[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AMAMZK2P5Y3MWZVTKHNWTB":3,"skill-same:01M7AMAMZK2P5Y3MWZVTKHNWTB":69},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":34,"commitSha":35,"skillDirectory":5,"forksCount":36,"repositoryHtmlUrl":37,"repositoryHomepageUrl":6,"ownerHtmlUrl":38,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":39,"install":40,"resources":52},"01M7AMAMZK2P5Y3MWZVTKHNWTB","google-gemini-embeddings",null,"This skill provides complete coverage of Google Gemini embeddings API (gemini-embedding-001) for building RAG systems, semantic search, document clustering, and similarity matching. Use when implementing vector search with Google's embedding models, integrating with Cloudflare Vectorize, or building retrieval-augmented generation systems. Covers SDK usage (@google\u002Fgenai), fetch-based Workers implementation, batch processing, 8 task types (RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, SEMANTIC_SIMILARITY, etc.), dimension optimization (128-3072), and cosine similarity calculations. Prevents 8+ embedding-specific errors including dimension mismatches, incorrect task types, rate limiting issues (100 RPM free tier), vector normalization mistakes, text truncation (2,048 token limit), and model version confusion. Includes production-ready RAG patterns with Cloudflare Vectorize integration, chunking strategies, and caching patterns. Token savings: ~60%. Production tested.\n\nKeywords: gemini embeddings, gemini-embedding-001, google embeddings, semantic search, RAG, vector search, document clustering, similarity search, retrieval augmented generation, vectorize integration, cloudflare vectorize embeddings, 768 dimensions, embed content gemini, batch embeddings, embeddings api, cosine similarity, vector normalization, retrieval query, retrieval document, task types, dimension mismatch, embeddings rate limit, text truncation, @google\u002Fgenai","davidanthonyn\u002Fagent-skills","davidanthonyn","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F55152980?v=4","01M7AJ4MY8S5C2RJCGX6626J0G","01M7AJ4MZB7EC7CAP27ANCSE7N",1,"MIT","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fgoogle-gemini-embeddings\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: google-gemini-embeddings\ndescription: |\n  This skill provides complete coverage of Google Gemini embeddings API (gemini-embedding-001) for building RAG systems, semantic search, document clustering, and similarity matching. Use when implementing vector search with Google's embedding models, integrating with Cloudflare Vectorize, or building retrieval-augmented generation systems. Covers SDK usage (@google\u002Fgenai), fetch-based Workers implementation, batch processing, 8 task types (RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, SEMANTIC_SIMILARITY, etc.), dimension optimization (128-3072), and cosine similarity calculations. Prevents 8+ embedding-specific errors including dimension mismatches, incorrect task types, rate limiting issues (100 RPM free tier), vector normalization mistakes, text truncation (2,048 token limit), and model version confusion. Includes production-ready RAG patterns with Cloudflare Vectorize integration, chunking strategies, and caching patterns. Token savings: ~60%. Production tested.\n\n  Keywords: gemini embeddings, gemini-embedding-001, google embeddings, semantic search, RAG, vector search, document clustering, similarity search, retrieval augmented generation, vectorize integration, cloudflare vectorize embeddings, 768 dimensions, embed content gemini, batch embeddings, embeddings api, cosine similarity, vector normalization, retrieval query, retrieval document, task types, dimension mismatch, embeddings rate limit, text truncation, @google\u002Fgenai\nlicense: MIT\nmetadata:\n  version: 1.0.0\n  last_updated: 2025-10-25\n  tested_package_version: \"@google\u002Fgenai@1.27.0\"\n  target_audience: \"Developers building RAG, semantic search, or vector-based applications\"\n  complexity: intermediate\n  estimated_reading_time: \"15 minutes\"\n  tokens_saved: \"~60%\"\n  errors_prevented: 8\n  production_tested: true\n---\n\n# Google Gemini Embeddings\n\n**Complete production-ready guide for Google Gemini embeddings API**\n\nThis skill provides comprehensive coverage of the `gemini-embedding-001` model for generating text embeddings, including SDK usage, REST API patterns, batch processing, RAG integration with Cloudflare Vectorize, and advanced use cases like semantic search and document clustering.\n\n---\n\n## Table of Contents\n\n1. [Quick Start](#1-quick-start)\n2. [gemini-embedding-001 Model](#2-gemini-embedding-001-model)\n3. [Basic Embeddings](#3-basic-embeddings)\n4. [Batch Embeddings](#4-batch-embeddings)\n5. [Task Types](#5-task-types)\n6. [RAG Patterns](#6-rag-patterns)\n7. [Semantic Search](#7-semantic-search)\n8. [Document Clustering](#8-document-clustering)\n9. [Error Handling](#9-error-handling)\n10. [Best Practices](#10-best-practices)\n\n---\n\n## 1. Quick Start\n\n### Installation\n\nInstall the Google Generative AI SDK:\n\n```bash\nnpm install @google\u002Fgenai@^1.27.0\n```\n\nFor TypeScript projects:\n\n```bash\nnpm install -D typescript@^5.0.0\n```\n\n### Environment Setup\n\nSet your Gemini API key as an environment variable:\n\n```bash\nexport GEMINI_API_KEY=\"your-api-key-here\"\n```\n\nGet your API key from: https:\u002F\u002Faistudio.google.com\u002Fapikey\n\n### First Embedding Example\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'What is the meaning of life?',\n  config: {\n    taskType: 'RETRIEVAL_QUERY',\n    outputDimensionality: 768\n  }\n});\n\nconsole.log(response.embedding.values); \u002F\u002F [0.012, -0.034, ...]\nconsole.log(response.embedding.values.length); \u002F\u002F 768\n```\n\n**Result**: A 768-dimension embedding vector representing the semantic meaning of the text.\n\n---\n\n## 2. gemini-embedding-001 Model\n\n### Model Specifications\n\n**Current Model**: `gemini-embedding-001` (stable, production-ready)\n- **Status**: Stable\n- **Experimental**: `gemini-embedding-exp-03-07` (deprecated October 2025, do not use)\n\n### Dimensions\n\nThe model supports flexible output dimensionality using **Matryoshka Representation Learning**:\n\n| Dimension | Use Case | Storage | Performance |\n|-----------|----------|---------|-------------|\n| **768** | Recommended for most use cases | Low | Fast |\n| **1536** | Balance between accuracy and efficiency | Medium | Medium |\n| **3072** | Maximum accuracy (default) | High | Slower |\n| 128-3071 | Custom (any value in range) | Variable | Variable |\n\n**Default**: 3072 dimensions\n**Recommended**: 768, 1536, or 3072 for optimal performance\n\n### Context Window\n\n- **Input Limit**: 2,048 tokens per text\n- **Input Type**: Text only (no images, audio, or video)\n\n### Rate Limits\n\n| Tier | RPM | TPM | RPD | Requirements |\n|------|-----|-----|-----|--------------|\n| **Free** | 100 | 30,000 | 1,000 | No billing account |\n| **Tier 1** | 3,000 | 1,000,000 | - | Billing account linked |\n| **Tier 2** | 5,000 | 5,000,000 | - | $250+ spending, 30-day wait |\n| **Tier 3** | 10,000 | 10,000,000 | - | $1,000+ spending, 30-day wait |\n\n**RPM** = Requests Per Minute\n**TPM** = Tokens Per Minute\n**RPD** = Requests Per Day\n\n### Output Format\n\n```typescript\n{\n  embedding: {\n    values: number[] \u002F\u002F Array of floating-point numbers\n  }\n}\n```\n\n---\n\n## 3. Basic Embeddings\n\n### SDK Approach (Node.js)\n\n**Single text embedding**:\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'The quick brown fox jumps over the lazy dog',\n  config: {\n    taskType: 'SEMANTIC_SIMILARITY',\n    outputDimensionality: 768\n  }\n});\n\nconsole.log(response.embedding.values);\n\u002F\u002F [0.00388, -0.00762, 0.01543, ...]\n```\n\n### Fetch Approach (Cloudflare Workers)\n\n**For Workers\u002Fedge environments without SDK support**:\n\n```typescript\nexport default {\n  async fetch(request: Request, env: Env): Promise\u003CResponse> {\n    const apiKey = env.GEMINI_API_KEY;\n    const text = \"What is the meaning of life?