Prism Parallel

https://github.com/yosefhayim/parallel-ai-image-generation
Admin
Prism Parallel is a high-performance multi-agent image generation arena that simultaneously deploys...
Tokens:5,666
Snippets:32
Trust Score:4
Update:7 months ago
Show doc for...
Tokens:
Raw
### Manage AI Models with Model Registry API

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The model registry provides a centralized configuration system for all supported AI models, including metadata about quality, cost, and resolution capabilities. Functions are available to retrieve all models, look up specific models by ID, get display names, check if a model is simulated, and access provider groupings.

```typescript
import {
  getAllModels,
  getModelById,
  getModelLabel,
  isModelSimulated,
  getModelMetadata,
  PROVIDER_GROUPS
} from '@/models/registry';

// Get all available models
const allModels = getAllModels();
console.log(allModels.length); // 22 models available

// Look up a specific model by ID
const model = getModelById('flux:pro');
console.log({
  id: model.id,                    // "flux:pro"
  name: model.name,                // "Flux Pro"
  provider: model.provider,        // "replicate"
  simulated: model.simulated,      // true
  baseModel: model.baseModel,      // "Gemini 3 Pro" (used for simulation)
  metadata: model.metadata         // { quality: "Premium", cost: "External", res: "1024x1024" }
});

// Get display-friendly model name
const label = getModelLabel(AIModelProvider.DALLE_3); // "DALL-E 3"

// Check if a model is simulated (uses prompt modifiers)
const isSimulated = isModelSimulated('midjourney:v6'); // true

// Access provider groupings for UI organization
PROVIDER_GROUPS.forEach(group => {
  console.log(`${group.name} (${group.id}): ${group.color}`);
});
// Output: "Google Gemini (gemini): #4285F4", "OpenAI (openai): #10a37f", etc.
```

--------------------------------

### Initialize Google GenAI Client

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The `getAIClient` function creates and returns a Google GenAI client instance configured with the appropriate API key from local storage or environment variables. It handles potential errors if the API key is not available.

```typescript
import { getAIClient } from '@/services/client';

// Get configured AI client (throws if no API key available)
try {
  const ai = getAIClient();

  // Use client for text generation
  const response = await ai.models.generateContent({
    model: 'gemini-3-flash-preview',
    contents: 'Explain quantum computing in simple terms'
  });

  console.log(response.text);
} catch (error) {
  console.error("Missing Gemini API key. Set GEMINI_API_KEY or connect in Settings.");
}
```

--------------------------------

### AI Client Initialization

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The `getAIClient` function creates and returns a Google GenAI client instance configured with the appropriate API key from local storage or environment variables.

```APIDOC
## GET /api/getAIClient

### Description
Initializes and returns a Google GenAI client instance. The client is configured using an API key found in local storage or environment variables. Throws an error if no API key is available.

### Method
GET (Conceptual - this is a function call, not a direct HTTP endpoint)

### Endpoint
(N/A - This is a client-side function call)

### Parameters
(None - relies on environment variables or local storage for configuration)

### Request Body
(Not applicable)

### Request Example
```typescript
import { getAIClient } from '@/services/client';

// Get configured AI client (throws if no API key available)
try {
  const ai = getAIClient();

  // Use client for text generation
  const response = await ai.models.generateContent({
    model: 'gemini-3-flash-preview',
    contents: 'Explain quantum computing in simple terms'
  });

  console.log(response.text);
} catch (error) {
  console.error("Missing Gemini API key. Set GEMINI_API_KEY or connect in Settings.");
}
```

### Response
#### Success Response (200)
- **aiClient** (object) - An initialized Google GenAI client instance.

#### Response Example
(The `ai` object returned in the example is the successful response.)

#### Error Response (400/500)
- **error** (string) - A message indicating that the API key is missing.
```

--------------------------------

### Configure Import Map Dependencies

Source: https://github.com/yosefhayim/parallel-ai-image-generation/blob/main/index.html

Defines the import map for the project, mapping package names to their respective ESM CDN URLs. This configuration ensures consistent resolution of dependencies like React, Framer Motion, and Google GenAI.

```json
{
  "imports": {
    "@google/genai": "https://esm.sh/@google/genai@^1.37.0",
    "lucide-react": "https://esm.sh/lucide-react@^0.562.0",
    "framer-motion": "https://esm.sh/framer-motion@^11.11.17",
    "react/": "https://esm.sh/react@^19.2.3/",
    "react": "https://esm.sh/react@^19.2.3",
    "react-dom/": "https://esm.sh/react-dom@^19.2.3/",
    "class-variance-authority": "https://esm.sh/class-variance-authority@^0.7.1",
    "clsx": "https://esm.sh/clsx@^2.1.1",
    "tailwind-merge": "https://esm.sh/tailwind-merge@^3.0.1",
    "@radix-ui/react-slot": "https://esm.sh/@radix-ui/react-slot@^1.1.2",
    "@/": "./src/"
  }
}
```

