src/lib/components/MonsterGenerator/MonsterGenerator.svelte

<script lang="ts">
  import type { MonsterGeneratorProps, MonsterWorkflowState, CaptionType, CaptionLength } from '$lib/types';
  import UploadStep from './UploadStep.svelte';
  import WorkflowProgress from './WorkflowProgress.svelte';
  import MonsterResult from './MonsterResult.svelte';
  
  interface Props extends MonsterGeneratorProps {}
  
  let { joyCaptionClient, rwkvClient, fluxClient }: Props = $props();
  
  let state: MonsterWorkflowState = $state({
    currentStep: 'upload',
    userImage: null,
    imageCaption: null,
    monsterConcept: null,
    imagePrompt: null,
    monsterImage: null,
    error: null,
    isProcessing: false
  });
  
  // Prompt templates
  const MONSTER_CONCEPT_PROMPT = (caption: string) => `User: Based on this image caption: "${caption}"

Come up with a monster idea (in the vein of Pokémon) that incorporates properties of what the pictured object is.
Assess how unique the pictured object is. If the image is not very unique then it should result in weak & common monster.
Meanwhile if the image is very unique the monster should be unique and powerful.

Create a creature that reflects the essence, materials, function, and characteristics of the object. Consider:
- What the object is made of (metal, wood, fabric, etc.)
- How it functions or is used
- Its physical properties (hard, soft, hot, cold, etc.)
- Its cultural or symbolic meaning
- Its shape, texture, and visual characteristics

The monster should have unique strengths and weaknesses based on these properties.

Include:
- Physical appearance and distinguishing features
- Special abilities or powers  
- Personality traits
- Habitat or origin

Assistant: **Monster Name:**`;

  const IMAGE_GENERATION_PROMPT = (concept: string) => `User: Convert this monster concept into a clear and succinct description of its appearance:
"${concept}"

Include all of its visual details, format the description as a single long sentence.

Assistant:`;
  
  async function handleImageSelected(file: File) {
    if (!joyCaptionClient || !rwkvClient || !fluxClient) {
      state.error = "Services not connected. Please wait...";
      return;
    }
    
    state.userImage = file;
    state.error = null;
    startWorkflow();
  }
  
  async function startWorkflow() {
    state.isProcessing = true;
    
    try {
      // Step 1: Caption the image
      await captionImage();
      await new Promise(resolve => setTimeout(resolve, 100)); // Small delay for state update
      
      // Step 2: Generate monster concept
      await generateMonsterConcept();
      await new Promise(resolve => setTimeout(resolve, 100)); // Small delay for state update
      
      // Step 3: Generate image prompt
      await generateImagePrompt();
      await new Promise(resolve => setTimeout(resolve, 100)); // Small delay for state update
      
      // Step 4: Generate monster image
      await generateMonsterImage();
      
      state.currentStep = 'complete';
    } catch (err) {
      console.error('Workflow error:', err);
      state.error = err instanceof Error ? err.message : 'An unknown error occurred';
    } finally {
      state.isProcessing = false;
    }
  }
  
  async function captionImage() {
    state.currentStep = 'captioning';
    
    if (!joyCaptionClient || !state.userImage) {
      throw new Error('Caption service not available or no image provided');
    }
    
    const output = await joyCaptionClient.predict("/stream_chat", [
      state.userImage,
      "Descriptive" as CaptionType,
      "very long" as CaptionLength,
      [], // extra_options
      "", // name_input
      ""  // custom_prompt
    ]);
    
    const [prompt, caption] = output.data;
    state.imageCaption = caption;
    console.log('Caption generated:', caption);
  }
  
  async function generateMonsterConcept() {
    state.currentStep = 'conceptualizing';
    
    if (!rwkvClient || !state.imageCaption) {
      throw new Error('Text generation service not available or no caption');
    }
    
    const conceptPrompt = MONSTER_CONCEPT_PROMPT(state.imageCaption);
    
    console.log('Generating monster concept with prompt:', conceptPrompt);
    
    const output = await rwkvClient.predict(0, [
      conceptPrompt,
      300,   // maxTokens
      0.7,   // temperature
      0.8,   // topP
      0.1,   // presencePenalty
      0.1    // countPenalty
    ]);
    
    console.log('RWKV output:', output);
    state.monsterConcept = output.data[0];
    console.log('Monster concept generated:', state.monsterConcept);
    
    if (!state.monsterConcept || state.monsterConcept.trim() === '') {
      throw new Error('Failed to generate monster concept - received empty response');
    }
  }
  
  async function generateImagePrompt() {
    state.currentStep = 'promptCrafting';
    
    if (!rwkvClient || !state.monsterConcept) {
      throw new Error('Text generation service not available or no concept');
    }
    