\";\n\n    const response = await fetch(\n      'https:\u002F\u002Fgenerativelanguage.googleapis.com\u002Fv1beta\u002Fmodels\u002Fgemini-embedding-001:embedContent',\n      {\n        method: 'POST',\n        headers: {\n          'x-goog-api-key': apiKey,\n          'Content-Type': 'application\u002Fjson'\n        },\n        body: JSON.stringify({\n          content: {\n            parts: [{ text }]\n          },\n          taskType: 'RETRIEVAL_QUERY',\n          outputDimensionality: 768\n        })\n      }\n    );\n\n    const data = await response.json();\n\n    \u002F\u002F Response format:\n    \u002F\u002F {\n    \u002F\u002F   embedding: {\n    \u002F\u002F     values: [0.012, -0.034, ...]\n    \u002F\u002F   }\n    \u002F\u002F }\n\n    return new Response(JSON.stringify(data), {\n      headers: { 'Content-Type': 'application\u002Fjson' }\n    });\n  }\n};\n```\n\n### Response Parsing\n\n```typescript\ninterface EmbeddingResponse {\n  embedding: {\n    values: number[];\n  };\n}\n\nconst response: EmbeddingResponse = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'Sample text',\n  config: { taskType: 'SEMANTIC_SIMILARITY' }\n});\n\nconst embedding: number[] = response.embedding.values;\nconst dimensions: number = embedding.length; \u002F\u002F 3072 by default\n```\n\n---\n\n## 4. Batch Embeddings\n\n### Multiple Texts in One Request (SDK)\n\nGenerate embeddings for multiple texts simultaneously:\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst texts = [\n  \"What is the meaning of life?\",\n  \"How does photosynthesis work?\",\n  \"Tell me about the history of the internet.\"\n];\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  contents: texts, \u002F\u002F Array of strings\n  config: {\n    taskType: 'RETRIEVAL_DOCUMENT',\n    outputDimensionality: 768\n  }\n});\n\n\u002F\u002F Process each embedding\nresponse.embeddings.forEach((embedding, index) => {\n  console.log(`Text ${index}: ${texts[index]}`);\n  console.log(`Embedding: ${embedding.values.slice(0, 5)}...`);\n  console.log(`Dimensions: ${embedding.values.length}`);\n});\n```\n\n### Batch REST API (fetch)\n\nUse the `batchEmbedContents` endpoint:\n\n```typescript\nconst response = await fetch(\n  'https:\u002F\u002Fgenerativelanguage.googleapis.com\u002Fv1beta\u002Fmodels\u002Fgemini-embedding-001:batchEmbedContents',\n  {\n    method: 'POST',\n    headers: {\n      'x-goog-api-key': apiKey,\n      'Content-Type': 'application\u002Fjson'\n    },\n    body: JSON.stringify({\n      requests: texts.map(text => ({\n        model: 'models\u002Fgemini-embedding-001',\n        content: {\n          parts: [{ text }]\n        },\n        taskType: 'RETRIEVAL_DOCUMENT'\n      }))\n    })\n  }\n);\n\nconst data = await response.json();\n\u002F\u002F data.embeddings: Array of {values: number[]}\n```\n\n### Chunking for Rate Limits\n\nWhen processing large datasets, chunk requests to stay within rate limits:\n\n```typescript\nasync function batchEmbedWithRateLimit(\n  texts: string[],\n  batchSize: number = 100, \u002F\u002F Free tier: 100 RPM\n  delayMs: number = 60000 \u002F\u002F 1 minute delay between batches\n): Promise\u003Cnumber[][]> {\n  const allEmbeddings: number[][] = [];\n\n  for (let i = 0; i \u003C texts.length; i += batchSize) {\n    const batch = texts.slice(i, i + batchSize);\n\n    console.log(`Processing batch ${i \u002F batchSize + 1} (${batch.length} texts)`);\n\n    const response = await ai.models.embedContent({\n      model: 'gemini-embedding-001',\n      contents: batch,\n      config: {\n        taskType: 'RETRIEVAL_DOCUMENT',\n        outputDimensionality: 768\n      }\n    });\n\n    allEmbeddings.push(...response.embeddings.map(e => e.values));\n\n    \u002F\u002F Wait before next batch (except last batch)\n    if (i + batchSize \u003C texts.length) {\n      await new Promise(resolve => setTimeout(resolve, delayMs));\n    }\n  }\n\n  return allEmbeddings;\n}\n\n\u002F\u002F Usage\nconst embeddings = await batchEmbedWithRateLimit(documents, 100);\n```\n\n### Performance Optimization\n\n**Tips**:\n1. Use batch API when embedding multiple texts (single request vs multiple requests)\n2. Choose lower dimensions (768) for faster processing and less storage\n3. Implement exponential backoff for rate limit errors\n4. Cache embeddings to avoid redundant API calls\n\n---\n\n## 5. Task Types\n\nThe `taskType` parameter optimizes embeddings for specific use cases. **Always specify a task type for best results.**\n\n### Available Task Types (8 total)\n\n| Task Type | Use Case | Example |\n|-----------|----------|---------|\n| **RETRIEVAL_QUERY** | User search queries | \"How do I fix a flat tire?\" |\n| **RETRIEVAL_DOCUMENT** | Documents to be indexed\u002Fsearched | Product descriptions, articles |\n| **SEMANTIC_SIMILARITY** | Comparing text similarity | Duplicate detection, clustering |\n| **CLASSIFICATION** | Categorizing texts | Spam detection, sentiment analysis |\n| **CLUSTERING** | Grouping similar texts | Topic modeling, content organization |\n| **CODE_RETRIEVAL_QUERY** | Code search queries | \"function to sort array\" |\n| **QUESTION_ANSWERING** | Questions seeking answers | FAQ matching |\n| **FACT_VERIFICATION** | Verifying claims with evidence | Fact-checking systems |\n\n### When to Use Which\n\n**RAG Systems** (Retrieval Augmented Generation):\n```typescript\n\u002F\u002F When embedding user queries\nconst queryEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery,\n  config: { taskType: 'RETRIEVAL_QUERY' } \u002F\u002F ← Use RETRIEVAL_QUERY\n});\n\n\u002F\u002F When embedding documents for indexing\nconst docEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: documentText,\n  config: { taskType: 'RETRIEVAL_DOCUMENT' } \u002F\u002F ← Use RETRIEVAL_DOCUMENT\n});\n```\n\n**Semantic Search**:\n```typescript\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'SEMANTIC_SIMILARITY' }\n});\n```\n\n**Document Clustering**:\n```typescript\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'CLUSTERING' }\n});\n```\n\n### Impact on Quality\n\nUsing the correct task type **significantly improves** retrieval quality:\n\n```typescript\n\u002F\u002F ❌ BAD: No task type specified\nconst embedding1 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery\n});\n\n\u002F\u002F ✅ GOOD: Task type specified\nconst embedding2 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery,\n  config: { taskType: 'RETRIEVAL_QUERY' }\n});\n```\n\n**Result**: Using the right task type can improve search relevance by 10-30%.\n\n---\n\n## 6. RAG Patterns\n\n**RAG** (Retrieval Augmented Generation) combines vector search with LLM generation to create AI systems that answer questions using custom knowledge bases.\n\n### Complete RAG Workflow\n\n```\n1. Document Ingestion\n   ├── Chunk documents into smaller pieces\n   ├── Generate embeddings (RETRIEVAL_DOCUMENT)\n   └── Store in Vectorize\n\n2. Query Processing\n   ├── User submits query\n   ├── Generate query embedding (RETRIEVAL_QUERY)\n   └── Search Vectorize for similar documents\n\n3. Response Generation\n   ├── Retrieve top-k similar documents\n   ├── Pass documents as context to LLM\n   └── Stream response to user\n```\n\n### Document Ingestion Pipeline\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\n\u002F\u002F 1. Chunk document into smaller pieces\nfunction chunkDocument(text: string, chunkSize: number = 500): string[] {\n  const words = text.split(' ');\n  const chunks: string[] = [];\n\n  for (let i = 0; i \u003C words.length; i += chunkSize) {\n    chunks.push(words.slice(i, i + chunkSize).join(' '));\n  }\n\n  return chunks;\n}\n\n\u002F\u002F 2. Generate embeddings for chunks\nasync function embedChunks(chunks: string[]): Promise\u003Cnumber[][]> {\n  const response = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    contents: chunks,\n    config: {\n      taskType: 'RETRIEVAL_DOCUMENT', \u002F\u002F ← Documents for indexing\n      outputDimensionality: 768 \u002F\u002F ← Match Vectorize index dimensions\n    }\n  });\n\n  return response.embeddings.map(e => e.values);\n}\n\n\u002F\u002F 3. Store in Cloudflare