[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AE1KB9P0MNFWDRTH1D6JBY":3,"skill-same:01M7AE1KB9P0MNFWDRTH1D6JBY":44},{"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":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M7AE1KB9P0MNFWDRTH1D6JBY","nodetool-rag-indexing",null,"Set up RAG pipelines, vector indexing, document ingestion, vector search, and knowledge base creation in NodeTool. Use when user asks about RAG, document indexing, vector search, chat with documents, knowledge base, embeddings, or collection management.","hanchenhao\u002Fai-canvas","hanchenhao","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F15713703?v=4","01M7ADWCF5AE30N3MG7MDF3RV2","01M7ADWCFVJS4V8VD9KWJG22WR",1,"AGPL-3.0","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fnodetool-rag-indexing\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: nodetool-rag-indexing\ndescription: Set up RAG pipelines, vector indexing, document ingestion, vector search, and knowledge base creation in NodeTool. Use when user asks about RAG, document indexing, vector search, chat with documents, knowledge base, embeddings, or collection management.\n---\n\nYou help users build Retrieval-Augmented Generation (RAG) pipelines in NodeTool.\n\n# RAG Architecture\n\n```\nINDEXING:  Documents → Load → Split → Embed → Store (vector collection)\nQUERY:     Question → Embed → Search → Format → LLM → Answer\n```\n\n# Vector Store Backends\n\nNodeTool's vector store (`@nodetool-ai\u002Fvectorstore`) is backend-pluggable. The\nworkflow nodes are the same regardless of backend — you pick the backend via\nconfiguration.\n\n| Backend | Best for | Notes |\n|---------|----------|-------|\n| **SQLite-vec** | Default, local, embedded | No external service |\n| **ChromaDB** | Self-host \u002F remote | `CHROMA_URL`, `CHROMA_PATH`, `CHROMA_TOKEN` |\n| **Pinecone** | Managed cloud | API-key based |\n| **Supabase (pgvector)** | Postgres-backed | Pairs with Supabase auth\u002Fstorage |\n\nThere are **no FAISS nodes**; the backends above cover local and hosted use.\n\n# Vector Nodes (`vector.*`)\n\nAll RAG nodes live under the single `vector.*` namespace (not `vector.chroma.*`\nor `vector.faiss.*`).\n\n| Node | Purpose |\n|------|---------|\n| `vector.Collection` | Reference\u002Fselect a collection by name (the collection ref other nodes consume) |\n| `vector.IndexTextChunk` | Index a single text chunk with its embedding |\n| `vector.IndexString` | Index a string value |\n| `vector.IndexAggregatedText` | Index aggregated text |\n| `vector.IndexEmbedding` | Index a precomputed embedding |\n| `vector.IndexImage` | Index an image |\n| `vector.QueryText` | Vector similarity search over text |\n| `vector.QueryImage` | Vector similarity search over images |\n| `vector.HybridSearch` | Vector + keyword search (best accuracy) |\n| `vector.GetDocuments` | Retrieve specific documents |\n| `vector.Count` | Count documents in a collection |\n| `vector.Peek` | Preview collection contents |\n| `vector.RemoveOverlap` | De-duplicate overlapping chunks in results |\n\nQuery nodes (`QueryText`, `QueryImage`, `HybridSearch`) output `ids`,\n`documents`, `metadatas`, and `distances` (HybridSearch also returns `scores`).\n\n# Document Loading & Splitting\n\n| Node | Namespace | Purpose |\n|------|-----------|---------|\n| `Code` | `nodetool.code` | Enumerate files with `await workspace.list(dir)` |\n| `LoadDocumentFile` | `nodetool.document` | Load a PDF\u002FTXT\u002FMD into a document |\n| `Chunk` | `nodetool.text` | Fixed-size word chunking with overlap (general purpose) |\n| `RegexSplit` | `nodetool.text` | Structure-aware splitting on a delimiter pattern |\n\n## Chunk Size Guidance\n\n| Content type | Chunk size | Overlap |\n|--------------|-----------|---------|\n| Technical docs | 200-500 tokens | 50 tokens |\n| Prose\u002Farticles | 300-600 tokens | 75 tokens |\n| Code | 100-300 tokens | 25 tokens |\n\n# Indexing — Workflow Pattern\n\n```\nListFiles → LoadDocumentFile → Chunk → IndexTextChunk(collection)\n```\n\nPair every index\u002Fquery node with a `vector.Collection` node (or a collection\nname) so they target the same store. Use the **same embedding model** for\nindexing and querying.\n\n# Indexing — HTTP API\n\n```bash\n# Index a file into a collection\ncurl -X POST http:\u002F\u002Flocalhost:7777\u002Fapi\u002Fcollections\u002F\u003Cname>\u002Findex \\\n  -H \"Authorization: Bearer TOKEN\" \\\n  -H \"Content-Type: application\u002Fjson\" \\\n  -d '{\"file_path\": \"\u002Fpath\u002Fto\u002Fdocument.pdf\"}'\n```\n\nThe server resolves the collection, runs its ingestion workflow if one is\nregistered, otherwise falls back to split → embed → store.