[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FCNKWT8FAPC53SXRGSEZWW":3,"skill-same:01M7FCNKWT8FAPC53SXRGSEZWW":78},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":42,"commitSha":43,"skillDirectory":44,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":6,"ownerDescription":48,"updatedAt":49,"install":50,"resources":63},"01M7FCNKWT8FAPC53SXRGSEZWW","llamaindex",null,"Data framework for building LLM applications with RAG. Specializes in document ingestion (300+ connectors), indexing, and querying. Features vector indices, query engines, agents, and multi-modal support. Use for document Q&A, chatbots, knowledge retrieval, or building RAG pipelines. Best for data-centric LLM applications.","CZX-LBJ\u002Fdeepseek-science","CZX-LBJ","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F204594689?v=4","01M7FC59WQ9PFXT04MKGJDCFFT","czx","01M7FC59XB2GW0BKWN9JTDBQS7",1,"MIT","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: llamaindex\ndescription: Data framework for building LLM applications with RAG. Specializes in document ingestion (300+ connectors), indexing, and querying. Features vector indices, query engines, agents, and multi-modal support. Use for document Q&A, chatbots, knowledge retrieval, or building RAG pipelines. Best for data-centric LLM applications.\nversion: 1.0.0\nauthor: Orchestra Research\nlicense: MIT\ntags: [Agents, LlamaIndex, RAG, Document Ingestion, Vector Indices, Query Engines, Knowledge Retrieval, Data Framework, Multimodal, Private Data, Connectors]\ndependencies: [llama-index, openai, anthropic]\n---\n\n# LlamaIndex - Data Framework for LLM Applications\n\nThe leading framework for connecting LLMs with your data.\n\n## When to use LlamaIndex\n\n**Use LlamaIndex when:**\n- Building RAG (retrieval-augmented generation) applications\n- Need document question-answering over private data\n- Ingesting data from multiple sources (300+ connectors)\n- Creating knowledge bases for LLMs\n- Building chatbots with enterprise data\n- Need structured data extraction from documents\n\n**Metrics**:\n- **45,100+ GitHub stars**\n- **23,000+ repositories** use LlamaIndex\n- **300+ data connectors** (LlamaHub)\n- **1,715+ contributors**\n- **v0.14.7** (stable)\n\n**Use alternatives instead**:\n- **LangChain**: More general-purpose, better for agents\n- **Haystack**: Production search pipelines\n- **txtai**: Lightweight semantic search\n- **Chroma**: Just need vector storage\n\n## Quick start\n\n### Installation\n\n```bash\n# Starter package (recommended)\npip install llama-index\n\n# Or minimal core + specific integrations\npip install llama-index-core\npip install llama-index-llms-openai\npip install llama-index-embeddings-openai\n```\n\n### 5-line RAG example\n\n```python\nfrom llama_index.core import VectorStoreIndex, SimpleDirectoryReader\n\n# Load documents\ndocuments = SimpleDirectoryReader(\"data\").load_data()\n\n# Create index\nindex = VectorStoreIndex.from_documents(documents)\n\n# Query\nquery_engine = index.as_query_engine()\nresponse = query_engine.query(\"What did the author do growing up?\")\nprint(response)\n```\n\n## Core concepts\n\n### 1. Data connectors - Load documents\n\n```python\nfrom llama_index.core import SimpleDirectoryReader, Document\nfrom llama_index.readers.web import SimpleWebPageReader\nfrom llama_index.readers.github import GithubRepositoryReader\n\n# Directory of files\ndocuments = SimpleDirectoryReader(\".\u002Fdata\").load_data()\n\n# Web pages\nreader = SimpleWebPageReader()\ndocuments = reader.load_data([\"https:\u002F\u002Fexample.com\"])\n\n# GitHub repository\nreader = GithubRepositoryReader(owner=\"user\", repo=\"repo\")\ndocuments = reader.load_data(branch=\"main\")\n\n# Manual document creation\ndoc = Document(\n    text=\"This is the document content\",\n    metadata={\"source\": \"manual\", \"date\": \"2025-01-01\"}\n)\n```\n\n### 2. Indices - Structure data\n\n```python\nfrom llama_index.core import VectorStoreIndex, ListIndex, TreeIndex\n\n# Vector index (most common - semantic search)\nvector_index = VectorStoreIndex.from_documents(documents)\n\n# List index (sequential scan)\nlist_index = ListIndex.from_documents(documents)\n\n# Tree index (hierarchical summary)\ntree_index = TreeIndex.from_documents(documents)\n\n# Save index\nindex.storage_context.persist(persist_dir=\".\u002Fstorage\")\n\n# Load index\nfrom llama_index.core import load_index_from_storage, StorageContext\nstorage_context = StorageContext.from_defaults(persist_dir=\".