Vectorize\nasync function storeInVectorize(\n  env: Env,\n  chunks: string[],\n  embeddings: number[][]\n) {\n  const vectors = chunks.map((chunk, i) => ({\n    id: `doc-${Date.now()}-${i}`,\n    values: embeddings[i],\n    metadata: { text: chunk }\n  }));\n\n  await env.VECTORIZE.insert(vectors);\n}\n\n\u002F\u002F Complete pipeline\nasync function ingestDocument(env: Env, documentText: string) {\n  const chunks = chunkDocument(documentText, 500);\n  const embeddings = await embedChunks(chunks);\n  await storeInVectorize(env, chunks, embeddings);\n\n  console.log(`Ingested ${chunks.length} chunks`);\n}\n```\n\n### Query Flow (Retrieve + Generate)\n\n```typescript\nasync function ragQuery(env: Env, userQuery: string): Promise\u003Cstring> {\n  \u002F\u002F 1. Embed user query\n  const queryResponse = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    content: userQuery,\n    config: {\n      taskType: 'RETRIEVAL_QUERY', \u002F\u002F ← Query, not document\n      outputDimensionality: 768\n    }\n  });\n\n  const queryEmbedding = queryResponse.embedding.values;\n\n  \u002F\u002F 2. Search Vectorize for similar documents\n  const results = await env.VECTORIZE.query(queryEmbedding, {\n    topK: 5,\n    returnMetadata: true\n  });\n\n  \u002F\u002F 3. Extract context from top results\n  const context = results.matches\n    .map(match => match.metadata.text)\n    .join('\\n\\n');\n\n  \u002F\u002F 4. Generate response with context\n  const response = await ai.models.generateContent({\n    model: 'gemini-2.5-flash',\n    contents: `Context:\\n${context}\\n\\nQuestion: ${userQuery}\\n\\nAnswer based on the context above:`\n  });\n\n  return response.text;\n}\n```\n\n### Integration with Cloudflare Vectorize\n\n**Create Vectorize Index** (768 dimensions for Gemini):\n\n```bash\nnpx wrangler vectorize create gemini-embeddings --dimensions 768 --metric cosine\n```\n\n**Bind in wrangler.jsonc**:\n\n```jsonc\n{\n  \"name\": \"my-rag-app\",\n  \"main\": \"src\u002Findex.ts\",\n  \"compatibility_date\": \"2025-10-25\",\n  \"vectorize\": {\n    \"bindings\": [\n      {\n        \"binding\": \"VECTORIZE\",\n        \"index_name\": \"gemini-embeddings\"\n      }\n    ]\n  }\n}\n```\n\n**Complete RAG Worker**:\n\nSee `templates\u002Frag-with-vectorize.ts` for full implementation.\n\n---\n\n## 7. Semantic Search\n\nSemantic search finds content based on **meaning**, not just keyword matching.\n\n### Cosine Similarity\n\nCosine similarity measures how similar two embeddings are (range: -1 to 1, where 1 = identical):\n\n```typescript\nfunction cosineSimilarity(a: number[], b: number[]): number {\n  if (a.length !== b.length) {\n    throw new Error('Vectors must have same length');\n  }\n\n  let dotProduct = 0;\n  let magnitudeA = 0;\n  let magnitudeB = 0;\n\n  for (let i = 0; i \u003C a.length; i++) {\n    dotProduct += a[i] * b[i];\n    magnitudeA += a[i] * a[i];\n    magnitudeB += b[i] * b[i];\n  }\n\n  if (magnitudeA === 0 || magnitudeB === 0) {\n    return 0;\n  }\n\n  return dotProduct \u002F (Math.sqrt(magnitudeA) * Math.sqrt(magnitudeB));\n}\n```\n\n### Vector Normalization\n\nNormalize vectors to unit length for faster similarity calculations:\n\n```typescript\nfunction normalizeVector(vector: number[]): number[] {\n  const magnitude = Math.sqrt(vector.reduce((sum, val) => sum + val * val, 0));\n\n  if (magnitude === 0) {\n    return vector;\n  }\n\n  return vector.map(val => val \u002F magnitude);\n}\n\n\u002F\u002F Normalized vectors allow dot product instead of cosine similarity\nfunction dotProduct(a: number[], b: number[]): number {\n  return a.reduce((sum, val, i) => sum + val * b[i], 0);\n}\n```\n\n### Top-K Search\n\nFind the most similar documents to a query:\n\n```typescript\ninterface Document {\n  id: string;\n  text: string;\n  embedding: number[];\n}\n\nfunction topKSimilar(\n  queryEmbedding: number[],\n  documents: Document[],\n  k: number = 5\n): Array\u003C{ document: Document; similarity: number }> {\n  \u002F\u002F Calculate similarity for each document\n  const similarities = documents.map(doc => ({\n    document: doc,\n    similarity: cosineSimilarity(queryEmbedding, doc.embedding)\n  }));\n\n  \u002F\u002F Sort by similarity (descending) and return top K\n  return similarities\n    .sort((a, b) => b.similarity - a.similarity)\n    .slice(0, k);\n}\n\n\u002F\u002F Usage\nconst results = topKSimilar(queryEmbedding, documents, 5);\n\nresults.forEach(result => {\n  console.log(`Similarity: ${result.similarity.toFixed(4)}`);\n  console.log(`Text: ${result.document.text}\\n`);\n});\n```\n\n### Complete Semantic Search Example\n\nSee `templates\u002Fsemantic-search.ts` for full implementation with Gemini API.\n\n---\n\n## 8. Document Clustering\n\nClustering groups similar documents together automatically.\n\n### K-Means Clustering\n\n```typescript\ninterface Cluster {\n  centroid: number[];\n  documents: number[][];\n}\n\nfunction kMeansClustering(\n  embeddings: number[][],\n  k: number = 3,\n  maxIterations: number = 100\n): Cluster[] {\n  \u002F\u002F 1. Initialize centroids randomly\n  const centroids: number[][] = [];\n  for (let i = 0; i \u003C k; i++) {\n    centroids.push(embeddings[Math.floor(Math.random() * embeddings.length)]);\n  }\n\n  \u002F\u002F 2. Iterate until convergence\n  for (let iter = 0; iter \u003C maxIterations; iter++) {\n    \u002F\u002F Assign each embedding to nearest centroid\n    const clusters: number[][][] = Array(k).fill(null).map(() => []);\n\n    embeddings.forEach(embedding => {\n      let minDistance = Infinity;\n      let closestCluster = 0;\n\n      centroids.forEach((centroid, i) => {\n        const distance = 1 - cosineSimilarity(embedding, centroid);\n        if (distance \u003C minDistance) {\n          minDistance = distance;\n          closestCluster = i;\n        }\n      });\n\n      clusters[closestCluster].push(embedding);\n    });\n\n    \u002F\u002F Update centroids\n    let changed = false;\n    clusters.forEach((cluster, i) => {\n      if (cluster.length > 0) {\n        const newCentroid = cluster[0].map((_, dim) =>\n          cluster.reduce((sum, emb) => sum + emb[dim], 0) \u002F cluster.length\n        );\n\n        if (cosineSimilarity(centroids[i], newCentroid) \u003C 0.9999) {\n          changed = true;\n        }\n\n        centroids[i] = newCentroid;\n      }\n    });\n\n    if (!changed) break;\n  }\n\n  \u002F\u002F Build final clusters\n  const finalClusters: Cluster[] = centroids.map((centroid, i) => ({\n    centroid,\n    documents: embeddings.filter(emb =>\n      cosineSimilarity(emb, centroid) === Math.max(\n        ...centroids.map(c => cosineSimilarity(emb, c))\n      )\n    )\n  }));\n\n  return finalClusters;\n}\n```\n\nSee `templates\u002Fclustering.ts` for complete implementation with examples.\n\n### Use Cases\n\n1. **Content Organization**: Automatically group similar articles, products, or documents\n2. **Duplicate Detection**: Find near-duplicate content\n3. **Topic Modeling**: Discover themes in large text corpora\n4. **Customer Feedback Analysis**: Group similar customer complaints or feature requests\n\n---\n\n## 9. Error Handling\n\n### Common Errors\n\n**1. API Key Missing or Invalid**\n\n```typescript\n\u002F\u002F ❌ Error: API key not set\nconst ai = new GoogleGenAI({});\n\n\u002F\u002F ✅ Correct\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nif (!process.env.GEMINI_API_KEY) {\n  throw new Error('GEMINI_API_KEY environment variable not set');\n}\n```\n\n**2. Dimension Mismatch**\n\n```typescript\n\u002F\u002F ❌ Error: Embedding has 3072 dims, Vectorize expects 768\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text\n  \u002F\u002F No outputDimensionality specified → defaults to 3072\n});\n\nawait env.VECTORIZE.insert([{\n  id: '1',\n  values: embedding.embedding.values \u002F\u002F 3072 dims, but index is 768!