--------------------------------

### Manage AI Provider API Keys

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Service utilities to securely store, retrieve, and clear API keys for various AI providers. Includes logic for environment variable fallbacks for critical keys like Gemini.

```typescript
import { getProviderKey, setProviderKey, clearProviderKey, getGeminiApiKey } from '@/services/keys';

setProviderKey('openai', 'sk-...');
const openaiKey = getProviderKey('openai');
clearProviderKey('openai');

const geminiKey = getGeminiApiKey();
```

--------------------------------

### Refine Prompts for AI Image Generation

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Enhances user prompts by adding descriptive keywords, lighting, and artistic styles optimized for AI models. It returns the refined prompt text and token usage information. Requires the 'refinePrompt' function from '@/services/analysis'.

```typescript
import { refinePrompt } from '@/services/analysis';

// Refine a basic prompt for better generation results
const original = "cat sitting on a couch";
const refined = await refinePrompt(original);

console.log(refined.text);
// Output: "A fluffy tabby cat sitting elegantly on a plush velvet couch,
// soft ambient lighting, shallow depth of field, cinematic composition,
// warm color palette, photorealistic fur texture"

console.log(refined.tokens); // { prompt: 45, candidates: 80, total: 125 }
```

--------------------------------

### Configure Gemini AI Model Providers

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Retrieves model-specific configurations for various Gemini model variants, including aspect ratios and image sizes. It takes an AIModelProvider enum value and returns an object with modelName and config. Requires the 'getGeminiConfig' function from '@/services/providers/gemini'.

```typescript
import { getGeminiConfig } from '@/services/providers/gemini';
import { AIModelProvider } from '@/types';

// Get configuration for Gemini Pro
const proConfig = getGeminiConfig(AIModelProvider.GEMINI_PRO);
console.log({
  modelName: proConfig.modelName,  // "gemini-3-pro-image-preview"
  config: proConfig.config         // { imageSize: "1K", aspectRatio: "1:1" }
});

// Get configuration for cinematic output
const cinemaConfig = getGeminiConfig(AIModelProvider.GEMINI_CINEMA);
console.log({
  modelName: cinemaConfig.modelName,  // "gemini-2.5-flash-image"
  config: cinemaConfig.config         // { aspectRatio: "16:9" }
});

// Get configuration for portrait mode
const verticalConfig = getGeminiConfig(AIModelProvider.GEMINI_VERTICAL);
console.log(verticalConfig.config.aspectRatio); // "9:16"
```

--------------------------------

### Manage Session History with useHistory Hook

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Provides persistent session history management using local storage, including automatic migration from legacy formats. It exposes functions to add items, clear history, and access the current history array. History is capped at 12 items.

```typescript
import { useHistory } from '@/hooks/useHistory';
import { HistoryItem } from '@/types';

function MyComponent() {
  const { history, addHistoryItem, clearHistory } = useHistory();

  // Add a new history entry after generation
  const saveGeneration = (prompt: string, results: GenerationResult[]) => {
    const historyItem: HistoryItem = {
      id: Math.random().toString(36).substr(2, 9),
      timestamp: Date.now(),
      prompt: prompt,
      results: results,
      globalEvaluation: {
        summary: "Analysis complete",
        averageScore: 8.2
      }
    };

    addHistoryItem(historyItem);
    // History is automatically capped at 12 items (MAX_HISTORY_ITEMS)
  };

  // Access stored history
  console.log(`${history.length} generations saved`);

  // Clear all history
  const handlePurge = () => {
    if (confirm("Clear all history?")) {
      clearHistory();
    }
  };
}
```

--------------------------------

### Evaluate Image Alignment with Gemini

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Analyzes generated images against their prompts using Gemini's multimodal capabilities. It returns alignment scores and insights, with token usage details. Requires the 'evaluateImage' function from '@/services/analysis'.

```typescript
import { evaluateImage } from '@/services/analysis';

// Evaluate how well an image matches its prompt
const imageBase64Url = "data:image/png;base64,iVBORw0KGgo...";
const evaluation = await evaluateImage(
  "A serene mountain landscape at dawn",
  imageBase64Url
);

console.log({
  score: evaluation.evaluation.score,     // 8.5 (1-10 scale)
  insight: evaluation.evaluation.insight, // "Strong composition with accurate lighting"
  tokens: evaluation.tokens               // { prompt: 200, candidates: 50, total: 250 }
});
```

