DIMENSIONS: Record<string, { width: number; height: number }> = {
"16:9": { width: 1920, height: 1080 },
@@ -48,22 +45,17 @@ export const ASPECT_RATIO_DIMENSIONS: Record<string, { width: number; height: nu
"3:4": { width: 1080, height: 1440 },
};
-export const PROMPT_INPUT_MODE_OPTIONS = [
+export const PROMPT_INPUT_MODE_OPTIONS: { value: PromptMode; label: string; icon: string }[] = [
{ value: "IMAGE", label: "Image", icon: "image" },
{ value: "VIDEO", label: "Video", icon: "film-video-export" },
{ value: "VOICE", label: "Voice", icon: "voice" },
{ value: "AUDIO", label: "Audio", icon: "audio" },
];
-export const PROMPT_INPUT_IMAGE_ASPECT_RATIO_OPTIONS = [
- { value: "16:9", label: "16:9 \u00B7 Wide", triggerLabel: "16:9", icon: "aspect-ratio-16-9" },
- { value: "9:16", label: "9:16 \u00B7 Vertical", triggerLabel: "9:16", icon: "aspect-ratio-9-16" },
- { value: "1:1", label: "1:1 \u00B7 Square", triggerLabel: "1:1", icon: "aspect-ratio-1-1" },
- { value: "4:3", label: "4:3 \u00B7 Classic", triggerLabel: "4:3", icon: "aspect-ratio-4-3" },
- { value: "3:4", label: "3:4 \u00B7 Tall", triggerLabel: "3:4", icon: "aspect-ratio-3-4" },
-];
+/** The aspect ratios the prompt box offers; a model may take fewer. */
+export type AspectRatio = "16:9" | "9:16" | "1:1" | "4:3" | "3:4";
-export const ALL_VIDEO_ASPECT_RATIO_OPTIONS = [
+export const ASPECT_RATIO_OPTIONS: { value: AspectRatio; label: string; triggerLabel: string; icon: string }[] = [
{ value: "16:9", label: "16:9 \u00B7 Wide", triggerLabel: "16:9", icon: "aspect-ratio-16-9" },
{ value: "9:16", label: "9:16 \u00B7 Vertical", triggerLabel: "9:16", icon: "aspect-ratio-9-16" },
{ value: "1:1", label: "1:1 \u00B7 Square", triggerLabel: "1:1", icon: "aspect-ratio-1-1" },
@@ -71,29 +63,6 @@ export const ALL_VIDEO_ASPECT_RATIO_OPTIONS = [
{ value: "3:4", label: "3:4 \u00B7 Tall", triggerLabel: "3:4", icon: "aspect-ratio-3-4" },
];
-export const PROMPT_INPUT_VARIANT_COUNT_OPTIONS = [
- { value: "1", label: "1" },
- { value: "2", label: "2" },
- { value: "3", label: "3" },
- { value: "4", label: "4" },
-];
-
-export const ALL_DURATION_OPTIONS = [
- { value: "3s", label: "3s" },
- { value: "4s", label: "4s" },
- { value: "5s", label: "5s" },
- { value: "6s", label: "6s" },
- { value: "7s", label: "7s" },
- { value: "8s", label: "8s" },
- { value: "9s", label: "9s" },
- { value: "10s", label: "10s" },
- { value: "11s", label: "11s" },
- { value: "12s", label: "12s" },
- { value: "13s", label: "13s" },
- { value: "14s", label: "14s" },
- { value: "15s", label: "15s" },
-];
-
export const PROMPT_INPUT_RESOLUTION_OPTIONS = [
{ value: "4K", label: "4K" },
{ value: "1440p", label: "1440p" },
@@ -101,138 +70,6 @@ export const PROMPT_INPUT_RESOLUTION_OPTIONS = [
{ value: "720p", label: "720p" },
];
-export type ImageModelOption = {
- name: string;
- id: ModelId;
- icon: string;
- description: string;
- maxReferences: number;
-};
-
-export const PROMPT_INPUT_IMAGE_MODEL_OPTIONS: ImageModelOption[] = [
- {
- name: "FLUX.2 Klein",
- id: "flux-2-klein",
- icon: "large-bfl",
- description: "Low budget, high quality, fast.",
- maxReferences: 4,
- },
- {
- name: "GPT Image 2",
- id: "gpt-image-2",
- icon: "large-openai",
- description: "Flexible sizes up to 4K, true aspect ratios.",
- maxReferences: 16,
- },
- {
- name: "Nano Banana 2",
- id: "nano-banana-2",
- icon: "large-google",
- description: "Fast, high quality, flexible edits.",
- maxReferences: 14,
- },
- {
- name: "Nano Banana Pro",
- id: "nano-banana-pro",
- icon: "large-google",
- description: "Pro control, readable text, ultra consistent.",
- maxReferences: 14,
- },
- {
- name: "Seedream 4.5",
- id: "seedream-4.5",
- icon: "large-bytedance",
- description: "Up to 4K, multi image edits, strong text.",
- maxReferences: 14,
- },
-];
-
-/**
- * What a video model takes besides the prompt: a first and a last frame, and
- * a switch for the sound it makes (models without one score their video
- * themselves, or not at all).