\u002Fstorage\")\nindex = load_index_from_storage(storage_context)\n```\n\n### 3. Query engines - Ask questions\n\n```python\n# Basic query\nquery_engine = index.as_query_engine()\nresponse = query_engine.query(\"What is the main topic?\")\nprint(response)\n\n# Streaming response\nquery_engine = index.as_query_engine(streaming=True)\nresponse = query_engine.query(\"Explain quantum computing\")\nfor text in response.response_gen:\n    print(text, end=\"\", flush=True)\n\n# Custom configuration\nquery_engine = index.as_query_engine(\n    similarity_top_k=3,          # Return top 3 chunks\n    response_mode=\"compact\",     # Or \"tree_summarize\", \"simple_summarize\"\n    verbose=True\n)\n```\n\n### 4. Retrievers - Find relevant chunks\n\n```python\n# Vector retriever\nretriever = index.as_retriever(similarity_top_k=5)\nnodes = retriever.retrieve(\"machine learning\")\n\n# With filtering\nretriever = index.as_retriever(\n    similarity_top_k=3,\n    filters={\"metadata.category\": \"tutorial\"}\n)\n\n# Custom retriever\nfrom llama_index.core.retrievers import BaseRetriever\n\nclass CustomRetriever(BaseRetriever):\n    def _retrieve(self, query_bundle):\n        # Your custom retrieval logic\n        return nodes\n```\n\n## Agents with tools\n\n### Basic agent\n\n```python\nfrom llama_index.core.agent import FunctionAgent\nfrom llama_index.llms.openai import OpenAI\n\n# Define tools\ndef multiply(a: int, b: int) -> int:\n    \"\"\"Multiply two numbers.\"\"\"\n    return a * b\n\ndef add(a: int, b: int) -> int:\n    \"\"\"Add two numbers.\"\"\"\n    return a + b\n\n# Create agent\nllm = OpenAI(model=\"gpt-4o\")\nagent = FunctionAgent.from_tools(\n    tools=[multiply, add],\n    llm=llm,\n    verbose=True\n)\n\n# Use agent\nresponse = agent.chat(\"What is 25 * 17 + 142?\")\nprint(response)\n```\n\n### RAG agent (document search + tools)\n\n```python\nfrom llama_index.core.tools import QueryEngineTool\n\n# Create index as before\nindex = VectorStoreIndex.from_documents(documents)\n\n# Wrap query engine as tool\nquery_tool = QueryEngineTool.from_defaults(\n    query_engine=index.as_query_engine(),\n    name=\"python_docs\",\n    description=\"Useful for answering questions about Python programming\"\n)\n\n# Agent with document search + calculator\nagent = FunctionAgent.from_tools(\n    tools=[query_tool, multiply, add],\n    llm=llm\n)\n\n# Agent decides when to search docs vs calculate\nresponse = agent.chat(\"According to the docs, what is Python used for?\")\n```\n\n## Advanced RAG patterns\n\n### Chat engine (conversational)\n\n```python\nfrom llama_index.core.chat_engine import CondensePlusContextChatEngine\n\n# Chat with memory\nchat_engine = index.as_chat_engine(\n    chat_mode=\"condense_plus_context\",  # Or \"context\", \"react\"\n    verbose=True\n)\n\n# Multi-turn conversation\nresponse1 = chat_engine.chat(\"What is Python?\")\nresponse2 = chat_engine.chat(\"Can you give examples?\")  # Remembers context\nresponse3 = chat_engine.chat(\"What about web frameworks?\")\n```\n\n### Metadata filtering\n\n```python\nfrom llama_index.core.vector_stores import MetadataFilters, ExactMatchFilter\n\n# Filter by metadata\nfilters = MetadataFilters(\n    filters=[\n        ExactMatchFilter(key=\"category\", value=\"tutorial\"),\n        ExactMatchFilter(key=\"difficulty\", value=\"beginner\")\n    ]\n)\n\nretriever = index.as_retriever(\n    similarity_top_k=3,\n    filters=filters\n)\n\nquery_engine = index.as_query_engine(filters=filters)\n```\n\n### Structured output\n\n```python\nfrom pydantic import BaseModel\nfrom llama_index.core.output_parsers import PydanticOutputParser\n\nclass Summary(BaseModel):\n    title: str\n    main_points: list[str]\n    conclusion: str\n\n# Get structured response\noutput_parser = PydanticOutputParser(output_cls=Summary)\nquery_engine = index.as_query_engine(output_parser=output_parser)\n\nresponse = query_engine.query(\"Summarize the document\")\nsummary = response  # Pydantic model\nprint(summary.title, summary.main_points)\n```\n\n## Data ingestion patterns\n\n### Multiple file types\n\n```python\n# Load all supported formats\ndocuments = SimpleDirectoryReader(\n    \".