    const promptGenerationPrompt = IMAGE_GENERATION_PROMPT(state.monsterConcept);
    
    console.log('Generating image prompt from concept');
    
    const output = await rwkvClient.predict(0, [
      promptGenerationPrompt,
      300,   // maxTokens
      0.7,   // temperature
      0.8,   // topP
      0.1,   // presencePenalty
      0.1    // countPenalty
    ]);
    
    console.log('Image prompt output:', output);
    state.imagePrompt = output.data[0];
    console.log('Image prompt generated:', state.imagePrompt);
    
    if (!state.imagePrompt || state.imagePrompt.trim() === '') {
      throw new Error('Failed to generate image prompt - received empty response');
    }
  }
  
  async function generateMonsterImage() {
    state.currentStep = 'generating';
    
    if (!fluxClient || !state.imagePrompt) {
      throw new Error('Image generation service not available or no prompt');
    }
    
    const output = await fluxClient.predict("/infer", [
      `${state.imagePrompt}\nNow generate an Anime-style image of the monster in an idle pose. The monster should not be attacking or in motion. The full monster must be visible within the frame.`,
      0,      // seed
      true,   // randomizeSeed
      1024,   // width
      1024,   // height
      4       // steps
    ]);
    
    const [image, usedSeed] = output.data;
    let url: string | undefined;
    
    if (typeof image === "string") url = image;
    else if (image && image.url) url = image.url;
    else if (image && image.path) url = image.path;
    
    if (url) {
      state.monsterImage = {
        imageUrl: url,
        seed: usedSeed,
        prompt: state.imagePrompt
      };
    } else {
      throw new Error('Failed to generate monster image');
    }
  }
  
  function reset() {
    state = {
      currentStep: 'upload',
      userImage: null,
      imageCaption: null,
      monsterConcept: null,
      imagePrompt: null,
      monsterImage: null,
      error: null,
      isProcessing: false
    };
  }
</script>

<div class="monster-generator">
  
  {#if state.currentStep !== 'upload'}
    <WorkflowProgress currentStep={state.currentStep} error={state.error} />
  {/if}
  
  {#if state.currentStep === 'upload'}
    <UploadStep 
      onImageSelected={handleImageSelected}
      isProcessing={state.isProcessing}
    />
  {:else if state.currentStep === 'complete'}
    <MonsterResult state={state} onReset={reset} />
  {:else}
    <div class="processing-container">
      <div class="spinner"></div>
      <p class="processing-text">
        {#if state.currentStep === 'captioning'}
          Analyzing your image...
        {:else if state.currentStep === 'conceptualizing'}
          Creating monster concept...
        {:else if state.currentStep === 'promptCrafting'}
          Crafting generation prompt...
        {:else if state.currentStep === 'generating'}
          Generating your monster...
        {/if}
      </p>
      
      {#if state.imageCaption && state.currentStep !== 'captioning'}
        <div class="intermediate-result">
          <h4>Image Description:</h4>
          <p>{state.imageCaption}</p>
        </div>
      {/if}
      
      {#if state.monsterConcept && state.currentStep !== 'captioning' && state.currentStep !== 'conceptualizing'}
        <div class="intermediate-result">
          <h4>Monster Concept:</h4>
          <p>{state.monsterConcept}</p>
        </div>
      {/if}
      
      {#if state.imagePrompt && state.currentStep === 'generating'}
        <div class="intermediate-result">
          <h4>Generation Prompt:</h4>
          <p>{state.imagePrompt}</p>
        </div>
      {/if}
    </div>
  {/if}
</div>

<style>
  .monster-generator {
    width: 100%;
    max-width: 1200px;
    margin: 0 auto;
    padding: 2rem;
  }
  
  
  .processing-container {
    display: flex;
    flex-direction: column;
    align-items: center;
    padding: 3rem 1rem;
  }
  
  .spinner {
    width: 60px;
    height: 60px;
    border: 3px solid #f3f3f3;
    border-top: 3px solid #007bff;
    border-radius: 50%;
    animation: spin 1s linear infinite;
    margin-bottom: 2rem;
  }
  
  @keyframes spin {
    0% { transform: rotate(0deg); }
    100% { transform: rotate(360deg); }
  }
  
  .processing-text {
    font-size: 1.2rem;
    color: #333;
    margin-bottom: 2rem;
  }
  
  .intermediate-result {
    width: 100%;
    max-width: 700px;
    background: #f8f9fa;
    padding: 1.5rem;
    border-radius: 8px;
    margin-bottom: 1.5rem;
    text-align: left;
  }
  
  .intermediate-result h4 {
    margin: 0 0 0.5rem 0;
    color: #495057;
  }
  
  .intermediate-result p {
    margin: 0;
    line-height: 1.6;
    color: #333;
  }
</style>