- */
-export type VideoModelFeature = "start-frame" | "end-frame" | "audio";
-
-export type VideoModelOption = {
- name: string;
- id: ModelId;
- icon: string;
- description: string;
- features: VideoModelFeature[];
- durations: string[];
- /** Empty for a model that takes its proportions from the start frame. */
- aspectRatios: string[];
-};
-
-const seconds = (from: number, to: number) => Array.from({ length: to - from + 1 }, (_, i) => `${from + i}s`);
-
-export const PROMPT_INPUT_VIDEO_MODEL_OPTIONS: VideoModelOption[] = [
- {
- name: "Kling 3.0",
- id: "kling-3-pro",
- icon: "large-kling",
- description: "Cinematic motion with built-in audio.",
- features: ["start-frame", "end-frame", "audio"],
- durations: seconds(3, 15),
- aspectRatios: ["16:9", "9:16", "1:1"],
- },
- {
- name: "Kling 3.0 Omni",
- id: "kling-o3-pro",
- icon: "large-kling",
- description: "Multi-modal reasoning, strong scenes.",
- features: ["start-frame", "audio"],
- durations: seconds(3, 15),
- aspectRatios: [],
- },
- {
- name: "Seedance 2.0",
- id: "seedance-2.0",
- icon: "large-bytedance",
- description: "Rich motion, lip-synced audio and SFX.",
- features: ["start-frame", "end-frame"],
- durations: seconds(4, 15),
- aspectRatios: ["16:9", "9:16", "1:1", "4:3", "3:4"],
- },
- {
- name: "Veo 3.1",
- id: "veo-3.1",
- icon: "large-google",
- description: "Realistic physics, complex scenes.",
- features: ["start-frame", "end-frame"],
- durations: ["4s", "6s", "8s"],
- aspectRatios: ["16:9", "9:16"],
- },
- {
- name: "Veo 3.1 Fast",
- id: "veo-3.1-fast",
- icon: "large-google",
- description: "Same Veo engine, lower cost.",
- features: ["start-frame", "end-frame"],
- durations: ["4s", "6s", "8s"],
- aspectRatios: ["16:9", "9:16"],
- },
- {
- name: "Wan 2.6",
- id: "wan-2.6",
- icon: "large-wan",
- description: "Expressive motion, stylized aesthetics.",
- features: ["start-frame"],
- durations: ["5s", "10s"],
- aspectRatios: ["16:9", "9:16"],
- },
- {
- name: "Hailuo 3 Max",
- id: "hailuo-3-max",
- icon: "large-hailuo",
- description: "Smooth motion, cinematic lighting.",
- features: ["start-frame", "end-frame"],
- durations: seconds(5, 15),
- aspectRatios: ["16:9", "9:16", "1:1", "4:3", "3:4"],
- },
-];
-
export const PROMPT_INPUT_VOICE_OPTIONS = [
{
value: "CwhRBWXzGAHq8TQ4Fs17",
@@ -397,17 +234,176 @@ export const PROMPT_INPUT_VOICE_OPTIONS = [
},
];
-export const PROMPT_INPUT_AUDIO_MODEL_OPTIONS: { name: string; id: ModelId; icon: string; description: string }[] = [
+export type VoiceOption = (typeof PROMPT_INPUT_VOICE_OPTIONS)[number];
+
+/**
+ * A model the prompt box makes requests with, and the settings it takes: a
+ * setting the model declares is offered (within the values given), one it
+ * leaves out is neither offered nor sent. The settings are the API's request
+ * fields, so a config becomes a request without knowing the model (see
+ * `requests.ts`).
+ */
+export interface ModelOption {
+ id: ModelId;
+ mode: PromptMode;
+ name: string;
+ icon: string;
+ description: string;
+ /** Empty for a model that takes its proportions from the start frame. */
+ aspectRatios?: AspectRatio[];
+ /** Images per generation. */
+ counts?: number[];
+ /** Seconds. */
+ durations?: number[];
+ voices?: VoiceOption[];
+ /** How many reference images the model takes. */
+ references?: number;
+ /** The frames a video can be made to start or end on. */
+ frames?: ("start" | "end")[];
+}
+
+const ALL_ASPECT_RATIOS = ASPECT_RATIO_OPTIONS.map((option) => option.value);
+const seconds = (from: number, to: number) => Array.from({ length: to - from + 1 }, (_, i) => from + i);
+
+const image = { mode: "IMAGE", aspectRatios: ALL_ASPECT_RATIOS, counts: [1, 2, 3, 4] } satisfies Partial<ModelOption>;
+const video = { mode: "VIDEO" } satisfies Partial<ModelOption>;
+
+export const MODEL_OPTIONS: ModelOption[] = [
+ {
+ ...image,
+ id: "flux-2-klein",
+ name: "FLUX.2 Klein",
+ icon: "large-bfl",
+ description: "Low budget, high quality, fast.",
+ references: 4,
+ },
+ {
+ ...image,
+ id: "gpt-image-2",
+ name: "GPT Image 2",
+ icon: "large-openai",
+ description: "Flexible sizes up to 4K, true aspect ratios.",
+ references: 16,
+ },
+ {
+ ...image,
+ id: "nano-banana-2",
+ name: "Nano Banana 2",
+ icon: "large-google",
+ description: "Fast, high quality, flexible edits.",
+ references: 14,
+ },
+ {
+ ...image,
+ id: "nano-banana-pro",
+ name: "Nano Banana Pro",
+ icon: "large-google",
+ description: "Pro control, readable text, ultra consistent.",
+ references: 14,
+ },
+ {
+ ...image,
+ id: "seedream-4.5",
+ name: "Seedream 4.5",
+ icon: "large-bytedance",
+ description: "Up to 4K, multi image edits, strong text.",
+ references: 14,
+ },
+ {
+ ...video,
+ id: "kling-3-pro",
+ name: "Kling 3.0",
+ icon: "large-kling",
+ description: "Cinematic motion with built-in audio.",
+ frames: ["start", "end"],
+ durations: seconds(3, 15),
+ aspectRatios: ["16:9", "9:16", "1:1"],
+ },
+ {
+ ...video,
+ id: "kling-o3-pro",
+ name: "Kling 3.0 Omni",
+ icon: "large-kling",
+ description: "Multi-modal reasoning, strong scenes.",
+ frames: ["start"],
+ durations: seconds(3, 15),
+ },
+ {
+ ...video,
+ id: "seedance-2.0",
+ name: "Seedance 2.0",
+ icon: "large-bytedance",
+ description: "Rich motion, lip-synced audio and SFX.",
+ frames: ["start", "end"],
+ durations: seconds(4, 15),
+ aspectRatios: ["16:9", "9:16", "1:1", "4:3", "3:4"],
+ },
+ {
+ ...video,
+ id: "veo-3.1",
+ name: "Veo 3.1",
+ icon: "large-google",
+ description: "Realistic physics, complex scenes.",
+ frames: ["start", "end"],
+ durations: [4, 6, 8],
+ aspectRatios: ["16:9", "9:16"],
+ },
+ {
+ ...video,
+ id: "veo-3.1-fast",
+ name: "Veo 3.1 Fast",
+ icon: "large-google",
+ description: "Same Veo engine, lower cost.",