\u002Fdata\",\n    recursive=True,\n    required_exts=[\".pdf\", \".docx\", \".txt\", \".md\"]\n).load_data()\n```\n\n### Web scraping\n\n```python\nfrom llama_index.readers.web import BeautifulSoupWebReader\n\nreader = BeautifulSoupWebReader()\ndocuments = reader.load_data(urls=[\n    \"https:\u002F\u002Fdocs.python.org\u002F3\u002Ftutorial\u002F\",\n    \"https:\u002F\u002Fdocs.python.org\u002F3\u002Flibrary\u002F\"\n])\n```\n\n### Database\n\n```python\nfrom llama_index.readers.database import DatabaseReader\n\nreader = DatabaseReader(\n    sql_database_uri=\"postgresql:\u002F\u002Fuser:pass@localhost\u002Fdb\"\n)\ndocuments = reader.load_data(query=\"SELECT * FROM articles\")\n```\n\n### API endpoints\n\n```python\nfrom llama_index.readers.json import JSONReader\n\nreader = JSONReader()\ndocuments = reader.load_data(\"https:\u002F\u002Fapi.example.com\u002Fdata.json\")\n```\n\n## Vector store integrations\n\n### Chroma (local)\n\n```python\nfrom llama_index.vector_stores.chroma import ChromaVectorStore\nimport chromadb\n\n# Initialize Chroma\ndb = chromadb.PersistentClient(path=\".\u002Fchroma_db\")\ncollection = db.get_or_create_collection(\"my_collection\")\n\n# Create vector store\nvector_store = ChromaVectorStore(chroma_collection=collection)\n\n# Use in index\nfrom llama_index.core import StorageContext\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n### Pinecone (cloud)\n\n```python\nfrom llama_index.vector_stores.pinecone import PineconeVectorStore\nimport pinecone\n\n# Initialize Pinecone\npinecone.init(api_key=\"your-key\", environment=\"us-west1-gcp\")\npinecone_index = pinecone.Index(\"my-index\")\n\n# Create vector store\nvector_store = PineconeVectorStore(pinecone_index=pinecone_index)\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\n\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n### FAISS (fast)\n\n```python\nfrom llama_index.vector_stores.faiss import FaissVectorStore\nimport faiss\n\n# Create FAISS index\nd = 1536  # Dimension of embeddings\nfaiss_index = faiss.IndexFlatL2(d)\n\nvector_store = FaissVectorStore(faiss_index=faiss_index)\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\n\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n## Customization\n\n### Custom LLM\n\n```python\nfrom llama_index.llms.anthropic import Anthropic\nfrom llama_index.core import Settings\n\n# Set global LLM\nSettings.llm = Anthropic(model=\"claude-sonnet-4-5-20250929\")\n\n# Now all queries use Anthropic\nquery_engine = index.as_query_engine()\n```\n\n### Custom embeddings\n\n```python\nfrom llama_index.embeddings.huggingface import HuggingFaceEmbedding\n\n# Use HuggingFace embeddings\nSettings.embed_model = HuggingFaceEmbedding(\n    model_name=\"sentence-transformers\u002Fall-mpnet-base-v2\"\n)\n\nindex = VectorStoreIndex.from_documents(documents)\n```\n\n### Custom prompt templates\n\n```python\nfrom llama_index.core import PromptTemplate\n\nqa_prompt = PromptTemplate(\n    \"Context: {context_str}\\n\"\n    \"Question: {query_str}\\n\"\n    \"Answer the question based only on the context. \"\n    \"If the answer is not in the context, say 'I don't know'.\\n\"\n    \"Answer: \"\n)\n\nquery_engine = index.as_query_engine(text_qa_template=qa_prompt)\n```\n\n## Multi-modal RAG\n\n### Image + text\n\n```python\nfrom llama_index.core import SimpleDirectoryReader\nfrom llama_index.multi_modal_llms.openai import OpenAIMultiModal\n\n# Load images and documents\ndocuments = SimpleDirectoryReader(\n    \".\u002Fdata\",\n    required_exts=[\".jpg\", \".png\", \".pdf\"]\n).load_data()\n\n# Multi-modal index\nindex = VectorStoreIndex.from_documents(documents)\n\n# Query with multi-modal LLM\nmulti_modal_llm = OpenAIMultiModal(model=\"gpt-4o\")\nquery_engine = index.as_query_engine(llm=multi_modal_llm)\n\nresponse = query_engine.query(\"What is in the diagram on page 3?\")\n```\n\n## Evaluation\n\n### Response quality\n\n```python\nfrom llama_index.core.evaluation import RelevancyEvaluator, FaithfulnessEvaluator\n\n# Evaluate relevance\nrelevancy = RelevancyEvaluator()\nresult = relevancy.evaluate_response(\n    query=\"What is Python?\",\n    response=response\n)\nprint(f\"Relevancy: {result.passing}\")\n\n# Evaluate faithfulness (no hallucination)\nfaithfulness = FaithfulnessEvaluator()\nresult = faithfulness.evaluate_response(\n    query=\"What is Python?\",\n    response=response\n)\nprint(f\"Faithfulness: {result.passing}\")\n```\n\n## Best practices\n\n1. **Use vector indices for most cases** - Best performance\n2. **Save indices to disk** - Avoid re-indexing\n3. **Chunk documents properly** - 512-1024 tokens optimal\n4. **Add metadata** - Enables filtering and tracking\n5. **Use streaming** - Better UX for long responses\n6. **Enable verbose during dev** - See retrieval process\n7. **Evaluate responses** - Check relevance and faithfulness\n8. **Use chat engine for conversations** - Built-in memory\n9. **Persist storage** - Don't lose your index\n10. **Monitor costs** - Track embedding and LLM usage\n\n## Common patterns\n\n### Document Q&A system\n\n```python\n# Complete RAG pipeline\ndocuments = SimpleDirectoryReader(\"docs\").load_data()\nindex = VectorStoreIndex.from_documents(documents)\nindex.storage_context.persist(persist_dir=\".