[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7D355B6JCV360TCM2T4G44S":3,"skill-same:01M7D355B6JCV360TCM2T4G44S":64},{"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":38,"commitSha":39,"skillDirectory":40,"forksCount":13,"repositoryHtmlUrl":41,"repositoryHomepageUrl":6,"ownerHtmlUrl":42,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":43,"install":44,"resources":57},"01M7D355B6JCV360TCM2T4G44S","pinecone",null,"Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + sparse), metadata filtering, and namespaces. Low latency (\u003C100ms p95). Use for production RAG, recommendation systems, or semantic search at scale. Best for serverless, managed infrastructure.","beyondai\u002Fsandbox","beyondai","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F2845958?v=4","01M7D24SBNM47BJGJ9J2HB13BP","01M7D24SCDC9DYE005XH6R553S",0,"MIT","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.cursor\u002Fskills\u002Fpinecone\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: pinecone\ndescription: Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + sparse), metadata filtering, and namespaces. Low latency (\u003C100ms p95). Use for production RAG, recommendation systems, or semantic search at scale. Best for serverless, managed infrastructure.\nversion: 1.0.0\nauthor: Orchestra Research\nlicense: MIT\ntags: [RAG, Pinecone, Vector Database, Managed Service, Serverless, Hybrid Search, Production, Auto-Scaling, Low Latency, Recommendations]\ndependencies: [pinecone-client]\n---\n\n# Pinecone - Managed Vector Database\n\nThe vector database for production AI applications.\n\n## When to use Pinecone\n\n**Use when:**\n- Need managed, serverless vector database\n- Production RAG applications\n- Auto-scaling required\n- Low latency critical (\u003C100ms)\n- Don't want to manage infrastructure\n- Need hybrid search (dense + sparse vectors)\n\n**Metrics**:\n- Fully managed SaaS\n- Auto-scales to billions of vectors\n- **p95 latency \u003C100ms**\n- 99.9% uptime SLA\n\n**Use alternatives instead**:\n- **Chroma**: Self-hosted, open-source\n- **FAISS**: Offline, pure similarity search\n- **Weaviate**: Self-hosted with more features\n\n## Quick start\n\n### Installation\n\n```bash\npip install pinecone-client\n```\n\n### Basic usage\n\n```python\nfrom pinecone import Pinecone, ServerlessSpec\n\n# Initialize\npc = Pinecone(api_key=\"your-api-key\")\n\n# Create index\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,  # Must match embedding dimension\n    metric=\"cosine\",  # or \"euclidean\", \"dotproduct\"\n    spec=ServerlessSpec(cloud=\"aws\", region=\"us-east-1\")\n)\n\n# Connect to index\nindex = pc.Index(\"my-index\")\n\n# Upsert vectors\nindex.upsert(vectors=[\n    {\"id\": \"vec1\", \"values\": [0.1, 0.2, ...], \"metadata\": {\"category\": \"A\"}},\n    {\"id\": \"vec2\", \"values\": [0.3, 0.4, ...], \"metadata\": {\"category\": \"B\"}}\n])\n\n# Query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    include_metadata=True\n)\n\nprint(results[\"matches\"])\n```\n\n## Core operations\n\n### Create index\n\n```python\n# Serverless (recommended)\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,\n    metric=\"cosine\",\n    spec=ServerlessSpec(\n        cloud=\"aws\",         # or \"gcp\", \"azure\"\n        region=\"us-east-1\"\n    )\n)\n\n# Pod-based (for consistent performance)\nfrom pinecone import PodSpec\n\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,\n    metric=\"cosine\",\n    spec=PodSpec(\n        environment=\"us-east1-gcp\",\n        pod_type=\"p1.x1\"\n    )\n)\n```\n\n### Upsert vectors\n\n```python\n# Single upsert\nindex.upsert(vectors=[\n    {\n        \"id\": \"doc1\",\n        \"values\": [0.1, 0.2, ...],  # 1536 dimensions\n        \"metadata\": {\n            \"text\": \"Document content\",\n            \"category\": \"tutorial\",\n            \"timestamp\": \"2025-01-01\"\n        }\n    }\n])\n\n# Batch upsert (recommended)\nvectors = [\n    {\"id\": f\"vec{i}\", \"values\": embedding, \"metadata\": metadata}\n    for i, (embedding, metadata) in enumerate(zip(embeddings, metadatas))\n]\n\nindex.upsert(vectors=vectors, batch_size=100)\n```\n\n### Query vectors\n\n```python\n# Basic query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=10,\n    include_metadata=True,\n    include_values=False\n)\n\n# With metadata filtering\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    filter={\"category\": {\"$eq\": \"tutorial\"}}\n)\n\n# Namespace query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    namespace=\"production\"\n)\n\n# Access results\nfor match in results[\"matches\"]:\n    print(f\"ID: {match['id']}\")\n    print(f\"Score: {match['score']}\")\n    print(f\"Metadata: {match['metadata']}\")\n```\n\n### Metadata filtering\n\n```python\n# Exact match\nfilter = {\"category\": \"tutorial\"}\n\n# Comparison\nfilter = {\"price\": {\"$gte\": 100}}  # $gt, $gte, $lt, $lte, $ne\n\n# Logical operators\nfilter = {\n    \"$and\": [\n        {\"category\": \"tutorial\"},\n        {\"difficulty\": {\"$lte\": 3}}\n    ]\n}  # Also: $or\n\n# In operator\nfilter = {\"tags\": {\"$in\": [\"python\", \"ml\"]}}\n```\n\n## Namespaces\n\n```python\n# Partition data by namespace\nindex.upsert(\n    vectors=[{\"id\": \"vec1\", \"values\": [...]}],\n    namespace=\"user-123\"\n)\n\n# Query specific namespace\nresults = index.query(\n    vector=[...],\n    namespace=\"user-123\",\n    top_k=5\n)\n\n# List namespaces\nstats = index.describe_index_stats()\nprint(stats['namespaces'])\n```\n\n## Hybrid search (dense + sparse)\n\n```python\n# Upsert with sparse vectors\nindex.upsert(vectors=[\n    {\n        \"id\": \"doc1\",\n        \"values\": [0.1, 0.2, ...],  # Dense vector\n        \"sparse_values\": {\n            \"indices\": [10, 45, 123],  # Token IDs\n            \"values\": [0.5, 0.3, 0.8]   # TF-IDF scores\n        },\n        \"metadata\": {\"text\": \"...\"}\n    }\n])\n\n# Hybrid query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    sparse_vector={\n        \"indices\": [10, 45],\n        \"values\": [0.5, 0.3]\n    },\n    top_k=5,\n    alpha=0.5  # 0=sparse, 1=dense, 0.5=hybrid\n)\n```\n\n## LangChain integration\n\n```python\nfrom langchain_pinecone import PineconeVectorStore\nfrom langchain_openai import OpenAIEmbeddings\n\n# Create vector store\nvectorstore = PineconeVectorStore.from_documents(\n    documents=docs,\n    embedding=OpenAIEmbeddings(),\n    index_name=\"my-index\"\n)\n\n# Query\nresults = vectorstore.similarity_search(\"query\", k=5)\n\n# With metadata filter\nresults = vectorstore.similarity_search(\n    \"query\",\n    k=5,\n    filter={\"category\": \"tutorial\"}\n)\n\n# As retriever\nretriever = vectorstore.as_retriever(search_kwargs={\"k\": 10})\n```\n\n## LlamaIndex integration\n\n```python\nfrom llama_index.vector_stores.pinecone import PineconeVectorStore\n\n# Connect to Pinecone\npc = Pinecone(api_key=\"your-key\")\npinecone_index = pc.Index(\"my-index\")\n\n# Create vector store\nvector_store = PineconeVectorStore(pinecone_index=pinecone_index)\n\n# Use in LlamaIndex\nfrom llama_index.core import StorageContext, VectorStoreIndex\n\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n## Index management\n\n```python\n# List indices\nindexes = pc.list_indexes()\n\n# Describe index\nindex_info = pc.describe_index(\"my-index\")\nprint(index_info)\n\n# Get index stats\nstats = index.describe_index_stats()\nprint(f\"Total vectors: {stats['total_vector_count']}\")\nprint(f\"Namespaces: {stats['namespaces']}\")\n\n# Delete index\npc.delete_index(\"my-index\")\n```\n\n## Delete vectors\n\n```python\n# Delete by ID\nindex.delete(ids=[\"vec1\", \"vec2\"])\n\n# Delete by filter\nindex.delete(filter={\"category\": \"old\"})\n\n# Delete all in namespace\nindex.delete(delete_all=True, namespace=\"test\")\n\n# Delete entire index\nindex.delete(delete_all=True)\n```\n\n## Best