--------------------------------

### Model Registry API

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The model registry provides a centralized configuration system for all supported AI models, including metadata about quality, cost, and resolution capabilities.

```APIDOC
## Model Registry Functions

### Description
Provides functions to access and query information about all supported AI models, including their metadata, simulation status, and provider groupings.

### Methods
- `getAllModels()`: Returns an array of all available AI models.
- `getModelById(id: string)`: Retrieves a specific model by its unique ID.
- `getModelLabel(provider: AIModelProvider)`: Gets a human-readable label for a given AI model provider.
- `isModelSimulated(modelId: string)`: Checks if a model is simulated (uses prompt modifiers).
- `getModelMetadata(modelId: string)`: Retrieves metadata for a specific model.

### Parameters
(Parameters are passed as arguments to the respective functions)

### Request Example
```typescript
import {
  getAllModels,
  getModelById,
  getModelLabel,
  isModelSimulated,
  getModelMetadata,
  PROVIDER_GROUPS
} from '@/models/registry';
import { AIModelProvider } from '@/types';

// Get all available models
const allModels = getAllModels();
console.log(allModels.length); // 22 models available

// Look up a specific model by ID
const model = getModelById('flux:pro');
console.log({
  id: model.id,                    // "flux:pro"
  name: model.name,                // "Flux Pro"
  provider: model.provider,        // "replicate"
  simulated: model.simulated,      // true
  baseModel: model.baseModel,      // "Gemini 3 Pro" (used for simulation)
  metadata: model.metadata         // { quality: "Premium", cost: "External", res: "1024x1024" }
});

// Get display-friendly model name
const label = getModelLabel(AIModelProvider.DALLE_3); // "DALL-E 3"

// Check if a model is simulated (uses prompt modifiers)
const isSimulated = isModelSimulated('midjourney:v6'); // true

// Access provider groupings for UI organization
PROVIDER_GROUPS.forEach(group => {
  console.log(`${group.name} (${group.id}): ${group.color}`);
});
// Output: "Google Gemini (gemini): #4285F4", "OpenAI (openai): #10a37f", etc.
```

### Response
(Responses are the direct return values of the functions as shown in the examples.)

#### Success Response (200)
(N/A - These are client-side functions)

#### Response Example
(See Request Example for detailed output structures.)
```

--------------------------------

### Generate Global Summary of Image Generations

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

Aggregates evaluation results from multiple model generations to provide a comparative analysis of swarm performance. It takes the original prompt and an array of generation results, returning a summary text and token usage. Requires the 'getGlobalSummary' function from '@/services/analysis'.

```typescript
import { getGlobalSummary } from '@/services/analysis';
import { GenerationResult } from '@/types';

// After generating images from multiple models
const results: GenerationResult[] = [
  {
    id: '1',
    model: 'Gemini 3 Pro',
    status: 'completed',
    duration: 2500,
    evaluation: { score: 8.5, insight: "Excellent detail" }
  },
  {
    id: '2',
    model: 'DALL-E 3',
    status: 'completed',
    duration: 3200,
    evaluation: { score: 7.8, insight: "Good composition" }
  }
];

const summary = await getGlobalSummary(
  "A futuristic Tokyo street scene",
  results
);

console.log(summary.text);
// Output: "In this parallel generation swarm, Gemini 3 Pro demonstrated
// superior performance with an 8.5 score, excelling in detail rendering.
// DALL-E 3 showed strong compositional skills with a 7.8 rating..."
```

--------------------------------

### Define CSS Gradient Variables

Source: https://github.com/yosefhayim/parallel-ai-image-generation/blob/main/index.html

Sets custom CSS properties for gradient backgrounds and border angles used across various UI components. These variables control color stops, positions, and spread values to create consistent visual themes.

```css
.gradient-badge:hover, .gradient-badge-active { --pos-x: 0%; --pos-y: 91.51%; --spread-x: 120.24%; --spread-y: 103.18%; --color-1: #c96287; --color-2: #c66c64; --color-3: #cc7d23; --color-4: #37140a; --color-5: #000; --border-angle: 190deg; --border-color-1: hsla(340, 78%, 90%, 0.1); --border-color-2: hsla(340, 75%, 90%, 0.6); --stop-1: 0%; --stop-2: 8.8%; --stop-3: 21.44%; --stop-4: 71.34%; --stop-5: 85.76%; }
```

--------------------------------

### Define Core AI Generation Data Structures

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

TypeScript interfaces and enums defining the application's domain models, including aspect ratios, provider identifiers, and generation result schemas.

```typescript
import { AIModelProvider, AspectRatio, GenerationResult, ProviderId } from '@/types';

const ratios: AspectRatio[] = ["1:1", "4:3", "3:4", "16:9", "9:16"];

const result: GenerationResult = {
  id: "abc123",
  model: AIModelProvider.GEMINI_PRO,
  imageUrl: "data:image/png;base64,...",
  duration: 2500,
  status: 'completed',
  evaluation: { score: 8.5, insight: "Excellent detail" },
  tokens: { prompt: 150, candidates: 100, total: 250 }
};
```