+ frames: ["start", "end"],
+ durations: [4, 6, 8],
+ aspectRatios: ["16:9", "9:16"],
+ },
+ {
+ ...video,
+ id: "wan-2.6",
+ name: "Wan 2.6",
+ icon: "large-wan",
+ description: "Expressive motion, stylized aesthetics.",
+ frames: ["start"],
+ durations: [5, 10],
+ aspectRatios: ["16:9", "9:16"],
+ },
+ {
+ ...video,
+ id: "hailuo-3-max",
+ name: "Hailuo 3 Max",
+ icon: "large-hailuo",
+ description: "Smooth motion, cinematic lighting.",
+ frames: ["start", "end"],
+ durations: seconds(5, 15),
+ aspectRatios: ["16:9", "9:16", "1:1", "4:3", "3:4"],
+ },
+ {
+ id: "elevenlabs-v3",
+ mode: "VOICE",
+ name: "ElevenLabs v3",
+ icon: "large-elevenlabs",
+ description: "Expressive speech in many voices.",
+ voices: PROMPT_INPUT_VOICE_OPTIONS,
+ },
{
- name: "ElevenLabs Music",
id: "elevenlabs-music",
+ mode: "AUDIO",
+ name: "ElevenLabs Music",
icon: "large-elevenlabs",
description: "Full compositions, stems, and lyrics.",
},
{
- name: "ElevenLabs SFX",
id: "elevenlabs-sfx",
+ mode: "AUDIO",
+ name: "ElevenLabs SFX",
icon: "large-elevenlabs",
description: "Sound effects up to 30s, seamless looping.",
},
];
+
+/** The models the prompt box offers in `mode`, the first one by default. */
+export const modelOptions = (mode: PromptMode) => MODEL_OPTIONS.filter((option) => option.mode === mode);
+
+/** The prompt box's option for `id`; undefined for a model it doesn't make requests with (tools, text). */
+export const modelOption = (id: ModelId) => MODEL_OPTIONS.find((option) => option.id === id);
diff --git a/apps/web/src/components/genai/prompt-input-actions.tsx b/apps/web/src/components/genai/prompt-input-actions.tsx
index eb2ddd2a..b4fc4416 100644
--- a/apps/web/src/components/genai/prompt-input-actions.tsx
+++ b/apps/web/src/components/genai/prompt-input-actions.tsx
@@ -22,7 +22,7 @@ import { useGenerate } from "./use-generate";
import { useMediaSelection } from "./selection";
import { useTransforms } from "./use-transforms";
-import type { TransformType } from "./schemas";
+import type { TransformType } from "./types";
export function PromptInputActions() {
const { isGenerated, totalCredits, firstJob } = useGenerationRecords();
diff --git a/apps/web/src/components/genai/prompt-input.tsx b/apps/web/src/components/genai/prompt-input.tsx
index a299df60..f4a04849 100644
--- a/apps/web/src/components/genai/prompt-input.tsx
+++ b/apps/web/src/components/genai/prompt-input.tsx
@@ -31,31 +31,20 @@ import {
} from "solid-js";
import {
- PROMPT_INPUT_IMAGE_MODEL_OPTIONS,
- PROMPT_INPUT_VIDEO_MODEL_OPTIONS,
- PROMPT_INPUT_VOICE_MODEL,
- PROMPT_INPUT_VOICE_OPTIONS,
+ ASPECT_RATIO_OPTIONS,
PROMPT_INPUT_MODE_OPTIONS,
- PROMPT_INPUT_IMAGE_ASPECT_RATIO_OPTIONS,
- PROMPT_INPUT_VARIANT_COUNT_OPTIONS,
- ALL_VIDEO_ASPECT_RATIO_OPTIONS,
- ALL_DURATION_OPTIONS,
- PROMPT_INPUT_AUDIO_MODEL_OPTIONS,
+ modelOption,
+ modelOptions,
// PROMPT_INPUT_RESOLUTION_OPTIONS
+ type AspectRatio,
type PromptMode,
} from "./config";
+import { createDefaultConfig, fitToModel } from "./requests";
import { createStoredSignal } from "@/lib/store";
import { store } from "@/init";
import { useGenerate } from "./use-generate";
import { PromptInputActions } from "./prompt-input-actions";
-import type {
- AspectRatio,
- GenerationConfig,
- ImageGenerationConfig,
- VideoGenerationConfig,
- VoiceGenerationConfig,
- AudioGenerationConfig,
-} from "./schemas";
+import type { GenerationConfig } from "./types";
import { AssetThumbnail } from "@/components/ui/asset-thumbnail";
import { useLibrary } from "@/engine/library";
import { useEditor } from "@/engine/hooks";
@@ -63,57 +52,15 @@ import { droppedFiles, importFiles, pickFiles } from "@/engine/asset-actions";
import { useMediaSelection } from "./selection";
import { ASSET_DRAG_TYPE } from "@/components/sidebar-left/folder-item";
+import type { ModelId } from "@diffusionstudio/api-contract";
import type { AssetCache } from "@diffusionstudio/assets";
import type { ThumbnailAsset } from "@/components/ui/asset-thumbnail";
-export type PromptInputMode = PromptMode;
const DEFAULT_BUTTON_CLASS = "text-muted-foreground gap-0 pr-2 pl-0";
/** As many references as the prompt box has room for; a model may take fewer. */
const MAX_IMAGE_REFERENCES = 5;
-type GenerationConfgs = {
- IMAGE: ImageGenerationConfig;
- VIDEO: VideoGenerationConfig;
- VOICE: VoiceGenerationConfig;
- AUDIO: AudioGenerationConfig;
-};
-
-export function createDefaultConfig<K extends keyof GenerationConfgs>(mode: K, prompt = ""): GenerationConfgs[K] {
- switch (mode) {
- case "IMAGE":
- return {
- mode: "IMAGE",
- model: PROMPT_INPUT_IMAGE_MODEL_OPTIONS[0].id,
- prompt,
- aspectRatio: "16:9",
- count: 1,
- } as GenerationConfgs[K];
- case "VIDEO":
- return {
- mode: "VIDEO",
- model: PROMPT_INPUT_VIDEO_MODEL_OPTIONS[0].id,
- prompt,
- aspectRatio: "16:9",
- duration: 6,
- generateAudio: true,
- } as GenerationConfgs[K];
- case "VOICE":
- return {
- mode: "VOICE",
- model: PROMPT_INPUT_VOICE_MODEL,
- prompt,
- voice: PROMPT_INPUT_VOICE_OPTIONS[0].value,
- } as GenerationConfgs[K];
- case "AUDIO":
- return {
- mode: "AUDIO",
- model: PROMPT_INPUT_AUDIO_MODEL_OPTIONS[0].id,
- prompt,
- } as GenerationConfgs[K];
- }
-}
-
export interface PromptInputProps {
initialConfig?: GenerationConfig;
}
@@ -134,52 +81,33 @@ export function PromptInput(props: PromptInputProps) {
/** Shallow-merge updates into the current config. */
const patch = (updates: Partial<GenerationConfig>) => {
- setConfig((prev) => ({ ...prev, ...updates }) as GenerationConfig);
+ setConfig((prev) => ({ ...prev, ...updates }));
};
const mode = () => config().mode;
const prompt = () => config().prompt;
- // These are only read inside their matching <Match when={mode() === …}>
- // blocks, so the narrowing is sound.