\n}]);\n\n\u002F\u002F ✅ Correct: Match dimensions\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { outputDimensionality: 768 } \u002F\u002F ← Match index dimensions\n});\n```\n\n**3. Rate Limiting**\n\n```typescript\n\u002F\u002F ❌ Error: 429 Too Many Requests\nfor (let i = 0; i \u003C 1000; i++) {\n  await ai.models.embedContent({ \u002F* ... *\u002F }); \u002F\u002F Exceeds 100 RPM on free tier\n}\n\n\u002F\u002F ✅ Correct: Implement rate limiting\nasync function embedWithRetry(text: string, maxRetries = 3) {\n  for (let attempt = 0; attempt \u003C maxRetries; attempt++) {\n    try {\n      return await ai.models.embedContent({\n        model: 'gemini-embedding-001',\n        content: text,\n        config: { taskType: 'SEMANTIC_SIMILARITY' }\n      });\n    } catch (error: any) {\n      if (error.status === 429 && attempt \u003C maxRetries - 1) {\n        const delay = Math.pow(2, attempt) * 1000; \u002F\u002F Exponential backoff\n        await new Promise(resolve => setTimeout(resolve, delay));\n        continue;\n      }\n      throw error;\n    }\n  }\n}\n```\n\nSee `references\u002Ftop-errors.md` for all 8 documented errors with detailed solutions.\n\n---\n\n## 10. Best Practices\n\n### Always Do\n\n✅ **Specify Task Type**\n```typescript\n\u002F\u002F Task type optimizes embeddings for your use case\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'RETRIEVAL_QUERY' } \u002F\u002F ← Always specify\n});\n```\n\n✅ **Match Dimensions with Vectorize**\n```typescript\n\u002F\u002F Ensure embeddings match your Vectorize index dimensions\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { outputDimensionality: 768 } \u002F\u002F ← Match index\n});\n```\n\n✅ **Implement Rate Limiting**\n```typescript\n\u002F\u002F Use exponential backoff for 429 errors\nasync function embedWithBackoff(text: string) {\n  \u002F\u002F Implementation from Error Handling section\n}\n```\n\n✅ **Cache Embeddings**\n```typescript\n\u002F\u002F Cache embeddings to avoid redundant API calls\nconst cache = new Map\u003Cstring, number[]>();\n\nasync function getCachedEmbedding(text: string): Promise\u003Cnumber[]> {\n  if (cache.has(text)) {\n    return cache.get(text)!;\n  }\n\n  const response = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    content: text,\n    config: { taskType: 'SEMANTIC_SIMILARITY' }\n  });\n\n  const embedding = response.embedding.values;\n  cache.set(text, embedding);\n  return embedding;\n}\n```\n\n✅ **Use Batch API for Multiple Texts**\n```typescript\n\u002F\u002F Single batch request vs multiple individual requests\nconst embeddings = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  contents: texts, \u002F\u002F Array of texts\n  config: { taskType: 'RETRIEVAL_DOCUMENT' }\n});\n```\n\n### Never Do\n\n❌ **Don't Skip Task Type**\n```typescript\n\u002F\u002F Reduces quality by 10-30%\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text\n  \u002F\u002F Missing taskType!\n});\n```\n\n❌ **Don't Mix Different Dimensions**\n```typescript\n\u002F\u002F Can't compare embeddings with different dimensions\nconst emb1 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text1,\n  config: { outputDimensionality: 768 }\n});\n\nconst emb2 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text2,\n  config: { outputDimensionality: 1536 } \u002F\u002F Different dimensions!\n});\n\n\u002F\u002F ❌ Can't calculate similarity between different dimensions\nconst similarity = cosineSimilarity(emb1.embedding.values, emb2.embedding.values);\n```\n\n❌ **Don't Use Wrong Task Type for RAG**\n```typescript\n\u002F\u002F Reduces search quality\nconst queryEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: query,\n  config: { taskType: 'RETRIEVAL_DOCUMENT' } \u002F\u002F Wrong! Should be RETRIEVAL_QUERY\n});\n```\n\n---\n\n## Using Bundled Resources\n\n### Templates (templates\u002F)\n\n- `package.json` - Package configuration with verified versions\n- `basic-embeddings.ts` - Single text embedding with SDK\n- `embeddings-fetch.ts` - Fetch-based for Cloudflare Workers\n- `batch-embeddings.ts` - Batch processing with rate limiting\n- `rag-with-vectorize.ts` - Complete RAG implementation with Vectorize\n- `semantic-search.ts` - Cosine similarity and top-K search\n- `clustering.ts` - K-means clustering implementation\n\n### References (references\u002F)\n\n- `model-comparison.md` - Compare Gemini vs OpenAI vs Workers AI embeddings\n- `vectorize-integration.md` - Cloudflare Vectorize setup and patterns\n- `rag-patterns.md` - Complete RAG implementation strategies\n- `dimension-guide.md` - Choosing the right dimensions (768 vs 1536 vs 3072)\n- `top-errors.md` - 8 common errors and detailed solutions\n\n### Scripts (scripts\u002F)\n\n- `check-versions.sh` - Verify @google\u002Fgenai package version is current\n\n---\n\n## Official Documentation\n\n- **Embeddings Guide**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Fembeddings\n- **Model Spec**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Fmodels\u002Fgemini#gemini-embedding-001\n- **Rate Limits**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Frate-limits\n- **SDK Reference**: https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@google\u002Fgenai\n- **Context7 Library ID**: `\u002Fwebsites\u002Fai_google_dev_gemini-api`\n\n---\n\n## Related Skills\n\n- **google-gemini-api** - Main Gemini API for text\u002Fimage generation\n- **cloudflare-vectorize** - Vector database for storing embeddings\n- **cloudflare-workers-ai** - Workers AI embeddings (BGE models)\n\n---\n\n## Success Metrics\n\n**Token Savings**: ~60% compared to manual implementation\n**Errors Prevented**: 8 documented errors with solutions\n**Production Tested**: ✅ Verified in RAG applications\n**Package Version**: @google\u002Fgenai@1.27.0\n**Last Updated**: 2025-10-25\n\n---\n\n## License\n\nMIT License - Free to use in personal and commercial projects.\n\n---\n\n**Questions or Issues?**\n\n- GitHub: https:\u002F\u002Fgithub.com\u002Fjezweb\u002Fclaude-skills\n- Email: jeremy@jezweb.net\n","# Google Gemini Embeddings\n\n**Complete production-ready guide for Google Gemini embeddings API**\n\nThis skill provides comprehensive coverage of the `gemini-embedding-001` model for generating text embeddings, including SDK usage, REST API patterns, batch processing, RAG integration with Cloudflare Vectorize, and advanced use cases like semantic search and document clustering.\n\n---\n\n## Table of Contents\n\n1. [Quick Start](#1-quick-start)\n2. [gemini-embedding-001 Model](#2-gemini-embedding-001-model)\n3. [Basic Embeddings](#3-basic-embeddings)\n4. [Batch Embeddings](#4-batch-embeddings)\n5. [Task Types](#5-task-types)\n6. [RAG Patterns](#6-rag-patterns)\n7. [Semantic Search](#7-semantic-search)\n8. [Document Clustering](#8-document-clustering)\n9. [Error Handling](#9-error-handling)\n10. [Best Practices](#10-best-practices)\n\n---\n\n## 1. Quick Start\n\n### Installation\n\nInstall the Google Generative AI SDK:\n\n```bash\nnpm install @google\u002Fgenai@^1.27.0\n```\n\nFor TypeScript projects:\n\n```bash\nnpm install -D typescript@^5.0.0\n```\n\n### Environment Setup\n\nSet your Gemini API key as an environment variable:\n\n```bash\nexport GEMINI_API_KEY=\"your-api-key-here\"\n```\n\nGet your API key from: https:\u002F\u002Faistudio.google.com\u002Fapikey\n\n### First Embedding Example\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'What is the meaning of life?',\n  config: {\n    taskType: 'RETRIEVAL_QUERY',\n    outputDimensionality: 768\n  }\n});\n\nconsole.log(response.embedding.values); \u002F\u002F [0.012, -0.034, ...]\nconsole.log(response.embedding.values.length); \u002F\u002F 768\n```\n\n**Result**: A 768-dimension embedding vector representing the semantic meaning of the text.