Search

  const input = v.parse(inputSchema, ["dogs", "cats", "felines", "canines"])
  const embeddings = await client.embed({
    input,
    model: "voyage-3-lite",
    return {
      word,
      embedding: embeddings.data?.at(i)?.embedding,
    }
  })

      matches({ type: "embedding", provider: "openai" }),
      async () => {
        const { OpenAIEmbeddings } = await import("https://esm.sh/langchain/embeddings/openai");
        return new OpenAIEmbeddings(args);
      },
    ],
      async () => {
        await import("https://esm.sh/@huggingface/inference");
        const { HuggingFaceInferenceEmbeddings } = await import("https://esm.sh/langchain/embeddings/hf");
        return new HuggingFaceInferenceEmbeddings(args);
      },
    ],

      matches({ type: "embedding", provider: "openai" }),
      async () => {
        const { OpenAIEmbeddings } = await import(
          "npm:langchain/embeddings/openai"
        );
        return new OpenAIEmbeddings(args);
      },
    ],
      async () => {
        await import("npm:@huggingface/inference");
        const { HuggingFaceInferenceEmbeddings } = await import(
          "npm:langchain/embeddings/hf"
        );
        return new HuggingFaceInferenceEmbeddings(args);
      },
    ],

// Get embedding for a single emoji
async function getEmbedding(emoji: string): Promise<number[]> {
  const result = await openai.embeddings.create({
    input: emoji,
    model: "text-embedding-3-small",
// // Get embedding for a single emoji
// async function getEmbedding(emoji: string): Promise<number[]> {
//   const result = await openai.embeddings.create({
//     input: emoji,
//     model: "text-embedding-3-small",
// }
// // Get embeddings for all emojis
// async function getAllEmbeddings(): Promise<EmojiEmbedding[]> {
//   const emojis = getAllEmojis();
//   const embeddings: EmojiEmbedding[] = [];
//   // Process emojis in batches to avoid rate limits
//     const batchResults = await Promise.all(batchPromises);
//     embeddings.push(...batchResults);
//     // Add a small delay between batches
//   }
//   return embeddings;
// }
// function findNearestNeighbors(
//   targetEmbedding: number[],
//   allEmbeddings: EmojiEmbedding[],
//   k: number = 5,
// ): { emoji: string; similarity: number }[] {
//   return allEmbeddings
//     .map(entry => ({
//       emoji: entry.emoji,
// async function main() {
//   try {
//     console.log("Getting embeddings for all emojis...");
//     const allEmbeddings = await getAllEmbeddings();
//     // Example: Find nearest neighbors for 😆
//     console.log(`Finding nearest neighbors for ${targetEmoji}...`);
//     const neighbors = findNearestNeighbors(targetEmbedding, allEmbeddings);
//     console.log("\nNearest neighbors:");

# Emoji Instant Search
Uses vector embeddings to get "vibes" search on emojis

// Generate embedding for a given text
async function generateEmbedding(text: string): Promise<number[]> {
  const response = await openai.embeddings.create({
    model: "text-embedding-ada-002",
    input: text,

      matches({ type: "embedding", provider: "openai" }),
      async () => {
        const { OpenAIEmbeddings } = await import(
          "npm:langchain/embeddings/openai"
        );
        return new OpenAIEmbeddings(args);
      },
    ],
      async () => {
        await import("npm:@huggingface/inference");
        const { HuggingFaceInferenceEmbeddings } = await import(
          "npm:langchain/embeddings/hf"
        );
        return new HuggingFaceInferenceEmbeddings(args);
      },
    ],

      matches({ type: "embedding", provider: "openai" }),
      async () => {
        const { OpenAIEmbeddings } = await import(
          "npm:langchain/embeddings/openai"
        );
        return new OpenAIEmbeddings(args);
      },
    ],
      async () => {
        await import("npm:@huggingface/inference");
        const { HuggingFaceInferenceEmbeddings } = await import(
          "npm:langchain/embeddings/hf"
        );
        return new HuggingFaceInferenceEmbeddings(args);
      },
    ],

app.post("/", async (c) => {
  const body = await c.req.json();
  const { url, opts = 'content', crawlMode = 'jina', useCache, getTags, getTagsPrompt, getSummary, getSummaryPrompt, getEmbeddings } = body;
  try {
    }
  
    const [summary, tags, embeddings] = await Promise.all([
      getSummary ? getSummaryFn(getSummaryPrompt, content) : null,
      getTags ? getTagsFn(getTagsPrompt, content) : null,
      getEmbeddings ? getEmbeddingsFn(content) : null
    ]);
  