- const imageConfig = () => {
- const c = config();
- return c.mode === "IMAGE" ? c : undefined;
- };
- const videoConfig = () => {
- const c = config();
- return c.mode === "VIDEO" ? c : undefined;
- };
- const voiceConfig = () => {
- const c = config();
- return c.mode === "VOICE" ? c : undefined;
- };
+ /** The model and the settings it takes: what the prompt box offers. */
+ const option = createMemo(() => modelOption(config().model) ?? modelOptions(mode())[0]!);
+ const options = createMemo(() => modelOptions(mode()));
+ const modeLabel = () => PROMPT_INPUT_MODE_OPTIONS.find((o) => o.value === mode())?.label.toLowerCase();
- const maxImageReferences = createMemo(() => Math.min(
- MAX_IMAGE_REFERENCES,
- PROMPT_INPUT_IMAGE_MODEL_OPTIONS.find((m) => m.id === config().model)?.maxReferences ?? MAX_IMAGE_REFERENCES,
- ));
+ const maxImageReferences = createMemo(() => Math.min(MAX_IMAGE_REFERENCES, option().references ?? 0));
- const currentVideoModel = createMemo(() =>
- PROMPT_INPUT_VIDEO_MODEL_OPTIONS.find((m) => m.id === config().model),
+ const aspectRatioOptions = createMemo(() =>
+ ASPECT_RATIO_OPTIONS.filter((o) => option().aspectRatios?.includes(o.value)),
+ );
+ const countOptions = createMemo(() => (option().counts ?? []).map((n) => ({ value: String(n), label: String(n) })));
+ const durationOptions = createMemo(() =>
+ (option().durations ?? []).map((seconds) => ({ value: String(seconds), label: `${seconds}s` })),
);
-
- const videoDurationOptions = createMemo(() => {
- const model = currentVideoModel();
- return model ? ALL_DURATION_OPTIONS.filter((o) => model.durations.includes(o.value)) : [];
- });
-
- const videoAspectRatioOptions = createMemo(() => {
- const model = currentVideoModel();
- return model ? ALL_VIDEO_ASPECT_RATIO_OPTIONS.filter((o) => model.aspectRatios.includes(o.value)) : [];
- });
// Pictures on the canvas are offered as references and as video frames:
// selecting one is another way of attaching it.
const effectiveImageRefIds = createMemo(() => {
const lib = library();
if (!lib) return [];
- const local = imageConfig()?.imageRefIds ?? [];
+ const local = config().imageRefIds ?? [];
const selected = selectedImages()
.map((entry) => entry.asset.id)
.filter((id) => !local.includes(id));
@@ -193,11 +121,10 @@ export function PromptInput(props: PromptInputProps) {
// A frame the model cannot take is not one the prompt box offers, and
// not one a canvas selection quietly attaches either: the declaration is
// checked against the model before it runs.
- if (!currentVideoModel()?.features.includes(`${frame}-frame`)) return null;
+ if (!option().frames?.includes(frame)) return null;
const lib = library();
- const config = videoConfig();
- const explicit = frame === "start" ? config?.startFrameImageId : config?.endFrameImageId;
+ const explicit = frame === "start" ? config().startFrameImageId : config().endFrameImageId;
const id = explicit ?? selectedImages()[frame === "start" ? 0 : 1]?.asset.id ?? null;
return id && lib?.get(id) ? id : null;
};
@@ -226,13 +153,13 @@ export function PromptInput(props: PromptInputProps) {
};
const handleModeChange = (value: string) => {
- const newMode = value as PromptInputMode;
+ const newMode = value as PromptMode;
if (newMode === mode()) return;
setConfig(createDefaultConfig(newMode, prompt()));
};
const removeImageReference = (assetId: string) => {
- const refs = imageConfig()?.imageRefIds ?? [];
+ const refs = config().imageRefIds ?? [];
if (refs.includes(assetId)) {
patch({ imageRefIds: refs.filter((id) => id !== assetId) });
return;
@@ -254,7 +181,7 @@ export function PromptInput(props: PromptInputProps) {
// `importFiles` reports its own failures.
const imported = await importFiles(lib, files, "");
const imageAssets = imported.filter((a) => a.type === "IMAGE");
- const current = imageConfig()?.imageRefIds ?? [];
+ const current = config().imageRefIds ?? [];
patch({
imageRefIds: [...current, ...imageAssets.map((a) => a.id)].slice(0, maxImageReferences()),
});
@@ -275,16 +202,12 @@ export function PromptInput(props: PromptInputProps) {
};
const swapVideoFrameImages = () => {
- const c = videoConfig();
- patch({
- startFrameImageId: c?.endFrameImageId,
- endFrameImageId: c?.startFrameImageId,
- });
+ const c = config();
+ patch({ startFrameImageId: c.endFrameImageId, endFrameImageId: c.startFrameImageId });
};
const clearVideoFrame = (frame: "start" | "end") => {
- const explicitId =
- frame === "start" ? videoConfig()?.startFrameImageId : videoConfig()?.endFrameImageId;
+ const explicitId = frame === "start" ? config().startFrameImageId : config().endFrameImageId;
if (explicitId) {
patch(frame === "start" ? { startFrameImageId: undefined } : { endFrameImageId: undefined });
return;
@@ -327,24 +250,14 @@ export function PromptInput(props: PromptInputProps) {
patch({ prompt: "" });
resizeTextarea();
- let submitted: GenerationConfig;
-
- if (currentConfig.mode === "IMAGE") {
- submitted = {
- ...currentConfig,
- imageRefIds: effectiveImageRefIds(),
- };
- } else if (currentConfig.mode === "VIDEO") {
- submitted = {
- ...currentConfig,
- startFrameImageId: effectiveStartFrameId() || undefined,
- endFrameImageId: effectiveEndFrameId() || undefined,
- };
- } else {
- submitted = currentConfig;
- }
-
- generate(submitted);
+ // What the model takes of the selection: none of it for a model without
+ // references or frames (see `maxImageReferences`, `frameId`).