practices\n\n1. **Use serverless** - Auto-scaling, cost-effective\n2. **Batch upserts** - More efficient (100-200 per batch)\n3. **Add metadata** - Enable filtering\n4. **Use namespaces** - Isolate data by user\u002Ftenant\n5. **Monitor usage** - Check Pinecone dashboard\n6. **Optimize filters** - Index frequently filtered fields\n7. **Test with free tier** - 1 index, 100K vectors free\n8. **Use hybrid search** - Better quality\n9. **Set appropriate dimensions** - Match embedding model\n10. **Regular backups** - Export important data\n\n## Performance\n\n| Operation | Latency | Notes |\n|-----------|---------|-------|\n| Upsert | ~50-100ms | Per batch |\n| Query (p50) | ~50ms | Depends on index size |\n| Query (p95) | ~100ms | SLA target |\n| Metadata filter | ~+10-20ms | Additional overhead |\n\n## Pricing (as of 2025)\n\n**Serverless**:\n- $0.096 per million read units\n- $0.06 per million write units\n- $0.06 per GB storage\u002Fmonth\n\n**Free tier**:\n- 1 serverless index\n- 100K vectors (1536 dimensions)\n- Great for prototyping\n\n## Resources\n\n- **Website**: https:\u002F\u002Fwww.pinecone.io\n- **Docs**: https:\u002F\u002Fdocs.pinecone.io\n- **Console**: https:\u002F\u002Fapp.pinecone.io\n- **Pricing**: https:\u002F\u002Fwww.pinecone.io\u002Fpricing\n\n\n","# Pinecone - Managed Vector Database\n\nThe vector database for production AI applications.\n\n## When to use Pinecone\n\n**Use when:**\n- Need managed, serverless vector database\n- Production RAG applications\n- Auto-scaling required\n- Low latency critical (\u003C100ms)\n- Don't want to manage infrastructure\n- Need hybrid search (dense + sparse vectors)\n\n**Metrics**:\n- Fully managed SaaS\n- Auto-scales to billions of vectors\n- **p95 latency \u003C100ms**\n- 99.9% uptime SLA\n\n**Use alternatives instead**:\n- **Chroma**: Self-hosted, open-source\n- **FAISS**: Offline, pure similarity search\n- **Weaviate**: Self-hosted with more features\n\n## Quick start\n\n### Installation\n\n```bash\npip install pinecone-client\n```\n\n### Basic usage\n\n```python\nfrom pinecone import Pinecone, ServerlessSpec\n\n# Initialize\npc = Pinecone(api_key=\"your-api-key\")\n\n# Create index\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,  # Must match embedding dimension\n    metric=\"cosine\",  # or \"euclidean\", \"dotproduct\"\n    spec=ServerlessSpec(cloud=\"aws\", region=\"us-east-1\")\n)\n\n# Connect to index\nindex = pc.Index(\"my-index\")\n\n# Upsert vectors\nindex.upsert(vectors=[\n    {\"id\": \"vec1\", \"values\": [0.1, 0.2, ...], \"metadata\": {\"category\": \"A\"}},\n    {\"id\": \"vec2\", \"values\": [0.3, 0.4, ...], \"metadata\": {\"category\": \"B\"}}\n])\n\n# Query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    include_metadata=True\n)\n\nprint(results[\"matches\"])\n```\n\n## Core operations\n\n### Create index\n\n```python\n# Serverless (recommended)\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,\n    metric=\"cosine\",\n    spec=ServerlessSpec(\n        cloud=\"aws\",         # or \"gcp\", \"azure\"\n        region=\"us-east-1\"\n    )\n)\n\n# Pod-based (for consistent performance)\nfrom pinecone import PodSpec\n\npc.create_index(\n    name=\"my-index\",\n    dimension=1536,\n    metric=\"cosine\",\n    spec=PodSpec(\n        environment=\"us-east1-gcp\",\n        pod_type=\"p1.x1\"\n    )\n)\n```\n\n### Upsert vectors\n\n```python\n# Single upsert\nindex.upsert(vectors=[\n    {\n        \"id\": \"doc1\",\n        \"values\": [0.1, 0.2, ...],  # 1536 dimensions\n        \"metadata\": {\n            \"text\": \"Document content\",\n            \"category\": \"tutorial\",\n            \"timestamp\": \"2025-01-01\"\n        }\n    }\n])\n\n# Batch upsert (recommended)\nvectors = [\n    {\"id\": f\"vec{i}\", \"values\": embedding, \"metadata\": metadata}\n    for i, (embedding, metadata) in enumerate(zip(embeddings, metadatas))\n]\n\nindex.upsert(vectors=vectors, batch_size=100)\n```\n\n### Query