\n\n---\n\n## 2. gemini-embedding-001 Model\n\n### Model Specifications\n\n**Current Model**: `gemini-embedding-001` (stable, production-ready)\n- **Status**: Stable\n- **Experimental**: `gemini-embedding-exp-03-07` (deprecated October 2025, do not use)\n\n### Dimensions\n\nThe model supports flexible output dimensionality using **Matryoshka Representation Learning**:\n\n| Dimension | Use Case | Storage | Performance |\n|-----------|----------|---------|-------------|\n| **768** | Recommended for most use cases | Low | Fast |\n| **1536** | Balance between accuracy and efficiency | Medium | Medium |\n| **3072** | Maximum accuracy (default) | High | Slower |\n| 128-3071 | Custom (any value in range) | Variable | Variable |\n\n**Default**: 3072 dimensions\n**Recommended**: 768, 1536, or 3072 for optimal performance\n\n### Context Window\n\n- **Input Limit**: 2,048 tokens per text\n- **Input Type**: Text only (no images, audio, or video)\n\n### Rate Limits\n\n| Tier | RPM | TPM | RPD | Requirements |\n|------|-----|-----|-----|--------------|\n| **Free** | 100 | 30,000 | 1,000 | No billing account |\n| **Tier 1** | 3,000 | 1,000,000 | - | Billing account linked |\n| **Tier 2** | 5,000 | 5,000,000 | - | $250+ spending, 30-day wait |\n| **Tier 3** | 10,000 | 10,000,000 | - | $1,000+ spending, 30-day wait |\n\n**RPM** = Requests Per Minute\n**TPM** = Tokens Per Minute\n**RPD** = Requests Per Day\n\n### Output Format\n\n```typescript\n{\n  embedding: {\n    values: number[] \u002F\u002F Array of floating-point numbers\n  }\n}\n```\n\n---\n\n## 3. Basic Embeddings\n\n### SDK Approach (Node.js)\n\n**Single text embedding**:\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'The quick brown fox jumps over the lazy dog',\n  config: {\n    taskType: 'SEMANTIC_SIMILARITY',\n    outputDimensionality: 768\n  }\n});\n\nconsole.log(response.embedding.values);\n\u002F\u002F [0.00388, -0.00762, 0.01543, ...]\n```\n\n### Fetch Approach (Cloudflare Workers)\n\n**For Workers\u002Fedge environments without SDK support**:\n\n```typescript\nexport default {\n  async fetch(request: Request, env: Env): Promise\u003CResponse> {\n    const apiKey = env.GEMINI_API_KEY;\n    const text = \"What is the meaning of life?\";\n\n    const response = await fetch(\n      'https:\u002F\u002Fgenerativelanguage.googleapis.com\u002Fv1beta\u002Fmodels\u002Fgemini-embedding-001:embedContent',\n      {\n        method: 'POST',\n        headers: {\n          'x-goog-api-key': apiKey,\n          'Content-Type': 'application\u002Fjson'\n        },\n        body: JSON.stringify({\n          content: {\n            parts: [{ text }]\n          },\n          taskType: 'RETRIEVAL_QUERY',\n          outputDimensionality: 768\n        })\n      }\n    );\n\n    const data = await response.json();\n\n    \u002F\u002F Response format:\n    \u002F\u002F {\n    \u002F\u002F   embedding: {\n    \u002F\u002F     values: [0.012, -0.034, ...]\n    \u002F\u002F   }\n    \u002F\u002F }\n\n    return new Response(JSON.stringify(data), {\n      headers: { 'Content-Type': 'application\u002Fjson' }\n    });\n  }\n};\n```\n\n### Response Parsing\n\n```typescript\ninterface EmbeddingResponse {\n  embedding: {\n    values: number[];\n  };\n}\n\nconst response: EmbeddingResponse = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: 'Sample text',\n  config: { taskType: 'SEMANTIC_SIMILARITY' }\n});\n\nconst embedding: number[] = response.embedding.values;\nconst dimensions: number = embedding.length; \u002F\u002F 3072 by default\n```\n\n---\n\n## 4. Batch Embeddings\n\n### Multiple Texts in One Request (SDK)\n\nGenerate embeddings for multiple texts simultaneously:\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nconst texts = [\n  \"What is the meaning of life?\",\n  \"How does photosynthesis work?\",\n  \"Tell me about the history of the internet.\"\n];\n\nconst response = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  contents: texts, \u002F\u002F Array of strings\n  config: {\n    taskType: 'RETRIEVAL_DOCUMENT',\n    outputDimensionality: 768\n  }\n});\n\n\u002F\u002F Process each embedding\nresponse.embeddings.forEach((embedding, index) => {\n  console.log(`Text ${index}: ${texts[index]}`);\n  console.log(`Embedding: ${embedding.values.slice(0, 5)}...`);\n  console.log(`Dimensions: ${embedding.values.length}`);\n});\n```\n\n### Batch REST API (fetch)\n\nUse the `batchEmbedContents` endpoint:\n\n```typescript\nconst response = await fetch(\n  'https:\u002F\u002Fgenerativelanguage.googleapis.com\u002Fv1beta\u002Fmodels\u002Fgemini-embedding-001:batchEmbedContents',\n  {\n    method: 'POST',\n    headers: {\n      'x-goog-api-key': apiKey,\n      'Content-Type': 'application\u002Fjson'\n    },\n    body: JSON.stringify({\n      requests: texts.map(text => ({\n        model: 'models\u002Fgemini-embedding-001',\n        content: {\n          parts: [{ text }]\n        },\n        taskType: 'RETRIEVAL_DOCUMENT'\n      }))\n    })\n  }\n);\n\nconst data = await response.json();\n\u002F\u002F data.embeddings: Array of {values: number[]}\n```\n\n### Chunking for Rate Limits\n\nWhen processing large datasets, chunk requests to stay within rate limits:\n\n```typescript\nasync function batchEmbedWithRateLimit(\n  texts: string[],\n  batchSize: number = 100, \u002F\u002F Free tier: 100 RPM\n  delayMs: number = 60000 \u002F\u002F 1 minute delay between batches\n): Promise\u003Cnumber[][]> {\n  const allEmbeddings: number[][] = [];\n\n  for (let i = 0; i \u003C texts.length; i += batchSize) {\n    const batch = texts.slice(i, i + batchSize);\n\n    console.log(`Processing batch ${i \u002F batchSize + 1} (${batch.length} texts)`);\n\n    const response = await ai.models.embedContent({\n      model: 'gemini-embedding-001',\n      contents: batch,\n      config: {\n        taskType: 'RETRIEVAL_DOCUMENT',\n        outputDimensionality: 768\n      }\n    });\n\n    allEmbeddings.push(...response.embeddings.map(e => e.values));\n\n    \u002F\u002F Wait before next batch (except last batch)\n    if (i + batchSize \u003C texts.length) {\n      await new Promise(resolve => setTimeout(resolve, delayMs));\n    }\n  }\n\n  return allEmbeddings;\n}\n\n\u002F\u002F Usage\nconst embeddings = await batchEmbedWithRateLimit(documents, 100);\n```\n\n### Performance Optimization\n\n**Tips**:\n1. Use batch API when embedding multiple texts (single request vs multiple requests)\n2. Choose lower dimensions (768) for faster processing and less storage\n3. Implement exponential backoff for rate limit errors\n4. Cache embeddings to avoid redundant API calls\n\n---\n\n## 5. Task Types\n\nThe `taskType` parameter optimizes embeddings for specific use cases. **Always specify a task type for best results.**\n\n### Available Task Types (8 total)\n\n| Task Type | Use Case | Example |\n|-----------|----------|---------|\n| **RETRIEVAL_QUERY** | User search queries | \"How do I fix a flat tire?\" |\n| **RETRIEVAL_DOCUMENT** | Documents to be indexed\u002Fsearched | Product descriptions, articles |\n| **SEMANTIC_SIMILARITY** | Comparing text similarity | Duplicate detection, clustering |\n| **CLASSIFICATION** | Categorizing texts | Spam detection, sentiment analysis |\n| **CLUSTERING** | Grouping similar texts | Topic modeling, content organization |\n| **CODE_RETRIEVAL_QUERY** | Code search queries | \"function to sort array\" |\n| **QUESTION_ANSWERING** | Questions seeking answers | FAQ matching |\n| **FACT_VERIFICATION** | Verifying claims with evidence | Fact-checking systems |\n\n### When to Use Which\n\n**RAG Systems** (Retrieval Augmented Generation):\n```typescript\n\u002F\u002F When embedding user queries\nconst queryEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery,\n  config: { taskType: 'RETRIEVAL_QUERY' } \u002F\u002F ← Use RETRIEVAL_QUERY\n});\n\n\u002F\u002F When embedding documents for indexing\nconst docEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: documentText,\n  config: { taskType: 'RETRIEVAL_DOCUMENT' } \u002F\u002F ← Use RETRIEVAL_DOCUMENT\n});\n```\n\n**Semantic Search**:\n```typescript\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'SEMANTIC_SIMILARITY' }\n});\n```\n\n**Document Clustering**:\n```typescript\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'CLUSTERING' }\n});\n```\n\n### Impact on Quality\n\nUsing the correct task type **significantly improves** retrieval quality:\n\n```typescript\n\u002F\u002F ❌ BAD: No task type specified\nconst embedding1 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery\n});\n\n\u002F\u002F ✅ GOOD: Task type specified\nconst embedding2 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: userQuery,\n  config: { taskType: 'RETRIEVAL_QUERY' }\n});\n```\n\n**Result**: Using the right task type can improve search relevance by 10-30%.