--------------------------------

### Core Image Generation Service

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The `generateUnifiedImage` function is the primary entry point for generating images across all supported models. It handles model resolution, prompt modification for simulated providers, and returns comprehensive generation metadata including token usage.

```APIDOC
## POST /api/generateUnifiedImage

### Description
Generates an image using the specified AI model provider. Supports both direct generation and style transfer using a reference image.

### Method
POST

### Endpoint
/api/generateUnifiedImage

### Parameters
#### Query Parameters
- **prompt** (string) - Required - The text prompt for image generation.
- **provider** (AIModelProvider) - Required - The AI model provider to use (e.g., GEMINI_PRO, DALLE_3).
- **referenceImageBase64** (string) - Optional - A Base64 encoded string of a reference image for style transfer.
- **aspectRatio** (string) - Optional - The desired aspect ratio of the generated image (e.g., "16:9", "1:1").

### Request Body
(Not applicable, parameters are passed via query string or function arguments in the example)

### Request Example
```typescript
import { generateUnifiedImage } from '@/services/unified-image-service';
import { AIModelProvider } from '@/types';

// Generate an image with a specific model
const result = await generateUnifiedImage(
  "A cyberpunk cityscape at sunset with neon lights",
  AIModelProvider.GEMINI_PRO,
  undefined,  // optional reference image URL
  "16:9"      // aspect ratio
);

// Generate with a reference image for style transfer
const referenceImageBase64 = "data:image/png;base64,iVBORw0KGgo...";
const styledResult = await generateUnifiedImage(
  "Transform this into a watercolor painting",
  AIModelProvider.GEMINI_FLASH,
  referenceImageBase64,
  "1:1"
);
```

### Response
#### Success Response (200)
- **imageUrl** (string) - Base64 data URL of the generated image.
- **duration** (number) - Generation time in milliseconds.
- **tokens** (object) - Token usage details: `{ prompt: number, candidates: number, total: number }`.
- **provider** (string) - The AI model provider used.
- **modelName** (string) - The specific model name used for generation.
- **simulated** (boolean) - Indicates if the model was simulated (true for non-Gemini models).

#### Response Example
```json
{
  "imageUrl": "data:image/png;base64,iVBORw0KGgo...",
  "duration": 5200,
  "tokens": { "prompt": 150, "candidates": 100, "total": 250 },
  "provider": "gemini",
  "modelName": "Gemini 3 Pro",
  "simulated": false
}
```
```

--------------------------------

### Persist React State with useLocalStorage Hook

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

A type-safe React hook that synchronizes component state with browser local storage. It supports functional updates, default values, and manual removal of stored data.

```typescript
import { useLocalStorage } from '@/hooks/useLocalStorage';

function SettingsComponent() {
  const [theme, setTheme, removeTheme] = useLocalStorage<'light' | 'dark'>(
    'app_theme',
    'dark'
  );

  const toggleTheme = () => {
    setTheme(prev => prev === 'dark' ? 'light' : 'dark');
  };

  const [count, setCount] = useLocalStorage('counter', 0);
  setCount(prev => prev + 1);

  const resetSettings = () => {
    removeTheme();
  };
}
```

--------------------------------

### Generate Unified Image with Prism Parallel

Source: https://context7.com/yosefhayim/parallel-ai-image-generation/llms.txt

The `generateUnifiedImage` function is the primary entry point for generating images across all supported models. It handles model resolution, prompt modification for simulated providers, and returns comprehensive generation metadata including token usage. It can also perform style transfer using a reference image.

```typescript
import { generateUnifiedImage } from '@/services/unified-image-service';
import { AIModelProvider } from '@/types';

// Generate an image with a specific model
const result = await generateUnifiedImage(
  "A cyberpunk cityscape at sunset with neon lights",
  AIModelProvider.GEMINI_PRO,
  undefined,  // optional reference image URL
  "16:9"      // aspect ratio
);

// Result structure
console.log({
  imageUrl: result.imageUrl,      // Base64 data URL: "data:image/png;base64,வைக்"
  duration: result.duration,       // Generation time in milliseconds
  tokens: result.tokens,           // { prompt: 150, candidates: 100, total: 250 }
  provider: result.provider,       // "gemini"
  modelName: result.modelName,     // "Gemini 3 Pro"
  simulated: result.simulated      // false (true for non-Gemini models)
});

// Generate with a reference image for style transfer
const referenceImageBase64 = "data:image/png;base64,iVBORw0KGgo...";
const styledResult = await generateUnifiedImage(
  "Transform this into a watercolor painting",
  AIModelProvider.GEMINI_FLASH,
  referenceImageBase64,
  "1:1"
);
```
Tokens:
Raw