+ generate({
+ ...currentConfig,
+ imageRefIds: effectiveImageRefIds(),
+ startFrameImageId: effectiveStartFrameId() ?? undefined,
+ endFrameImageId: effectiveEndFrameId() ?? undefined,
+ });
};
const handleKeyDown = (e: KeyboardEvent) => {
@@ -399,7 +312,7 @@ export function PromptInput(props: PromptInputProps) {
dragCounter = 0;
setIsDragging(false);
- if (mode() !== "IMAGE" && mode() !== "VIDEO") return;
+ if (maxImageReferences() === 0 && !option().frames) return;
const lib = library();
if (!lib) return;
@@ -420,48 +333,32 @@ export function PromptInput(props: PromptInputProps) {
const imageIds = droppedIds.filter((id) => lib.get(id)?.type === "IMAGE");
if (imageIds.length === 0) return;
- if (mode() === "VIDEO") {
- const c = videoConfig();
- if (!c?.startFrameImageId && imageIds[0]) {
+ if (option().frames) {
+ const c = config();
+ if (!c.startFrameImageId && imageIds[0]) {
patch({ startFrameImageId: imageIds[0] });
}
- if (!c?.endFrameImageId && imageIds[1]) {
+ if (!c.endFrameImageId && imageIds[1] && option().frames?.includes("end")) {
patch({ endFrameImageId: imageIds[1] });
}
return;
}
- const current = imageConfig()?.imageRefIds ?? [];
+ const current = config().imageRefIds ?? [];
patch({ imageRefIds: [...current, ...imageIds].slice(0, maxImageReferences()) });
};
- const handleVideoModelChange = (id: string) => {
- const model = PROMPT_INPUT_VIDEO_MODEL_OPTIONS.find((m) => m.id === id);
- if (!model) return;
-
- const c = videoConfig() ?? createDefaultConfig("VIDEO");
- const durationStr = `${c.duration}s`;
-
- setConfig({
- ...c,
- model: id,
- endFrameImageId: model.features.includes("end-frame") ? c.endFrameImageId : undefined,
- generateAudio: model.features.includes("audio") ? (c.generateAudio ?? true) : false,
- duration: model.durations.includes(durationStr) ? c.duration : parseInt(model.durations[0], 10),
- aspectRatio: model.aspectRatios.length === 0 || model.aspectRatios.includes(c.aspectRatio)
- ? c.aspectRatio
- : (model.aspectRatios[0] as AspectRatio),
- });
+ /** Another model of the mode, the settings carried over where it takes them. */
+ const handleModelChange = (id: string) => {
+ setConfig(fitToModel({ ...config(), model: id as ModelId }));
};
// ── Accessors for UI menus (string ↔ config conversions) ────────────
- const aspectRatioAccessor: Accessor<string> = () =>
- imageConfig()?.aspectRatio ?? videoConfig()?.aspectRatio ?? "16:9";
- const variantCountAccessor: Accessor<string> = () => String(imageConfig()?.count ?? 1);
- const durationAccessor: Accessor<string> = () => `${videoConfig()?.duration ?? 6}s`;
+ const aspectRatioAccessor: Accessor<string> = () => config().aspectRatio ?? "";
+ const countAccessor: Accessor<string> = () => String(config().count ?? "");
+ const durationAccessor: Accessor<string> = () => String(config().duration ?? "");
const modelAccessor: Accessor<string> = () => config().model;
- const voiceAccessor: Accessor<string> = () => voiceConfig()?.voice ?? PROMPT_INPUT_VOICE_OPTIONS[0].value;
- const audioEnabledAccessor = () => videoConfig()?.generateAudio ?? true;
+ const voiceAccessor: Accessor<string> = () => config().voice ?? "";
return (
<div
@@ -473,7 +370,7 @@ export function PromptInput(props: PromptInputProps) {
>
<PromptInputActions />
<Switch>
- <Match when={mode() === "IMAGE"}>
+ <Match when={maxImageReferences() > 0}>
<div class="flex w-full items-start gap-2 overflow-x-auto">
<For each={effectiveImageRefIds()}>
{(assetId) => (
@@ -497,9 +394,9 @@ export function PromptInput(props: PromptInputProps) {
</div>
</Match>
- <Match when={mode() === "VIDEO"}>
+ <Match when={option().frames}>
<div class="flex w-full items-center gap-2 overflow-x-auto">
- <Show when={currentVideoModel()?.features.includes("start-frame")}>
+ <Show when={option().frames?.includes("start")}>
<Show
when={effectiveStartFrameId()}
fallback={
@@ -522,7 +419,7 @@ export function PromptInput(props: PromptInputProps) {
)}
</Show>
</Show>
- <Show when={currentVideoModel()?.features.includes("end-frame")}>
+ <Show when={option().frames?.includes("end")}>
<Tooltip>
<TooltipTrigger
as={Button}
@@ -626,98 +523,55 @@ export function PromptInput(props: PromptInputProps) {
</Tooltip>
<Show when={settingsVisible()}>
<Separator orientation="vertical" class="min-h-5" />
- <Switch>
- <Match when={mode() === "IMAGE"}>
- <>
- <ModelMenu
- searchPlaceholder="Search in image models"
- options={PROMPT_INPUT_IMAGE_MODEL_OPTIONS}
- value={modelAccessor}
- onChange={(v) => patch({ model: v })}
- />
- <PromptInputCompactMenu
- aria-label="Select aspect ratio"
- menuLabel="Aspect ratio"
- value={aspectRatioAccessor}
- options={PROMPT_INPUT_IMAGE_ASPECT_RATIO_OPTIONS}
- onChange={(v) => patch({ aspectRatio: v as AspectRatio })}
- triggerIcon="aspect-ratio-16-9"
- />
- <PromptInputCompactMenu
- aria-label="Select amount of variants"
- menuLabel="Amount of variants"
- value={variantCountAccessor}
- options={PROMPT_INPUT_VARIANT_COUNT_OPTIONS}
- onChange={(v) => patch({ count: Number(v) })}
- triggerIcon="variants"
- />
- </>
- </Match>
- <Match when={mode() === "VIDEO"}>
- <>
- <ModelMenu
- searchPlaceholder="Search in video models"