\u002Fstorage\")\n\n# Query\nquery_engine = index.as_query_engine(\n    similarity_top_k=3,\n    response_mode=\"compact\",\n    verbose=True\n)\nresponse = query_engine.query(\"What is the main topic?\")\nprint(response)\nprint(f\"Sources: {[node.metadata['file_name'] for node in response.source_nodes]}\")\n```\n\n### Chatbot with memory\n\n```python\n# Conversational interface\nchat_engine = index.as_chat_engine(\n    chat_mode=\"condense_plus_context\",\n    verbose=True\n)\n\n# Multi-turn chat\nwhile True:\n    user_input = input(\"You: \")\n    if user_input.lower() == \"quit\":\n        break\n    response = chat_engine.chat(user_input)\n    print(f\"Bot: {response}\")\n```\n\n## Performance benchmarks\n\n| Operation | Latency | Notes |\n|-----------|---------|-------|\n| Index 100 docs | ~10-30s | One-time, can persist |\n| Query (vector) | ~0.5-2s | Retrieval + LLM |\n| Streaming query | ~0.5s first token | Better UX |\n| Agent with tools | ~3-8s | Multiple tool calls |\n\n## LlamaIndex vs LangChain\n\n| Feature | LlamaIndex | LangChain |\n|---------|------------|-----------|\n| **Best for** | RAG, document Q&A | Agents, general LLM apps |\n| **Data connectors** | 300+ (LlamaHub) | 100+ |\n| **RAG focus** | Core feature | One of many |\n| **Learning curve** | Easier for RAG | Steeper |\n| **Customization** | High | Very high |\n| **Documentation** | Excellent | Good |\n\n**Use LlamaIndex when:**\n- Your primary use case is RAG\n- Need many data connectors\n- Want simpler API for document Q&A\n- Building knowledge retrieval system\n\n**Use LangChain when:**\n- Building complex agents\n- Need more general-purpose tools\n- Want more flexibility\n- Complex multi-step workflows\n\n## References\n\n- **[Query Engines Guide](references\u002Fquery_engines.md)** - Query modes, customization, streaming\n- **[Agents Guide](references\u002Fagents.md)** - Tool creation, RAG agents, multi-step reasoning\n- **[Data Connectors Guide](references\u002Fdata_connectors.md)** - 300+ connectors, custom loaders\n\n## Resources\n\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Frun-llama\u002Fllama_index ⭐ 45,100+\n- **Docs**: https:\u002F\u002Fdevelopers.llamaindex.ai\u002Fpython\u002Fframework\u002F\n- **LlamaHub**: https:\u002F\u002Fllamahub.ai (data connectors)\n- **LlamaCloud**: https:\u002F\u002Fcloud.llamaindex.ai (enterprise)\n- **Discord**: https:\u002F\u002Fdiscord.gg\u002FdGcwcsnxhU\n- **Version**: 0.14.7+\n- **License**: MIT\n\n\n","# LlamaIndex - Data Framework for LLM Applications\n\nThe leading framework for connecting LLMs with your data.\n\n## When to use LlamaIndex\n\n**Use LlamaIndex when:**\n- Building RAG (retrieval-augmented generation) applications\n- Need document question-answering over private data\n- Ingesting data from multiple sources (300+ connectors)\n- Creating knowledge bases for LLMs\n- Building chatbots with enterprise data\n- Need structured data extraction from documents\n\n**Metrics**:\n- **45,100+ GitHub stars**\n- **23,000+ repositories** use LlamaIndex\n- **300+ data connectors** (LlamaHub)\n- **1,715+ contributors**\n- **v0.14.7** (stable)\n\n**Use alternatives instead**:\n- **LangChain**: More general-purpose, better for agents\n- **Haystack**: Production search pipelines\n- **txtai**: Lightweight semantic search\n- **Chroma**: Just need vector storage\n\n## Quick start\n\n### Installation\n\n```bash\n# Starter package (recommended)\npip install llama-index\n\n# Or minimal core + specific integrations\npip install llama-index-core\npip install llama-index-llms-openai\npip install llama-index-embeddings-openai\n```\n\n### 5-line RAG example\n\n```python\nfrom llama_index.core import VectorStoreIndex, SimpleDirectoryReader\n\n# Load documents\ndocuments = SimpleDirectoryReader(\"data\").load_data()\n\n# Create index\nindex = VectorStoreIndex.from_documents(documents)\n\n# Query\nquery_engine = index.as_query_engine()\nresponse = query_engine.query(\"What did the author do growing up?\")\nprint(response)\n```\n\n## Core concepts\n\n### 1. Data connectors - Load documents\n\n```python\nfrom llama_index.core import SimpleDirectoryReader, Document\nfrom llama_index.readers.web import SimpleWebPageReader\nfrom llama_index.readers.github import GithubRepositoryReader\n\n# Directory of files\ndocuments = SimpleDirectoryReader(\".