vectors\n\n```python\n# Basic query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=10,\n    include_metadata=True,\n    include_values=False\n)\n\n# With metadata filtering\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    filter={\"category\": {\"$eq\": \"tutorial\"}}\n)\n\n# Namespace query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    top_k=5,\n    namespace=\"production\"\n)\n\n# Access results\nfor match in results[\"matches\"]:\n    print(f\"ID: {match['id']}\")\n    print(f\"Score: {match['score']}\")\n    print(f\"Metadata: {match['metadata']}\")\n```\n\n### Metadata filtering\n\n```python\n# Exact match\nfilter = {\"category\": \"tutorial\"}\n\n# Comparison\nfilter = {\"price\": {\"$gte\": 100}}  # $gt, $gte, $lt, $lte, $ne\n\n# Logical operators\nfilter = {\n    \"$and\": [\n        {\"category\": \"tutorial\"},\n        {\"difficulty\": {\"$lte\": 3}}\n    ]\n}  # Also: $or\n\n# In operator\nfilter = {\"tags\": {\"$in\": [\"python\", \"ml\"]}}\n```\n\n## Namespaces\n\n```python\n# Partition data by namespace\nindex.upsert(\n    vectors=[{\"id\": \"vec1\", \"values\": [...]}],\n    namespace=\"user-123\"\n)\n\n# Query specific namespace\nresults = index.query(\n    vector=[...],\n    namespace=\"user-123\",\n    top_k=5\n)\n\n# List namespaces\nstats = index.describe_index_stats()\nprint(stats['namespaces'])\n```\n\n## Hybrid search (dense + sparse)\n\n```python\n# Upsert with sparse vectors\nindex.upsert(vectors=[\n    {\n        \"id\": \"doc1\",\n        \"values\": [0.1, 0.2, ...],  # Dense vector\n        \"sparse_values\": {\n            \"indices\": [10, 45, 123],  # Token IDs\n            \"values\": [0.5, 0.3, 0.8]   # TF-IDF scores\n        },\n        \"metadata\": {\"text\": \"...\"}\n    }\n])\n\n# Hybrid query\nresults = index.query(\n    vector=[0.1, 0.2, ...],\n    sparse_vector={\n        \"indices\": [10, 45],\n        \"values\": [0.5, 0.3]\n    },\n    top_k=5,\n    alpha=0.5  # 0=sparse, 1=dense, 0.5=hybrid\n)\n```\n\n## LangChain integration\n\n```python\nfrom langchain_pinecone import PineconeVectorStore\nfrom langchain_openai import OpenAIEmbeddings\n\n# Create vector store\nvectorstore = PineconeVectorStore.from_documents(\n    documents=docs,\n    embedding=OpenAIEmbeddings(),\n    index_name=\"my-index\"\n)\n\n# Query\nresults = vectorstore.similarity_search(\"query\", k=5)\n\n# With metadata filter\nresults = vectorstore.similarity_search(\n    \"query\",\n    k=5,\n    filter={\"category\": \"tutorial\"}\n)\n\n# As retriever\nretriever = vectorstore.as_retriever(search_kwargs={\"k\": 10})\n```\n\n## LlamaIndex integration\n\n```python\nfrom llama_index.vector_stores.pinecone import PineconeVectorStore\n\n# Connect to Pinecone\npc = Pinecone(api_key=\"your-key\")\npinecone_index = pc.Index(\"my-index\")\n\n# Create vector store\nvector_store = PineconeVectorStore(pinecone_index=pinecone_index)\n\n# Use in LlamaIndex\nfrom llama_index.core import StorageContext, VectorStoreIndex\n\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n## Index management\n\n```python\n# List indices\nindexes = pc.list_indexes()\n\n# Describe index\nindex_info = pc.describe_index(\"my-index\")\nprint(index_info)\n\n# Get index stats\nstats = index.describe_index_stats()\nprint(f\"Total vectors: {stats['total_vector_count']}\")\nprint(f\"Namespaces: {stats['namespaces']}\")\n\n# Delete index\npc.delete_index(\"my-index\")\n```\n\n## Delete vectors\n\n```python\n# Delete by ID\nindex.delete(ids=[\"vec1\", \"vec2\"])\n\n# Delete by filter\nindex.delete(filter={\"category\": \"old\"})\n\n# Delete all in namespace\nindex.delete(delete_all=True, namespace=\"test\")\n\n# Delete entire index\nindex.delete(delete_all=True)\n```\n\n## Best practices\n\n1. **Use serverless** - Auto-scaling, cost-effective\n2. **Batch upserts** - More efficient (100-200 per batch)\n3. **Add metadata** - Enable