- options={PROMPT_INPUT_VIDEO_MODEL_OPTIONS}
- value={modelAccessor}
- onChange={handleVideoModelChange}
- />
- <Show when={videoAspectRatioOptions().length > 0}>
- <PromptInputCompactMenu
- aria-label="Select aspect ratio"
- menuLabel="Aspect ratio"
- value={aspectRatioAccessor}
- options={videoAspectRatioOptions()}
- onChange={(v) => patch({ aspectRatio: v as AspectRatio })}
- triggerIcon="aspect-ratio-16-9"
- />
- </Show>
- {/* <PromptInputCompactMenu
- aria-label="Select resolution"
- menuLabel="Resolution"
- value={resolution}
- options={PROMPT_INPUT_RESOLUTION_OPTIONS}
- onChange={setResolution}
- triggerIcon="resolution"
- /> */}
- <PromptInputCompactMenu
- aria-label="Select duration"
- menuLabel="Duration"
- value={durationAccessor}
- options={videoDurationOptions()}
- onChange={(v) => patch({ duration: parseInt(v) })}
- triggerIcon="duration"
- />
- <Show when={currentVideoModel()?.features.includes("audio")}>
- <Button
- variant="ghost"
- onClick={() => patch({ generateAudio: !audioEnabledAccessor() })}
- class={DEFAULT_BUTTON_CLASS}
- >
- <Icon
- name={audioEnabledAccessor() ? "audio-on" : "audio-off"}
- class="size-6"
- />
- {audioEnabledAccessor() ? "On" : "Off"}
- </Button>
- </Show>
- </>
- </Match>
- <Match when={mode() === "VOICE"}>
- <VoiceMenu
- options={PROMPT_INPUT_VOICE_OPTIONS}
- value={voiceAccessor}
- onChange={(v) => patch({ voice: v })}
- />
- </Match>
- <Match when={mode() === "AUDIO"}>
- <ModelMenu
- searchPlaceholder="Search in audio models"
- options={PROMPT_INPUT_AUDIO_MODEL_OPTIONS}
- value={modelAccessor}
- onChange={(v) => patch({ model: v })}
- />
- </Match>
- </Switch>
+ <Show when={options().length > 1}>
+ <ModelMenu
+ searchPlaceholder={`Search in ${modeLabel()} models`}
+ options={options()}
+ value={modelAccessor}
+ onChange={handleModelChange}
+ />
+ </Show>
+ <Show when={aspectRatioOptions().length > 0}>
+ <PromptInputCompactMenu
+ aria-label="Select aspect ratio"
+ menuLabel="Aspect ratio"
+ value={aspectRatioAccessor}
+ options={aspectRatioOptions()}
+ onChange={(v) => patch({ aspectRatio: v as AspectRatio })}
+ triggerIcon="aspect-ratio-16-9"
+ />
+ </Show>
+ {/* <PromptInputCompactMenu
+ aria-label="Select resolution"
+ menuLabel="Resolution"
+ value={resolution}
+ options={PROMPT_INPUT_RESOLUTION_OPTIONS}
+ onChange={setResolution}
+ triggerIcon="resolution"
+ /> */}
+ <Show when={countOptions().length > 0}>
+ <PromptInputCompactMenu
+ aria-label="Select amount of variants"
+ menuLabel="Amount of variants"
+ value={countAccessor}
+ options={countOptions()}
+ onChange={(v) => patch({ count: Number(v) })}
+ triggerIcon="variants"
+ />
+ </Show>
+ <Show when={durationOptions().length > 0}>
+ <PromptInputCompactMenu
+ aria-label="Select duration"
+ menuLabel="Duration"
+ value={durationAccessor}
+ options={durationOptions()}
+ onChange={(v) => patch({ duration: Number(v) })}
+ triggerIcon="duration"
+ />
+ </Show>
+ <Show when={option().voices}>
+ {(voices) => <VoiceMenu options={voices()} value={voiceAccessor} onChange={(v) => patch({ voice: v })} />}
+ </Show>
</Show>
</div>
<Tooltip>
diff --git a/apps/web/src/components/genai/requests.ts b/apps/web/src/components/genai/requests.ts
index 57f583e1..9fbf2c19 100644
--- a/apps/web/src/components/genai/requests.ts
+++ b/apps/web/src/components/genai/requests.ts
@@ -10,13 +10,13 @@
import { uploadAsset } from "@/engine/generate";
import { AUDIO_SIZE } from "@/engine/insert-asset";
-import { ASPECT_RATIO_DIMENSIONS, MODEL_MODES, PROMPT_INPUT_VIDEO_MODEL_OPTIONS, PROMPT_INPUT_VOICE_OPTIONS } from "./config";
-import { aspectRatioSchema } from "./schemas";
+import { ASPECT_RATIO_DIMENSIONS, MODEL_MODES, modelOption, modelOptions } from "./config";
import type { AssetLibrary } from "@diffusionstudio/assets";
import type { AssetRef, GenerateRequest, GenerateRequestInput, ModelId } from "@diffusionstudio/api-contract";
import type { GenerationOutput } from "@/engine/generate";
-import type { GenerationConfig } from "./schemas";
+import type { AspectRatio, PromptMode } from "./config";
+import type { GenerationConfig } from "./types";
const TITLES = {
IMAGE: "Image generation failed",
@@ -55,42 +55,53 @@ function refs(library: AssetLibrary, ids: (string | undefined)[]): Promise<(Asse
}));
}
+/** `value` when `allowed` has it, else `fallback` when it has that, else its first; undefined without `allowed`. */
+function pick<T>(value: T | undefined, allowed: readonly T[] | undefined, fallback?: T): T | undefined {
+ if (!allowed?.length) return undefined;
+ if (value !== undefined && allowed.includes(value)) return value;
+ if (fallback !== undefined && allowed.includes(fallback)) return fallback;
+ return allowed[0];
+}
+
/**
- * The request `config` makes, its library inputs uploaded. The API validates
- * the rest: a model that does not take a setting says so.
+ * `config` within what its model takes (see `ModelOption`): a setting the
+ * model allows is kept, one it allows other values of falls back to a
+ * default, and one it doesn't take is dropped. Every config the prompt box
+ * holds has been through here, so its requests are within the model's bounds.