\u002Fdata\").load_data()\n\n# Web pages\nreader = SimpleWebPageReader()\ndocuments = reader.load_data([\"https:\u002F\u002Fexample.com\"])\n\n# GitHub repository\nreader = GithubRepositoryReader(owner=\"user\", repo=\"repo\")\ndocuments = reader.load_data(branch=\"main\")\n\n# Manual document creation\ndoc = Document(\n    text=\"This is the document content\",\n    metadata={\"source\": \"manual\", \"date\": \"2025-01-01\"}\n)\n```\n\n### 2. Indices - Structure data\n\n```python\nfrom llama_index.core import VectorStoreIndex, ListIndex, TreeIndex\n\n# Vector index (most common - semantic search)\nvector_index = VectorStoreIndex.from_documents(documents)\n\n# List index (sequential scan)\nlist_index = ListIndex.from_documents(documents)\n\n# Tree index (hierarchical summary)\ntree_index = TreeIndex.from_documents(documents)\n\n# Save index\nindex.storage_context.persist(persist_dir=\".\u002Fstorage\")\n\n# Load index\nfrom llama_index.core import load_index_from_storage, StorageContext\nstorage_context = StorageContext.from_defaults(persist_dir=\".\u002Fstorage\")\nindex = load_index_from_storage(storage_context)\n```\n\n### 3. Query engines - Ask questions\n\n```python\n# Basic query\nquery_engine = index.as_query_engine()\nresponse = query_engine.query(\"What is the main topic?\")\nprint(response)\n\n# Streaming response\nquery_engine = index.as_query_engine(streaming=True)\nresponse = query_engine.query(\"Explain quantum computing\")\nfor text in response.response_gen:\n    print(text, end=\"\", flush=True)\n\n# Custom configuration\nquery_engine = index.as_query_engine(\n    similarity_top_k=3,          # Return top 3 chunks\n    response_mode=\"compact\",     # Or \"tree_summarize\", \"simple_summarize\"\n    verbose=True\n)\n```\n\n### 4. Retrievers - Find relevant chunks\n\n```python\n# Vector retriever\nretriever = index.as_retriever(similarity_top_k=5)\nnodes = retriever.retrieve(\"machine learning\")\n\n# With filtering\nretriever = index.as_retriever(\n    similarity_top_k=3,\n    filters={\"metadata.category\": \"tutorial\"}\n)\n\n# Custom retriever\nfrom llama_index.core.retrievers import BaseRetriever\n\nclass CustomRetriever(BaseRetriever):\n    def _retrieve(self, query_bundle):\n        # Your custom retrieval logic\n        return nodes\n```\n\n## Agents with tools\n\n### Basic agent\n\n```python\nfrom llama_index.core.agent import FunctionAgent\nfrom llama_index.llms.openai import OpenAI\n\n# Define tools\ndef multiply(a: int, b: int) -> int:\n    \"\"\"Multiply two numbers.\"\"\"\n    return a * b\n\ndef add(a: int, b: int) -> int:\n    \"\"\"Add two numbers.\"\"\"\n    return a + b\n\n# Create agent\nllm = OpenAI(model=\"gpt-4o\")\nagent = FunctionAgent.from_tools(\n    tools=[multiply, add],\n    llm=llm,\n    verbose=True\n)\n\n# Use agent\nresponse = agent.chat(\"What is 25 * 17 + 142?\")\nprint(response)\n```\n\n### RAG agent (document search + tools)\n\n```python\nfrom llama_index.core.tools import QueryEngineTool\n\n# Create index as before\nindex = VectorStoreIndex.from_documents(documents)\n\n# Wrap query engine as tool\nquery_tool = QueryEngineTool.from_defaults(\n    query_engine=index.as_query_engine(),\n    name=\"python_docs\",\n    description=\"Useful for answering questions about Python programming\"\n)\n\n# Agent with document search + calculator\nagent = FunctionAgent.from_tools(\n    tools=[query_tool, multiply, add],\n    llm=llm\n)\n\n# Agent decides when to search docs vs calculate\nresponse = agent.chat(\"According to the docs, what is Python used for?\")\n```\n\n## Advanced RAG patterns\n\n### Chat engine (conversational)\n\n```python\nfrom llama_index.core.chat_engine import CondensePlusContextChatEngine\n\n# Chat with memory\nchat_engine = index.as_chat_engine(\n    chat_mode=\"condense_plus_context\",  # Or \"context\", \"react\"\n    verbose=True\n)\n\n# Multi-turn conversation\nresponse1 = chat_engine.chat(\"What is Python?\")\nresponse2 = chat_engine.chat(\"Can you give examples?\")  # Remembers context\nresponse3 = chat_engine.chat(\"What about web frameworks?\")\n```\n\n### Metadata filtering\n\n```python\nfrom llama_index.core.vector_stores import MetadataFilters, ExactMatchFilter\n\n# Filter by metadata\nfilters = MetadataFilters(\n    filters=[\n        ExactMatchFilter(key=\"category\", value=\"tutorial\"),\n        ExactMatchFilter(key=\"difficulty\", value=\"beginner\")\n    ]\n)\n\nretriever = index.as_retriever(\n    similarity_top_k=3,\n    filters=filters\n)\n\nquery_engine = index.as_query_engine(filters=filters)\n```\n\n### Structured output\n\n```python\nfrom pydantic import BaseModel\nfrom llama_index.core.output_parsers import PydanticOutputParser\n\nclass Summary(BaseModel):\n    title: str\n    main_points: list[str]\n    conclusion: str\n\n# Get structured response\noutput_parser = PydanticOutputParser(output_cls=Summary)\nquery_engine = index.as_query_engine(output_parser=output_parser)\n\nresponse = query_engine.query(\"Summarize the document\")\nsummary = response  # Pydantic model\nprint(summary.title, summary.main_points)\n```\n\n## Data ingestion patterns\n\n### Multiple file types\n\n```python\n# Load all supported formats\ndocuments = SimpleDirectoryReader(\n    \".