filtering\n4. **Use namespaces** - Isolate data by user\u002Ftenant\n5. **Monitor usage** - Check Pinecone dashboard\n6. **Optimize filters** - Index frequently filtered fields\n7. **Test with free tier** - 1 index, 100K vectors free\n8. **Use hybrid search** - Better quality\n9. **Set appropriate dimensions** - Match embedding model\n10. **Regular backups** - Export important data\n\n## Performance\n\n| Operation | Latency | Notes |\n|-----------|---------|-------|\n| Upsert | ~50-100ms | Per batch |\n| Query (p50) | ~50ms | Depends on index size |\n| Query (p95) | ~100ms | SLA target |\n| Metadata filter | ~+10-20ms | Additional overhead |\n\n## Pricing (as of 2025)\n\n**Serverless**:\n- $0.096 per million read units\n- $0.06 per million write units\n- $0.06 per GB storage\u002Fmonth\n\n**Free tier**:\n- 1 serverless index\n- 100K vectors (1536 dimensions)\n- Great for prototyping\n\n## Resources\n\n- **Website**: https:\u002F\u002Fwww.pinecone.io\n- **Docs**: https:\u002F\u002Fdocs.pinecone.io\n- **Console**: https:\u002F\u002Fapp.pinecone.io\n- **Pricing**: https:\u002F\u002Fwww.pinecone.io\u002Fpricing\n\n\n",{"name":5,"tags":23,"author":34,"license":14,"version":35,"description":7,"dependencies":36},[24,25,26,27,28,29,30,31,32,33],"RAG","Pinecone","Vector Database","Managed Service","Serverless","Hybrid Search","Production","Auto-Scaling","Low Latency","Recommendations","Orchestra Research","1.0.0",[37],"pinecone-client","main","46e03791269608c5ac886a64c08f5d94147e8786","labs\u002F.cursor\u002Fskills\u002Fpinecone","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox","https:\u002F\u002Fgithub.com\u002Fbeyondai","2026-11-08 04:18:42",{"source":8,"skillName":5,"commands":45,"prompt":55,"zipAvailable":56},[46,49,52],{"tool":47,"command":48},"npx","npx skills add beyondai\u002Fsandbox --skill pinecone",{"tool":50,"command":51},"bunx","bunx skills add beyondai\u002Fsandbox --skill pinecone",{"tool":53,"command":54},"pnpm","pnpm dlx skills add beyondai\u002Fsandbox --skill pinecone","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.cursor\u002Fskills\u002Fpinecone\u002FSKILL.md",true,[58],{"path":59,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Fdeployment.md","REFERENCE","参考文档",3557,"https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.cursor\u002Fskills\u002Fpinecone\u002Freferences\u002Fdeployment.md",[65,72,79,86,93,100],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":18},"01M7D3GMPGJZRG8EPY404YF8TR","weights-and-biases","Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, and manage model registry with W&B - collaborative MLOps platform","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Fweights-and-biases\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M7D3GFZH440V2J532BSZMPVP","serving-llms-vllm","Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency\u002Fthroughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ\u002FAWQ\u002FFP8), and tensor parallelism.","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Fvllm\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":18},"01M7D3GAE8MQCPGF72EP9G9GKY","verl-rl-training","Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Fverl\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":18},"01M7D3G7KKNE5RGNYJNXV4F0MM","unsloth","Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA\u002FQLoRA optimization","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Funsloth\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":18},"01M7D3G2RCZQ08GS0KAW5CBWSC","fine-tuning-with-trl","Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Ftrl-fine-tuning\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":18},"01M7D3FX08BPB172CYGM597NZF","torchforge-rl-training","Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.","https:\u002F\u002Fgithub.com\u002Fbeyondai\u002Fsandbox\u002Fblob\u002Fmain\u002Flabs\u002F.opencode\u002Fskills\u002Ftorchforge\u002FSKILL.md",[],[]]