*/
-export async function toRequest(library: AssetLibrary, config: GenerationConfig): Promise<GenerateRequestInput> {
- const model = config.model as ModelId;
+export function fitToModel(config: GenerationConfig): GenerationConfig {
+ const option = modelOption(config.model) ?? modelOptions(config.mode)[0]!;
+ const frame = (frame: "start" | "end", id: string | undefined) => (option.frames?.includes(frame) ? id : undefined);
+ return {
+ mode: option.mode,
+ model: option.id,
+ prompt: config.prompt,
+ aspectRatio: pick(config.aspectRatio, option.aspectRatios, "16:9"),
+ count: pick(config.count, option.counts, 1),
+ duration: pick(config.duration, option.durations, 6),
+ voice: pick(config.voice, option.voices?.map((voice) => voice.value)),
+ imageRefIds: option.references ? config.imageRefIds?.slice(0, option.references) : undefined,
+ startFrameImageId: frame("start", config.startFrameImageId),
+ endFrameImageId: frame("end", config.endFrameImageId),
+ };
+}
- switch (config.mode) {
- case "IMAGE": {
- const images = (await refs(library, config.imageRefIds ?? [])).filter((ref) => ref !== undefined);
- return {
- model,
- prompt: config.prompt,
- aspectRatio: config.aspectRatio,
- count: config.count,
- ...(images.length ? { images } : {}),
- } as GenerateRequestInput;
- }
- case "VIDEO": {
- const option = PROMPT_INPUT_VIDEO_MODEL_OPTIONS.find((candidate) => candidate.id === model);
- const [startFrame, endFrame] = await refs(library, [config.startFrameImageId, config.endFrameImageId]);
- return {
- model,
- prompt: config.prompt,
- duration: config.duration,
- ...(option?.aspectRatios.length ? { aspectRatio: config.aspectRatio } : {}),
- ...(option?.features.includes("audio") ? { generateAudio: config.generateAudio ?? false } : {}),
- ...(startFrame ? { startFrame } : {}),
- ...(endFrame ? { endFrame } : {}),
- } as GenerateRequestInput;
- }
- case "VOICE":
- return { model: "elevenlabs-v3", prompt: config.prompt, voice: config.voice };
- case "AUDIO":
- return { model, prompt: config.prompt } as GenerateRequestInput;
- }
+/** The prompt box in `mode`, set up for the mode's first model. */
+export function createDefaultConfig(mode: PromptMode, prompt = ""): GenerationConfig {
+ return fitToModel({ mode, model: modelOptions(mode)[0]!.id, prompt });
+}
+
+/** The request `config` makes, its library inputs uploaded. Settings the config leaves unset are left out. */
+export async function toRequest(library: AssetLibrary, config: GenerationConfig): Promise<GenerateRequestInput> {
+ const { model, prompt, aspectRatio, count, duration, voice } = config;
+ const [startFrame, endFrame, ...images] = await refs(library, [
+ config.startFrameImageId,
+ config.endFrameImageId,
+ ...(config.imageRefIds ?? []),
+ ]);
+ const uploaded = images.filter((ref) => ref !== undefined);
+ const request = { model, prompt, aspectRatio, count, duration, voice, startFrame, endFrame, images: uploaded.length ? uploaded : undefined };
+ return Object.fromEntries(Object.entries(request).filter(([, value]) => value !== undefined)) as GenerateRequestInput;
}
/**
@@ -99,49 +110,17 @@ export async function toRequest(library: AssetLibrary, config: GenerationConfig)
* asset. Undefined for a request the prompt box does not make (a tool).
*/
export function toConfig(request: GenerateRequest): GenerationConfig | undefined {
- const aspectRatio = (value: unknown) => aspectRatioSchema.safeParse(value).data ?? "16:9";
-
- switch (request.model) {
- case "gpt-image-2":
- case "nano-banana-2":
- case "nano-banana-pro":
- case "seedream-4.5":
- case "flux-2-klein":
- return {
- mode: "IMAGE",
- model: request.model,
- prompt: request.prompt,
- aspectRatio: aspectRatio(request.aspectRatio),
- count: Math.min(request.count, 4),
- };
- case "kling-3-pro":
- case "kling-o3-pro":
- case "wan-2.6":
- case "hailuo-3-max":
- case "seedance-2.0":
- case "veo-3.1":
- case "veo-3.1-fast":
- return {
- mode: "VIDEO",
- model: request.model,
- prompt: request.prompt,
- aspectRatio: aspectRatio("aspectRatio" in request ? request.aspectRatio : undefined),
- duration: request.duration,
- generateAudio: "generateAudio" in request ? request.generateAudio : undefined,
- };
- case "elevenlabs-v3":
- return {
- mode: "VOICE",
- model: request.model,
- prompt: request.prompt,
- voice: PROMPT_INPUT_VOICE_OPTIONS.some((option) => option.value === request.voice)
- ? request.voice
- : PROMPT_INPUT_VOICE_OPTIONS[0]!.value,
- };
- case "elevenlabs-music":
- case "elevenlabs-sfx":
- return { mode: "AUDIO", model: request.model, prompt: request.prompt };
- default:
- return undefined;
- }
+ const option = modelOption(request.model);
+ if (!option || !("prompt" in request) || request.prompt === undefined) return undefined;
+ // The settings a request of the prompt box's models may carry; `fitToModel` checks them.
+ const settings = request as { aspectRatio?: string; count?: number; duration?: number; voice?: string };
+ return fitToModel({
+ mode: option.mode,
+ model: option.id,
+ prompt: request.prompt,
+ aspectRatio: settings.aspectRatio as AspectRatio | undefined,
+ count: settings.count,
+ duration: settings.duration,
+ voice: settings.voice,
+ });
}
diff --git a/apps/web/src/components/genai/schemas.ts b/apps/web/src/components/genai/schemas.ts
deleted file mode 100644
index 4705baa2..00000000
--- a/apps/web/src/components/genai/schemas.ts
+++ /dev/null
@@ -1,59 +0,0 @@
-/* This Source Code Form is subject to the terms of the Mozilla Public
- * License, v. 2.0. If a copy of the MPL was not distributed with this
- * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
-
-import { z } from "zod";
-
-/** What a generation request says, in the vocabulary the prompt input speaks. */
-
-export const aspectRatioSchema = z.enum(["1:1", "4:3", "3:4", "16:9", "9:16"]);
-export type AspectRatio = z.infer<typeof aspectRatioSchema>;
-
-export const imageGenerationConfigSchema = z.object({
- mode: z.literal("IMAGE"),
- model: z.string(),
- prompt: z.string(),
- aspectRatio: aspectRatioSchema,
- count: z.number().int().min(1).max(4),
- imageRefIds: z.array(z.string()).optional(),
-});
-
-export const videoGenerationConfigSchema = z.object({
- mode: z.literal("VIDEO"),
- model: z.string(),
- prompt: z.string(),
- aspectRatio: aspectRatioSchema,
- duration: z.number().int().min(1),
- generateAudio: z.boolean().optional(),
- startFrameImageId: z.string().optional(),
- endFrameImageId: z.string().optional(),
-});
-
-export const voiceGenerationConfigSchema = z.object({
- mode: z.literal("VOICE"),
- model: z.string(),
- prompt: z.string(),
- voice: z.string(),
-});
-
-export const audioGenerationConfigSchema = z.object({
- mode: z.literal("AUDIO"),
- model: z.string(),
- prompt: z.string(),
-});
-
-export const generationConfigSchema = z.discriminatedUnion("mode", [
- imageGenerationConfigSchema,
- videoGenerationConfigSchema,
- voiceGenerationConfigSchema,
- audioGenerationConfigSchema,
-]);
-
-export type ImageGenerationConfig = z.infer<typeof imageGenerationConfigSchema>;
-export type VideoGenerationConfig = z.infer<typeof videoGenerationConfigSchema>;
-export type VoiceGenerationConfig = z.infer<typeof voiceGenerationConfigSchema>;
-export type AudioGenerationConfig = z.infer<typeof audioGenerationConfigSchema>;
-export type GenerationConfig = z.infer<typeof generationConfigSchema>;
-
-/** What can be done to an asset the selection shows. */
-export type TransformType = "upscale" | "removeBackground" | "addAudio";
diff --git a/apps/web/src/components/genai/types.ts b/apps/web/src/components/genai/types.ts
new file mode 100644
index 00000000..df7c05ef
--- /dev/null
+++ b/apps/web/src/components/genai/types.ts
@@ -0,0 +1,28 @@
+/* This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+import type { ModelId } from "@diffusionstudio/api-contract";
+import type { AspectRatio, PromptMode } from "./config";
+
+/**
+ * The prompt box's settings: the fields of the request it makes, with inputs
+ * named by library asset rather than upload. Which settings are set follows
+ * from the model (see `ModelOption` and `fitToModel`), not from the mode.