\u002Fdata\",\n    recursive=True,\n    required_exts=[\".pdf\", \".docx\", \".txt\", \".md\"]\n).load_data()\n```\n\n### Web scraping\n\n```python\nfrom llama_index.readers.web import BeautifulSoupWebReader\n\nreader = BeautifulSoupWebReader()\ndocuments = reader.load_data(urls=[\n    \"https:\u002F\u002Fdocs.python.org\u002F3\u002Ftutorial\u002F\",\n    \"https:\u002F\u002Fdocs.python.org\u002F3\u002Flibrary\u002F\"\n])\n```\n\n### Database\n\n```python\nfrom llama_index.readers.database import DatabaseReader\n\nreader = DatabaseReader(\n    sql_database_uri=\"postgresql:\u002F\u002Fuser:pass@localhost\u002Fdb\"\n)\ndocuments = reader.load_data(query=\"SELECT * FROM articles\")\n```\n\n### API endpoints\n\n```python\nfrom llama_index.readers.json import JSONReader\n\nreader = JSONReader()\ndocuments = reader.load_data(\"https:\u002F\u002Fapi.example.com\u002Fdata.json\")\n```\n\n## Vector store integrations\n\n### Chroma (local)\n\n```python\nfrom llama_index.vector_stores.chroma import ChromaVectorStore\nimport chromadb\n\n# Initialize Chroma\ndb = chromadb.PersistentClient(path=\".\u002Fchroma_db\")\ncollection = db.get_or_create_collection(\"my_collection\")\n\n# Create vector store\nvector_store = ChromaVectorStore(chroma_collection=collection)\n\n# Use in index\nfrom llama_index.core import StorageContext\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n### Pinecone (cloud)\n\n```python\nfrom llama_index.vector_stores.pinecone import PineconeVectorStore\nimport pinecone\n\n# Initialize Pinecone\npinecone.init(api_key=\"your-key\", environment=\"us-west1-gcp\")\npinecone_index = pinecone.Index(\"my-index\")\n\n# Create vector store\nvector_store = PineconeVectorStore(pinecone_index=pinecone_index)\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\n\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n### FAISS (fast)\n\n```python\nfrom llama_index.vector_stores.faiss import FaissVectorStore\nimport faiss\n\n# Create FAISS index\nd = 1536  # Dimension of embeddings\nfaiss_index = faiss.IndexFlatL2(d)\n\nvector_store = FaissVectorStore(faiss_index=faiss_index)\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\n\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\n```\n\n## Customization\n\n### Custom LLM\n\n```python\nfrom llama_index.llms.anthropic import Anthropic\nfrom llama_index.core import Settings\n\n# Set global LLM\nSettings.llm = Anthropic(model=\"claude-sonnet-4-5-20250929\")\n\n# Now all queries use Anthropic\nquery_engine = index.as_query_engine()\n```\n\n### Custom embeddings\n\n```python\nfrom llama_index.embeddings.huggingface import HuggingFaceEmbedding\n\n# Use HuggingFace embeddings\nSettings.embed_model = HuggingFaceEmbedding(\n    model_name=\"sentence-transformers\u002Fall-mpnet-base-v2\"\n)\n\nindex = VectorStoreIndex.from_documents(documents)\n```\n\n### Custom prompt templates\n\n```python\nfrom llama_index.core import PromptTemplate\n\nqa_prompt = PromptTemplate(\n    \"Context: {context_str}\\n\"\n    \"Question: {query_str}\\n\"\n    \"Answer the question based only on the context. \"\n    \"If the answer is not in the context, say 'I don't know'.\\n\"\n    \"Answer: \"\n)\n\nquery_engine = index.as_query_engine(text_qa_template=qa_prompt)\n```\n\n## Multi-modal RAG\n\n### Image + text\n\n```python\nfrom llama_index.core import SimpleDirectoryReader\nfrom llama_index.multi_modal_llms.openai import OpenAIMultiModal\n\n# Load images and documents\ndocuments = SimpleDirectoryReader(\n    \".\u002Fdata\",\n    required_exts=[\".jpg\", \".png\", \".pdf\"]\n).load_data()\n\n# Multi-modal index\nindex = VectorStoreIndex.from_documents(documents)\n\n# Query with multi-modal LLM\nmulti_modal_llm = OpenAIMultiModal(model=\"gpt-4o\")\nquery_engine = index.as_query_engine(llm=multi_modal_llm)\n\nresponse = query_engine.query(\"What is in the diagram on page 3?\")\n```\n\n## Evaluation\n\n### Response quality\n\n```python\nfrom llama_index.core.evaluation import RelevancyEvaluator, FaithfulnessEvaluator\n\n# Evaluate relevance\nrelevancy = RelevancyEvaluator()\nresult = relevancy.evaluate_response(\n    query=\"What is Python?