\n\n---\n\n## 6. RAG Patterns\n\n**RAG** (Retrieval Augmented Generation) combines vector search with LLM generation to create AI systems that answer questions using custom knowledge bases.\n\n### Complete RAG Workflow\n\n```\n1. Document Ingestion\n   ├── Chunk documents into smaller pieces\n   ├── Generate embeddings (RETRIEVAL_DOCUMENT)\n   └── Store in Vectorize\n\n2. Query Processing\n   ├── User submits query\n   ├── Generate query embedding (RETRIEVAL_QUERY)\n   └── Search Vectorize for similar documents\n\n3. Response Generation\n   ├── Retrieve top-k similar documents\n   ├── Pass documents as context to LLM\n   └── Stream response to user\n```\n\n### Document Ingestion Pipeline\n\n```typescript\nimport { GoogleGenAI } from \"@google\u002Fgenai\";\n\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\n\u002F\u002F 1. Chunk document into smaller pieces\nfunction chunkDocument(text: string, chunkSize: number = 500): string[] {\n  const words = text.split(' ');\n  const chunks: string[] = [];\n\n  for (let i = 0; i \u003C words.length; i += chunkSize) {\n    chunks.push(words.slice(i, i + chunkSize).join(' '));\n  }\n\n  return chunks;\n}\n\n\u002F\u002F 2. Generate embeddings for chunks\nasync function embedChunks(chunks: string[]): Promise\u003Cnumber[][]> {\n  const response = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    contents: chunks,\n    config: {\n      taskType: 'RETRIEVAL_DOCUMENT', \u002F\u002F ← Documents for indexing\n      outputDimensionality: 768 \u002F\u002F ← Match Vectorize index dimensions\n    }\n  });\n\n  return response.embeddings.map(e => e.values);\n}\n\n\u002F\u002F 3. Store in Cloudflare Vectorize\nasync function storeInVectorize(\n  env: Env,\n  chunks: string[],\n  embeddings: number[][]\n) {\n  const vectors = chunks.map((chunk, i) => ({\n    id: `doc-${Date.now()}-${i}`,\n    values: embeddings[i],\n    metadata: { text: chunk }\n  }));\n\n  await env.VECTORIZE.insert(vectors);\n}\n\n\u002F\u002F Complete pipeline\nasync function ingestDocument(env: Env, documentText: string) {\n  const chunks = chunkDocument(documentText, 500);\n  const embeddings = await embedChunks(chunks);\n  await storeInVectorize(env, chunks, embeddings);\n\n  console.log(`Ingested ${chunks.length} chunks`);\n}\n```\n\n### Query Flow (Retrieve + Generate)\n\n```typescript\nasync function ragQuery(env: Env, userQuery: string): Promise\u003Cstring> {\n  \u002F\u002F 1. Embed user query\n  const queryResponse = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    content: userQuery,\n    config: {\n      taskType: 'RETRIEVAL_QUERY', \u002F\u002F ← Query, not document\n      outputDimensionality: 768\n    }\n  });\n\n  const queryEmbedding = queryResponse.embedding.values;\n\n  \u002F\u002F 2. Search Vectorize for similar documents\n  const results = await env.VECTORIZE.query(queryEmbedding, {\n    topK: 5,\n    returnMetadata: true\n  });\n\n  \u002F\u002F 3. Extract context from top results\n  const context = results.matches\n    .map(match => match.metadata.text)\n    .join('\\n\\n');\n\n  \u002F\u002F 4. Generate response with context\n  const response = await ai.models.generateContent({\n    model: 'gemini-2.5-flash',\n    contents: `Context:\\n${context}\\n\\nQuestion: ${userQuery}\\n\\nAnswer based on the context above:`\n  });\n\n  return response.text;\n}\n```\n\n### Integration with Cloudflare Vectorize\n\n**Create Vectorize Index** (768 dimensions for Gemini):\n\n```bash\nnpx wrangler vectorize create gemini-embeddings --dimensions 768 --metric cosine\n```\n\n**Bind in wrangler.jsonc**:\n\n```jsonc\n{\n  \"name\": \"my-rag-app\",\n  \"main\": \"src\u002Findex.ts\",\n  \"compatibility_date\": \"2025-10-25\",\n  \"vectorize\": {\n    \"bindings\": [\n      {\n        \"binding\": \"VECTORIZE\",\n        \"index_name\": \"gemini-embeddings\"\n      }\n    ]\n  }\n}\n```\n\n**Complete RAG Worker**:\n\nSee `templates\u002Frag-with-vectorize.ts` for full implementation.\n\n---\n\n## 7. Semantic Search\n\nSemantic search finds content based on **meaning**, not just keyword matching.\n\n### Cosine Similarity\n\nCosine similarity measures how similar two embeddings are (range: -1 to 1, where 1 = identical):\n\n```typescript\nfunction cosineSimilarity(a: number[], b: number[]): number {\n  if (a.length !== b.length) {\n    throw new Error('Vectors must have same length');\n  }\n\n  let dotProduct = 0;\n  let magnitudeA = 0;\n  let magnitudeB = 0;\n\n  for (let i = 0; i \u003C a.length; i++) {\n    dotProduct += a[i] * b[i];\n    magnitudeA += a[i] * a[i];\n    magnitudeB += b[i] * b[i];\n  }\n\n  if (magnitudeA === 0 || magnitudeB === 0) {\n    return 0;\n  }\n\n  return dotProduct \u002F (Math.sqrt(magnitudeA) * Math.sqrt(magnitudeB));\n}\n```\n\n### Vector Normalization\n\nNormalize vectors to unit length for faster similarity calculations:\n\n```typescript\nfunction normalizeVector(vector: number[]): number[] {\n  const magnitude = Math.sqrt(vector.reduce((sum, val) => sum + val * val, 0));\n\n  if (magnitude === 0) {\n    return vector;\n  }\n\n  return vector.map(val => val \u002F magnitude);\n}\n\n\u002F\u002F Normalized vectors allow dot product instead of cosine similarity\nfunction dotProduct(a: number[], b: number[]): number {\n  return a.reduce((sum, val, i) => sum + val * b[i], 0);\n}\n```\n\n### Top-K Search\n\nFind the most similar documents to a query:\n\n```typescript\ninterface Document {\n  id: string;\n  text: string;\n  embedding: number[];\n}\n\nfunction topKSimilar(\n  queryEmbedding: number[],\n  documents: Document[],\n  k: number = 5\n): Array\u003C{ document: Document; similarity: number }> {\n  \u002F\u002F Calculate similarity for each document\n  const similarities = documents.map(doc => ({\n    document: doc,\n    similarity: cosineSimilarity(queryEmbedding, doc.embedding)\n  }));\n\n  \u002F\u002F Sort by similarity (descending) and return top K\n  return similarities\n    .sort((a, b) => b.similarity - a.similarity)\n    .slice(0, k);\n}\n\n\u002F\u002F Usage\nconst results = topKSimilar(queryEmbedding, documents, 5);\n\nresults.forEach(result => {\n  console.log(`Similarity: ${result.similarity.toFixed(4)}`);\n  console.log(`Text: ${result.document.text}\\n`);\n});\n```\n\n### Complete Semantic Search Example\n\nSee `templates\u002Fsemantic-search.ts` for full implementation with Gemini API.\n\n---\n\n## 8. Document Clustering\n\nClustering groups similar documents together automatically.\n\n### K-Means Clustering\n\n```typescript\ninterface Cluster {\n  centroid: number[];\n  documents: number[][];\n}\n\nfunction kMeansClustering(\n  embeddings: number[][],\n  k: number = 3,\n  maxIterations: number = 100\n): Cluster[] {\n  \u002F\u002F 1. Initialize centroids randomly\n  const centroids: number[][] = [];\n  for (let i = 0; i \u003C k; i++) {\n    centroids.push(embeddings[Math.floor(Math.random() * embeddings.length)]);\n  }\n\n  \u002F\u002F 2. Iterate until convergence\n  for (let iter = 0; iter \u003C maxIterations; iter++) {\n    \u002F\u002F Assign each embedding to nearest centroid\n    const clusters: number[][][] = Array(k).fill(null).map(() => []);\n\n    embeddings.forEach(embedding => {\n      let minDistance = Infinity;\n      let closestCluster = 0;\n\n      centroids.forEach((centroid, i) => {\n        const distance = 1 - cosineSimilarity(embedding, centroid);\n        if (distance \u003C minDistance) {\n          minDistance = distance;\n          closestCluster = i;\n        }\n      });\n\n      clusters[closestCluster].push(embedding);\n    });\n\n    \u002F\u002F Update centroids\n    let changed = false;\n    clusters.forEach((cluster, i) => {\n      if (cluster.length > 0) {\n        const newCentroid = cluster[0].map((_, dim) =>\n          cluster.reduce((sum, emb) => sum + emb[dim], 0) \u002F cluster.length\n        );\n\n        if (cosineSimilarity(centroids[i], newCentroid) \u003C 0.9999) {\n          changed = true;\n        }\n\n        centroids[i] = newCentroid;\n      }\n    });\n\n    if (!changed) break;\n  }\n\n  \u002F\u002F Build final clusters\n  const finalClusters: Cluster[] = centroids.map((centroid, i) => ({\n    centroid,\n    documents: embeddings.filter(emb =>\n      cosineSimilarity(emb, centroid) === Math.max(\n        ...centroids.map(c => cosineSimilarity(emb, c))\n      )\n    )\n  }));\n\n  return finalClusters;\n}\n```\n\nSee `templates\u002Fclustering.ts` for complete implementation with examples.