    return c.json({ summary: summary?.text, tags: tags?.text, content, embeddings });
  } catch (error) {
    return c.json({ error: error.message || error || "Unknown error."})
export const getBlurbFromUrl = async (url: string, { content,
  getSummary = true, getSummaryPrompt, getTags = true, getTagsPrompt, getEmbeddings,
  model, provider
} = {}) => {
  }
  let summary, tags, embeddings
  if (getSummary) {
    summary = await getSummaryFn(getSummaryPrompt, content, { model, provider });
    tags = await getTagsFn(getTagsPrompt, content, { model, provider });
  }
  if (getEmbeddings) {
    embeddings = await getEmbeddingsFn(content + " " + summary + " " + tags);
  }
    summary: summary.text,
    tags: tags.text,
    embeddings
  }
};
}
export const getEmbeddingsFn = async (content) => {
  let result = await ai({
    provider: "openai",

 * This microservice implements a high-performance dimensionality reduction API using UMAP.
 * It uses the umap-js library for efficient UMAP computation and implements caching for improved performance.
 * The service accepts POST requests with JSON payloads containing high-dimensional embeddings and configuration parameters.
 * It returns 2D coordinates as the result of dimensionality reduction.
 *
 *
 * Common use cases include:
 * - Visualizing word embeddings or document vectors in NLP tasks
 * - Analyzing gene expression data in bioinformatics
 * - Exploring customer segmentation in marketing analytics
 * - Visualizing image embeddings in computer vision tasks
 */
  try {
    const { embeddings, config } = await request.json();
    // Input validation
    if (!Array.isArray(embeddings) || embeddings.length === 0) {
      return new Response("Invalid input: embeddings must be a non-empty array", { status: 400 });
    }
    if (embeddings.length > MAX_POINTS) {
      return new Response(`Input too large: maximum ${MAX_POINTS} points allowed`, { status: 413 });
    }
    if (embeddings[0].length > MAX_DIMENSIONS) {
      return new Response(`Input too high-dimensional: maximum ${MAX_DIMENSIONS} dimensions allowed`, { status: 413 });
    }
    // Generate cache key
    const encoder = new TextEncoder();
    const data = encoder.encode(JSON.stringify({ embeddings, config }));
    const hashBuffer = await crypto.subtle.digest("MD5", data);
    const hashArray = Array.from(new Uint8Array(hashBuffer));
    // Perform UMAP with timeout
    const result = await Promise.race([
      umap.fit(embeddings),
      new Promise((_, reject) => setTimeout(() => reject(new Error("Computation timed out")), TIMEOUT_MS)),
    ]);
    <h2>Common Use Cases</h2>
    <ul>
        <li>Visualizing word embeddings or document vectors in NLP tasks</li>
        <li>Analyzing gene expression data in bioinformatics</li>
        <li>Exploring customer segmentation in marketing analytics</li>
        <li>Visualizing image embeddings in computer vision tasks</li>
    </ul>
    <h2>How to Use</h2>
    <p>Send a POST request to this endpoint with your data to use the API. The request should include an array of embeddings and optional configuration parameters.</p>
    <div class="example">
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
        embeddings: [[1,2,3], [4,5,6], [7,8,9]],
        config: { nNeighbors: 15, minDist: 0.1, spread: 1.0 }
    })
        <h3>Curl Example:</h3>
        <pre>
        curl -X POST -H "Content-Type: application/json" -d '{"embeddings": [[1,2,3], [4,5,6], [7,8,9]], "config": { "nNeighbors": 15, "minDist": 0.1, "spread": 1.0 }}' https://ejfox-umap.web.val.run/
        </pre>
    </div>
    <div class="example">
        <h3>Example with OpenAI Embeddings:</h3>
        <p>This example shows how to use the UMAP service with OpenAI embeddings:</p>
        <pre>
// First, generate embeddings using OpenAI API
import { OpenAI } from "https://esm.town/v/std/openai";
const openai = new OpenAI();
async function getEmbeddings(texts) {
    const response = await openai.embeddings.create({
        model: "text-embedding-ada-002",
        input: texts,
}
// Then, use these embeddings with the UMAP service
const texts = ["Hello world", "OpenAI is amazing", "UMAP reduces dimensions"];
const embeddings = await getEmbeddings(texts);
fetch("/", {
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
        embeddings: embeddings,
        config: { nNeighbors: 15, minDist: 0.1, spread: 1.0 }
    })
        function runDemo() {
            const testData = {
                embeddings: Array.from({length: 100}, () => Array.from({length: 10}, () => Math.random())),
                config: { nNeighbors: 15, minDist: 0.1, spread: 1.0 }
            };