\",\n    response=response\n)\nprint(f\"Relevancy: {result.passing}\")\n\n# Evaluate faithfulness (no hallucination)\nfaithfulness = FaithfulnessEvaluator()\nresult = faithfulness.evaluate_response(\n    query=\"What is Python?\",\n    response=response\n)\nprint(f\"Faithfulness: {result.passing}\")\n```\n\n## Best practices\n\n1. **Use vector indices for most cases** - Best performance\n2. **Save indices to disk** - Avoid re-indexing\n3. **Chunk documents properly** - 512-1024 tokens optimal\n4. **Add metadata** - Enables filtering and tracking\n5. **Use streaming** - Better UX for long responses\n6. **Enable verbose during dev** - See retrieval process\n7. **Evaluate responses** - Check relevance and faithfulness\n8. **Use chat engine for conversations** - Built-in memory\n9. **Persist storage** - Don't lose your index\n10. **Monitor costs** - Track embedding and LLM usage\n\n## Common patterns\n\n### Document Q&A system\n\n```python\n# Complete RAG pipeline\ndocuments = SimpleDirectoryReader(\"docs\").load_data()\nindex = VectorStoreIndex.from_documents(documents)\nindex.storage_context.persist(persist_dir=\".\u002Fstorage\")\n\n# Query\nquery_engine = index.as_query_engine(\n    similarity_top_k=3,\n    response_mode=\"compact\",\n    verbose=True\n)\nresponse = query_engine.query(\"What is the main topic?\")\nprint(response)\nprint(f\"Sources: {[node.metadata['file_name'] for node in response.source_nodes]}\")\n```\n\n### Chatbot with memory\n\n```python\n# Conversational interface\nchat_engine = index.as_chat_engine(\n    chat_mode=\"condense_plus_context\",\n    verbose=True\n)\n\n# Multi-turn chat\nwhile True:\n    user_input = input(\"You: \")\n    if user_input.lower() == \"quit\":\n        break\n    response = chat_engine.chat(user_input)\n    print(f\"Bot: {response}\")\n```\n\n## Performance benchmarks\n\n| Operation | Latency | Notes |\n|-----------|---------|-------|\n| Index 100 docs | ~10-30s | One-time, can persist |\n| Query (vector) | ~0.5-2s | Retrieval + LLM |\n| Streaming query | ~0.5s first token | Better UX |\n| Agent with tools | ~3-8s | Multiple tool calls |\n\n## LlamaIndex vs LangChain\n\n| Feature | LlamaIndex | LangChain |\n|---------|------------|-----------|\n| **Best for** | RAG, document Q&A | Agents, general LLM apps |\n| **Data connectors** | 300+ (LlamaHub) | 100+ |\n| **RAG focus** | Core feature | One of many |\n| **Learning curve** | Easier for RAG | Steeper |\n| **Customization** | High | Very high |\n| **Documentation** | Excellent | Good |\n\n**Use LlamaIndex when:**\n- Your primary use case is RAG\n- Need many data connectors\n- Want simpler API for document Q&A\n- Building knowledge retrieval system\n\n**Use LangChain when:**\n- Building complex agents\n- Need more general-purpose tools\n- Want more flexibility\n- Complex multi-step workflows\n\n## References\n\n- **[Query Engines Guide](references\u002Fquery_engines.md)** - Query modes, customization, streaming\n- **[Agents Guide](references\u002Fagents.md)** - Tool creation, RAG agents, multi-step reasoning\n- **[Data Connectors Guide](references\u002Fdata_connectors.md)** - 300+ connectors, custom loaders\n\n## Resources\n\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Frun-llama\u002Fllama_index ⭐ 45,100+\n- **Docs**: https:\u002F\u002Fdevelopers.llamaindex.ai\u002Fpython\u002Fframework\u002F\n- **LlamaHub**: https:\u002F\u002Fllamahub.ai (data connectors)\n- **LlamaCloud**: https:\u002F\u002Fcloud.llamaindex.ai (enterprise)\n- **Discord**: https:\u002F\u002Fdiscord.gg\u002FdGcwcsnxhU\n- **Version**: 0.14.7+\n- **License**: MIT\n\n\n",{"name":5,"tags":24,"author":36,"license":15,"version":37,"description":7,"dependencies":38},[25,26,27,28,29,30,31,32,33,34,35],"Agents","LlamaIndex","RAG","Document Ingestion","Vector Indices","Query Engines","Knowledge Retrieval","Data Framework","Multimodal","Private Data","Connectors","Orchestra Research","1.0.0",[39,40,41],"llama-index","openai","anthropic","main","18193d71ea3441d7c033dbf195b8a7c75ac96f75","packages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex",0,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science","https:\u002F\u002Fgithub.com\u002FCZX-LBJ","一个没事喜欢随便写点代码的小菜鸡\r\n","2026-10-22 09:48:04",{"source":8,"skillName":5,"commands":51,"prompt":61,"zipAvailable":62},[52,55,58],{"tool":53,"command":54},"npx","npx skills add CZX-LBJ\u002Fdeepseek-science --skill llamaindex",{"tool":56,"command":57},"bunx","bunx skills add CZX-LBJ\u002Fdeepseek-science --skill llamaindex",{"tool":59,"command":60},"pnpm","pnpm dlx skills add