\n\n### Use Cases\n\n1. **Content Organization**: Automatically group similar articles, products, or documents\n2. **Duplicate Detection**: Find near-duplicate content\n3. **Topic Modeling**: Discover themes in large text corpora\n4. **Customer Feedback Analysis**: Group similar customer complaints or feature requests\n\n---\n\n## 9. Error Handling\n\n### Common Errors\n\n**1. API Key Missing or Invalid**\n\n```typescript\n\u002F\u002F ❌ Error: API key not set\nconst ai = new GoogleGenAI({});\n\n\u002F\u002F ✅ Correct\nconst ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });\n\nif (!process.env.GEMINI_API_KEY) {\n  throw new Error('GEMINI_API_KEY environment variable not set');\n}\n```\n\n**2. Dimension Mismatch**\n\n```typescript\n\u002F\u002F ❌ Error: Embedding has 3072 dims, Vectorize expects 768\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text\n  \u002F\u002F No outputDimensionality specified → defaults to 3072\n});\n\nawait env.VECTORIZE.insert([{\n  id: '1',\n  values: embedding.embedding.values \u002F\u002F 3072 dims, but index is 768!\n}]);\n\n\u002F\u002F ✅ Correct: Match dimensions\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { outputDimensionality: 768 } \u002F\u002F ← Match index dimensions\n});\n```\n\n**3. Rate Limiting**\n\n```typescript\n\u002F\u002F ❌ Error: 429 Too Many Requests\nfor (let i = 0; i \u003C 1000; i++) {\n  await ai.models.embedContent({ \u002F* ... *\u002F }); \u002F\u002F Exceeds 100 RPM on free tier\n}\n\n\u002F\u002F ✅ Correct: Implement rate limiting\nasync function embedWithRetry(text: string, maxRetries = 3) {\n  for (let attempt = 0; attempt \u003C maxRetries; attempt++) {\n    try {\n      return await ai.models.embedContent({\n        model: 'gemini-embedding-001',\n        content: text,\n        config: { taskType: 'SEMANTIC_SIMILARITY' }\n      });\n    } catch (error: any) {\n      if (error.status === 429 && attempt \u003C maxRetries - 1) {\n        const delay = Math.pow(2, attempt) * 1000; \u002F\u002F Exponential backoff\n        await new Promise(resolve => setTimeout(resolve, delay));\n        continue;\n      }\n      throw error;\n    }\n  }\n}\n```\n\nSee `references\u002Ftop-errors.md` for all 8 documented errors with detailed solutions.\n\n---\n\n## 10. Best Practices\n\n### Always Do\n\n✅ **Specify Task Type**\n```typescript\n\u002F\u002F Task type optimizes embeddings for your use case\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { taskType: 'RETRIEVAL_QUERY' } \u002F\u002F ← Always specify\n});\n```\n\n✅ **Match Dimensions with Vectorize**\n```typescript\n\u002F\u002F Ensure embeddings match your Vectorize index dimensions\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text,\n  config: { outputDimensionality: 768 } \u002F\u002F ← Match index\n});\n```\n\n✅ **Implement Rate Limiting**\n```typescript\n\u002F\u002F Use exponential backoff for 429 errors\nasync function embedWithBackoff(text: string) {\n  \u002F\u002F Implementation from Error Handling section\n}\n```\n\n✅ **Cache Embeddings**\n```typescript\n\u002F\u002F Cache embeddings to avoid redundant API calls\nconst cache = new Map\u003Cstring, number[]>();\n\nasync function getCachedEmbedding(text: string): Promise\u003Cnumber[]> {\n  if (cache.has(text)) {\n    return cache.get(text)!;\n  }\n\n  const response = await ai.models.embedContent({\n    model: 'gemini-embedding-001',\n    content: text,\n    config: { taskType: 'SEMANTIC_SIMILARITY' }\n  });\n\n  const embedding = response.embedding.values;\n  cache.set(text, embedding);\n  return embedding;\n}\n```\n\n✅ **Use Batch API for Multiple Texts**\n```typescript\n\u002F\u002F Single batch request vs multiple individual requests\nconst embeddings = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  contents: texts, \u002F\u002F Array of texts\n  config: { taskType: 'RETRIEVAL_DOCUMENT' }\n});\n```\n\n### Never Do\n\n❌ **Don't Skip Task Type**\n```typescript\n\u002F\u002F Reduces quality by 10-30%\nconst embedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text\n  \u002F\u002F Missing taskType!\n});\n```\n\n❌ **Don't Mix Different Dimensions**\n```typescript\n\u002F\u002F Can't compare embeddings with different dimensions\nconst emb1 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text1,\n  config: { outputDimensionality: 768 }\n});\n\nconst emb2 = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: text2,\n  config: { outputDimensionality: 1536 } \u002F\u002F Different dimensions!\n});\n\n\u002F\u002F ❌ Can't calculate similarity between different dimensions\nconst similarity = cosineSimilarity(emb1.embedding.values, emb2.embedding.values);\n```\n\n❌ **Don't Use Wrong Task Type for RAG**\n```typescript\n\u002F\u002F Reduces search quality\nconst queryEmbedding = await ai.models.embedContent({\n  model: 'gemini-embedding-001',\n  content: query,\n  config: { taskType: 'RETRIEVAL_DOCUMENT' } \u002F\u002F Wrong! Should be RETRIEVAL_QUERY\n});\n```\n\n---\n\n## Using Bundled Resources\n\n### Templates (templates\u002F)\n\n- `package.json` - Package configuration with verified versions\n- `basic-embeddings.ts` - Single text embedding with SDK\n- `embeddings-fetch.ts` - Fetch-based for Cloudflare Workers\n- `batch-embeddings.ts` - Batch processing with rate limiting\n- `rag-with-vectorize.ts` - Complete RAG implementation with Vectorize\n- `semantic-search.ts` - Cosine similarity and top-K search\n- `clustering.ts` - K-means clustering implementation\n\n### References (references\u002F)\n\n- `model-comparison.md` - Compare Gemini vs OpenAI vs Workers AI embeddings\n- `vectorize-integration.md` - Cloudflare Vectorize setup and patterns\n- `rag-patterns.md` - Complete RAG implementation strategies\n- `dimension-guide.md` - Choosing the right dimensions (768 vs 1536 vs 3072)\n- `top-errors.md` - 8 common errors and detailed solutions\n\n### Scripts (scripts\u002F)\n\n- `check-versions.sh` - Verify @google\u002Fgenai package version is current\n\n---\n\n## Official Documentation\n\n- **Embeddings Guide**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Fembeddings\n- **Model Spec**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Fmodels\u002Fgemini#gemini-embedding-001\n- **Rate Limits**: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Frate-limits\n- **SDK Reference**: https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@google\u002Fgenai\n- **Context7 Library ID**: `\u002Fwebsites\u002Fai_google_dev_gemini-api`\n\n---\n\n## Related Skills\n\n- **google-gemini-api** - Main Gemini API for text\u002Fimage generation\n- **cloudflare-vectorize** - Vector database for storing embeddings\n- **cloudflare-workers-ai** - Workers AI embeddings (BGE models)\n\n---\n\n## Success Metrics\n\n**Token Savings**: ~60% compared to manual implementation\n**Errors Prevented**: 8 documented errors with solutions\n**Production Tested**: ✅ Verified in RAG applications\n**Package Version**: @google\u002Fgenai@1.27.0\n**Last Updated**: 2025-10-25\n\n---\n\n## License\n\nMIT License - Free to use in personal and commercial projects.\n\n---\n\n**Questions or Issues?