+ */
+export interface GenerationConfig {
+ mode: PromptMode;
+ model: ModelId;
+ prompt: string;
+ aspectRatio?: AspectRatio;
+ count?: number;
+ /** Seconds. */
+ duration?: number;
+ voice?: string;
+ imageRefIds?: string[];
+ startFrameImageId?: string;
+ endFrameImageId?: string;
+}
+
+/** What can be done to an asset the selection shows. */
+export type TransformType = "upscale" | "removeBackground" | "addAudio";
diff --git a/apps/web/src/components/genai/use-generate.ts b/apps/web/src/components/genai/use-generate.ts
index 03f4f69f..58fe8c90 100644
--- a/apps/web/src/components/genai/use-generate.ts
+++ b/apps/web/src/components/genai/use-generate.ts
@@ -8,7 +8,7 @@ import { generate } from "@/engine/generate";
import { outputOf, toRequest } from "./requests";
import type { GenerateRequest } from "@diffusionstudio/api-contract";
-import type { GenerationConfig } from "./schemas";
+import type { GenerationConfig } from "./types";
/**
* Generating onto the canvas: from the prompt box's settings (`generate`), or
@@ -20,11 +20,7 @@ export function useGenerate() {
const run = (config: GenerationConfig): Promise<void> => {
const library = world.get(Library);
- const output = outputOf({
- model: config.model as GenerateRequest["model"],
- count: config.mode === "IMAGE" ? config.count : 1,
- aspectRatio: "aspectRatio" in config ? config.aspectRatio : undefined,
- });
+ const output = outputOf(config);
if (!library || !output) return Promise.resolve();
return generate(world, output, () => toRequest(library, config));
};
diff --git a/apps/web/src/components/genai/use-generation-records.ts b/apps/web/src/components/genai/use-generation-records.ts
index eb50f04d..4f961b30 100644
--- a/apps/web/src/components/genai/use-generation-records.ts
+++ b/apps/web/src/components/genai/use-generation-records.ts
@@ -13,7 +13,7 @@ import { useMediaSelection } from "./selection";
import { toConfig } from "./requests";
import type { Job } from "@diffusionstudio/api-contract";
-import type { GenerationConfig } from "./schemas";
+import type { GenerationConfig } from "./types";
export function useGenerationRecords() {
const { media } = useMediaSelection();
diff --git a/apps/web/src/components/genai/use-transforms.ts b/apps/web/src/components/genai/use-transforms.ts
index 9dca04a5..5544d2a7 100644
--- a/apps/web/src/components/genai/use-transforms.ts
+++ b/apps/web/src/components/genai/use-transforms.ts
@@ -13,7 +13,7 @@ import { useMediaSelection } from "./selection";
import type { AssetRef, GenerateRequestInput } from "@diffusionstudio/api-contract";
import type { NodeMedia } from "@/engine/generate";
-import type { TransformType } from "./schemas";
+import type { TransformType } from "./types";
interface Tool {
/** What a failure is reported under. */
diff --git a/apps/web/src/components/sidebar-left/assets.tsx b/apps/web/src/components/sidebar-left/assets.tsx
index 1768eb33..4bbcda89 100644
--- a/apps/web/src/components/sidebar-left/assets.tsx
+++ b/apps/web/src/components/sidebar-left/assets.tsx
@@ -4,7 +4,7 @@
import { basename, dirname } from "@diffusionstudio/assets";
import { usePromptInput } from "@/context/prompt-input";
-import { createDefaultConfig } from "@/components/genai/prompt-input";
+import { createDefaultConfig } from "@/components/genai/requests";
import { createEffect, createMemo, createSignal, For, onCleanup, onMount, Show } from "solid-js";
import { SidebarTabs } from "@/agent-chat";
import { Button } from "../ui/button";
diff --git a/apps/web/src/components/sidebar-right/inspector/asset-picker.tsx b/apps/web/src/components/sidebar-right/inspector/asset-picker.tsx
index 01e75e8a..3c7e9b17 100644
--- a/apps/web/src/components/sidebar-right/inspector/asset-picker.tsx
+++ b/apps/web/src/components/sidebar-right/inspector/asset-picker.tsx
@@ -18,7 +18,7 @@ import { Icon } from "@/components/ui/icon";
import { Separator } from "@/components/ui/separator";
import { LazyAssetItem } from "@/components/sidebar-left/asset-item";
import { usePromptInput } from "@/context/prompt-input";
-import { createDefaultConfig } from "@/components/genai/prompt-input";
+import { createDefaultConfig } from "@/components/genai/requests";
import { assetName } from "@diffusionstudio/assets";
import { useTrait } from "@diffusionstudio/koota-solid";
import { AssetId, isScene } from "@diffusionstudio/runtime";
diff --git a/apps/web/src/context/prompt-input.tsx b/apps/web/src/context/prompt-input.tsx
index 7777b575..7b36826f 100644
--- a/apps/web/src/context/prompt-input.tsx
+++ b/apps/web/src/context/prompt-input.tsx
@@ -4,7 +4,7 @@
import { createContext, createSignal, useContext } from "solid-js";
import type { JSX } from "solid-js";
-import type { GenerationConfig } from "@/components/genai/schemas";
+import type { GenerationConfig } from "@/components/genai/types";
type PromptInputContextValue = {
openPromptInput: (config: GenerationConfig) => void;