CZX-LBJ\u002Fdeepseek-science --skill llamaindex","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex\u002FSKILL.md",true,[64,70,74],{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},"references\u002Fagents.md","REFERENCE","参考文档",1958,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex\u002Freferences\u002Fagents.md",{"path":71,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":72,"sourceHtmlUrl":73},"references\u002Fdata_connectors.md",2307,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex\u002Freferences\u002Fdata_connectors.md",{"path":75,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":76,"sourceHtmlUrl":77},"references\u002Fquery_engines.md",9905,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fai-research-skills\u002F14-agents\u002Fllamaindex\u002Freferences\u002Fquery_engines.md",[79,86,93,100,107,114],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":19},"01M7FEEKSNMRMKW6JEK93PFDFY","bio-free-energy-calculations","Performs alchemical free-energy calculations including relative binding free energy (RBFE \u002F FEP+) and absolute binding free energy (ABFE) via OpenFE, FEP+, GROMACS, AMBER pmemd, and OpenMM with explicit lambda scheduling, soft-core potentials, MBAR\u002FBAR analysis, cycle-closure validation, and protocol-appropriate enhanced sampling. Compares ML alternatives (Boltz-2 affinity, DeepDock). Use when ranking analogs by binding affinity beyond docking accuracy, performing prospective lead optimization, or validating SAR predictions.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Ffree-energy-calculations\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":19},"01M7FEE5CTE5QK55CEMXNNVWY9","bio-conformer-generation","Generates 3D conformer ensembles using RDKit ETKDGv3 with knowledge-enhanced distance geometry, MMFF94\u002FUFF force-field optimization, CREST + GFN2-xTB semi-empirical refinement, and macrocycle-aware torsion preferences. Provides explicit decision rules for single vs ensemble conformer use, RMSD pruning, energy windows, conformer count, and force-field choice. Use when preparing 3D ligands for docking, generating descriptor input for 3D QSAR, or sampling macrocycle\u002Fpeptide conformational ensembles.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fconformer-generation\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":19},"01M7FEDNGFEMFWD1MSEDR5H7CY","bio-admet-prediction","Predicts ADMET properties using ADMETlab 3.0 (119 platform features, including 77 prediction models with modeled-endpoint uncertainty), ADMET-AI, DeepChem MolNet, and chemprop D-MPNN with explicit handling of OECD QSAR principles, applicability domain assessment, calibration, hERG\u002FCYP\u002FAMES endpoints, and PAINS \u002F Lipinski \u002F Ro5 \u002F Veber \u002F BBB druglikeness filters. Use when filtering compounds for drug-likeness, prioritizing leads by predicted safety, or building an in-house ADMET QSAR model.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fadmet-prediction\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":19},"01M7FED69GFB54FRF7AV5066DN","writing-systems-papers","Paragraph-level structural blueprint for 10-12 page systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys. Provides page allocation, paragraph templates, and writing patterns. Use when user says \"写系统论文\", \"systems paper structure\", \"OSDI paper\", \"SOSP paper\", or wants fine-grained structural guidance for a systems conference submission.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwriting-systems-papers\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":19},"01M7FECVVDWZ0X0VEDTCE4B4DZ","wiki-enrich","Fill in the per-paper TODO sections of research-wiki\u002Fpapers\u002F\u003Cslug>.md pages that literature-ingest skills leave as bare scaffolds. Use when user says 'enrich wiki', 'fill paper TODOs', 'wiki body 補完', '把 paper 摘要寫進 wiki', 'research-wiki 自動填', or after a batch ingest that left papers\u002F as TODO scaffolds.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwiki-enrich\u002FSKILL.md",[],[],{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":118,"categories":119,"keywords":120,"publishedAt":19},"01M7FECD4BCJ2MTSBDDSE3ZAJ5","web-debug-search","Search GitHub, Stack Exchange, Chinese technical communities, official documentation, and general developer web sources for software errors, compatibility problems, API usage questions, and real-world workarounds. Use for debugging and discovery only; results are not paper-citation evidence.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fweb-debug-search\u002FSKILL.md",[],[]]