**\n\n- GitHub: https:\u002F\u002Fgithub.com\u002Fjezweb\u002Fclaude-skills\n- Email: jeremy@jezweb.net\n",{"name":5,"license":14,"metadata":23,"description":33},{"version":24,"complexity":25,"last_updated":26,"tokens_saved":27,"target_audience":28,"errors_prevented":29,"production_tested":30,"estimated_reading_time":31,"tested_package_version":32},"1.0.0","intermediate",1761350400000,"~60%","Developers building RAG, semantic search, or vector-based applications",8,true,"15 minutes","@google\u002Fgenai@1.27.0","This skill provides complete coverage of Google Gemini embeddings API (gemini-embedding-001) for building RAG systems, semantic search, document clustering, and similarity matching. Use when implementing vector search with Google's embedding models, integrating with Cloudflare Vectorize, or building retrieval-augmented generation systems. Covers SDK usage (@google\u002Fgenai), fetch-based Workers implementation, batch processing, 8 task types (RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, SEMANTIC_SIMILARITY, etc.), dimension optimization (128-3072), and cosine similarity calculations. Prevents 8+ embedding-specific errors including dimension mismatches, incorrect task types, rate limiting issues (100 RPM free tier), vector normalization mistakes, text truncation (2,048 token limit), and model version confusion. Includes production-ready RAG patterns with Cloudflare Vectorize integration, chunking strategies, and caching patterns. Token savings: ~60%. Production tested.\n\nKeywords: gemini embeddings, gemini-embedding-001, google embeddings, semantic search, RAG, vector search, document clustering, similarity search, retrieval augmented generation, vectorize integration, cloudflare vectorize embeddings, 768 dimensions, embed content gemini, batch embeddings, embeddings api, cosine similarity, vector normalization, retrieval query, retrieval document, task types, dimension mismatch, embeddings rate limit, text truncation, @google\u002Fgenai\n","main","6b73b29e4866f997b1eb47fe8d9631396265fc24",0,"https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn","2026-09-09 16:06:07",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":30},[42,45,48],{"tool":43,"command":44},"npx","npx skills add davidanthonyn\u002Fagent-skills --skill google-gemini-embeddings",{"tool":46,"command":47},"bunx","bunx skills add davidanthonyn\u002Fagent-skills --skill google-gemini-embeddings",{"tool":49,"command":50},"pnpm","pnpm dlx skills add davidanthonyn\u002Fagent-skills --skill google-gemini-embeddings","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fgoogle-gemini-embeddings\u002FSKILL.md",[53,59,63],{"path":54,"resourceType":55,"resourceTypeDesc":56,"sizeBytes":57,"sourceHtmlUrl":58},"README.md","REFERENCE","参考文档",8056,"https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fgoogle-gemini-embeddings\u002FREADME.md",{"path":60,"resourceType":55,"resourceTypeDesc":56,"sizeBytes":61,"sourceHtmlUrl":62},"SKILL.json",28542,"https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fgoogle-gemini-embeddings\u002FSKILL.json",{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"SKILL.md.backup","OTHER","其他",27122,"https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fgoogle-gemini-embeddings\u002FSKILL.md.backup",[70,77,85,92,99,108],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":18},"01M7AQNJDXVNVMN08N5BHBGZBF","zustand-state-management","Production-tested setup for Zustand state management in React applications with TypeScript.\nThis skill provides comprehensive patterns for building scalable, type-safe global state.\n\nUse when: setting up global state in React, migrating from Redux or Context API, implementing\nstate persistence with localStorage, configuring TypeScript with Zustand, using slices pattern\nfor modular stores, adding devtools middleware for debugging, handling Next.js SSR hydration,\nor encountering hydration errors, TypeScript inference issues, or persist middleware problems.\n\nPrevents 5 documented issues: Next.js hydration mismatches, TypeScript double parentheses\nsyntax errors, persist middleware export errors, infinite render loops, and slices pattern\ntype inference failures.\n\nKeywords: zustand, state management, React state, TypeScript state, persist middleware,\ndevtools, slices pattern, global state, React hooks, create store, useBoundStore,\nStateCreator, hydration error, text content mismatch, infinite render, localStorage,\nsessionStorage, immer middleware, shallow equality, selector pattern, zustand v5","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fzustand-state-management\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":81,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M7AQN2DW9DTPM8BYNZSHQANC","zetic-mlange","Help users integrate the ZETIC Melange (MLange) on-device AI SDK into Android (Kotlin\u002FJava) and iOS (Swift) mobile apps. Use this skill whenever the user mentions Zetic, MLange, Melange, ZeticMLangeModel, ZeticMLangeLLMModel, ZeticMLangeHFModel, on-device NPU inference for mobile, or deploying ML models to Android\u002FiOS with NPU acceleration. Also trigger when users ask about running LLMs on-device with GGUF quantization, mobile NPU model deployment, or converting PyTorch\u002FONNX models for mobile inference. Covers general model inference, LLM inference (streaming token generation), and HuggingFace model loading.","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fzetic-mlange\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fzetic-mlange\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":18},"01M7AQMEK370FBPC75N2W8WA4W","youtube-downloader","Download videos, audio, playlists, and channels from YouTube and 1000+ websites using yt-dlp. Supports quality selection, format conversion, subtitle download, playlist filtering, metadata extraction, thumbnail download, and batch operations. Use when downloading YouTube videos in any quality (4K, 8K, HDR), extracting audio as MP3\u002FM4A\u002FFLAC, downloading entire playlists\u002Fchannels, getting subtitles in multiple languages, converting to specific formats, downloading live streams, archiving content, or batch processing multiple URLs. Optimized for reliability with automatic retries, rate limiting, and error handling.","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fyoutube-downloader\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":18},"01M7AQHKQ4VQB04NH6BG4SV9Q7","wordpress-plugin-core","This skill provides comprehensive knowledge for WordPress plugin development, covering core patterns, security best practices, database interactions, hooks\u002Ffilters, Settings API, custom post types, REST API, and AJAX. This skill should be used when creating WordPress plugins, troubleshooting security issues, implementing custom post types\u002Ftaxonomies, building admin interfaces, or working with the WordPress database.\n\nUse when: Creating new WordPress plugins, implementing nonces\u002Fsanitization\u002Fescaping, working with $wpdb and prepared statements, building Settings API pages, registering custom post types or taxonomies, implementing REST API endpoints, handling AJAX requests, debugging plugin activation\u002Fdeactivation issues, preventing SQL injection\u002FXSS\u002FCSRF vulnerabilities.\n\nKeywords: wordpress plugin development, wordpress security, wordpress hooks, wordpress filters, wordpress database, wpdb prepare, sanitize_text_field, esc_html, wp_nonce, custom post type, register_post_type, settings api, rest api, admin-ajax, wordpress sql injection, wordpress xss, wordpress csrf, plugin header, activation hook, deactivation hook, wordpress coding standards, wordpress plugin architecture","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fwordpress-plugin-core\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":103,"licenseFileUrl":104,"sourceHtmlUrl":105,"categories":106,"keywords":107,"publishedAt":18},"01M7AQH4JW8F2MJFS1C43TZ2TY","webapp-testing","Test local web applications with Playwright. Use when asked to verify frontend functionality, debug UI behavior, capture browser screenshots, or inspect browser logs.","Apache-2.0","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fwebapp-testing\u002FSKILL.md",[],[],{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":18},"01M7AQH0PWD6Z1NN66V15DT25M","web-performance-audit","Audit web performance using Chrome DevTools MCP. Use when asked to audit, profile, debug, or optimize page load performance, Lighthouse scores, or site speed.","https:\u002F\u002Fgithub.com\u002Fdavidanthonyn\u002Fagent-skills\u002Fblob\u002Fmain\u002Fweb-performance-audit\u002FSKILL.md",[],[]]