\n\n# Query — Workflow Pattern\n\n```\nChatInput → HybridSearch(collection, top_k) → FormatText → Agent → Output\n```\n\n| Node | Purpose |\n|------|---------|\n| `ChatInput` | User question |\n| `vector.HybridSearch` | Vector + keyword retrieval (best accuracy) |\n| `vector.QueryText` | Vector-only retrieval (faster) |\n| `FormatText` | Build the context string for the LLM |\n| `Agent` | Generate the answer from context + question |\n| `Output` | Return the answer |\n\n# Complete RAG Example\n\n## Index\n```\nListFiles(\"\u002Fdocs\u002F\") → LoadDocumentFile → Chunk(length=400, overlap=50)\n                                                  ↓\n                                  IndexTextChunk(collection=\"my-docs\")\n```\n\n## Query\n```\nChatInput(\"What is...?\") → HybridSearch(collection=\"my-docs\", top_k=5)\n                                  ↓\n                         FormatText(template=\"Context:\\n{documents}\\n\\nQuestion: {query}\")\n                                  ↓\n                         Agent(model=gpt-5.4, system=\"Answer using only the context provided.\")\n                                  ↓\n                         Output\n```\n\n# Environment Variables\n\n```bash\n# ChromaDB backend (only when using Chroma — SQLite-vec needs no config)\nCHROMA_URL=                          # Remote Chroma URL (empty = local)\nCHROMA_PATH=~\u002F.local\u002Fshare\u002Fnodetool\u002Fchroma  # Local storage path\nCHROMA_TOKEN=                        # Optional auth token\n```\n\nThe embedding model is chosen on the index\u002Fquery nodes via model selection\n(e.g. `text-embedding-3-small`, or a local sentence-transformers model).\n\n# Common Pitfalls\n\n- **Embedding model mismatch**: use the same embedding model for indexing and search.\n- **Chunks too large**: dilute the LLM context — keep to 200-500 tokens.\n- **Chunks too small**: sentences get fragmented; use 10-20% overlap.\n- **Empty collection**: index before querying — an unindexed collection returns nothing.\n- **Wrong node names**: it's `vector.IndexTextChunk` \u002F `vector.QueryText` \u002F `vector.HybridSearch`, not `IndexTextChunks` \u002F `TextSearch`, and there is no `vector.chroma.*`\u002F`vector.faiss.*` namespace.\n- **No `nodetool collections` CLI**: manage collections through the editor UI or the `\u002Fapi\u002Fcollections\u002F...` endpoints.\n","You help users build Retrieval-Augmented Generation (RAG) pipelines in NodeTool.\n\n# RAG Architecture\n\n```\nINDEXING:  Documents → Load → Split → Embed → Store (vector collection)\nQUERY:     Question → Embed → Search → Format → LLM → Answer\n```\n\n# Vector Store Backends\n\nNodeTool's vector store (`@nodetool-ai\u002Fvectorstore`) is backend-pluggable. The\nworkflow nodes are the same regardless of backend — you pick the backend via\nconfiguration.\n\n| Backend | Best for | Notes |\n|---------|----------|-------|\n| **SQLite-vec** | Default, local, embedded | No external service |\n| **ChromaDB** | Self-host \u002F remote | `CHROMA_URL`, `CHROMA_PATH`, `CHROMA_TOKEN` |\n| **Pinecone** | Managed cloud | API-key based |\n| **Supabase (pgvector)** | Postgres-backed | Pairs with Supabase auth\u002Fstorage |\n\nThere are **no FAISS nodes**; the backends above cover local and hosted use.\n\n# Vector Nodes (`vector.*`)\n\nAll RAG nodes live under the single `vector.*` namespace (not `vector.chroma.*`\nor `vector.faiss.*`).\n\n| Node | Purpose |\n|------|---------|\n| `vector.Collection` | Reference\u002Fselect a collection by name (the collection ref other nodes consume) |\n| `vector.IndexTextChunk` | Index a single text chunk with its embedding |\n| `vector.IndexString` | Index a string value |\n| `vector.IndexAggregatedText` | Index aggregated text |\n| `vector.IndexEmbedding` | Index a precomputed embedding |\n| `vector.IndexImage` | Index an image |\n| `vector.QueryText` | Vector similarity search over text |\n| `vector.QueryImage` | Vector similarity search over images |\n| `vector.HybridSearch` | Vector + keyword search (best accuracy) |\n| `vector.GetDocuments` | Retrieve specific documents |\n| `vector.Count` | Count documents in a collection |\n| `vector.Peek` | Preview collection contents |\n| `vector.RemoveOverlap` | De-duplicate overlapping chunks in results |\n\nQuery nodes (`QueryText`, `QueryImage`, `HybridSearch`) output `ids`,\n`documents`, `metadatas`, and `distances` (HybridSearch also returns `scores`).\n\n# Document Loading & Splitting\n\n| Node | Namespace | Purpose |\n|------|-----------|---------|\n| `Code` | `nodetool.code` | Enumerate files with `await workspace.list(dir)` |\n| `LoadDocumentFile` | `nodetool.document` | Load a PDF\u002FTXT\u002FMD into a document |\n| `Chunk` | `nodetool.text` | Fixed-size word chunking with overlap (general purpose) |\n| `RegexSplit` | `nodetool.text` | Structure-aware splitting on a delimiter pattern |\n\n## Chunk Size Guidance\n\n| Content type | Chunk size | Overlap |\n|--------------|-----------|---------|\n| Technical docs | 200-500 tokens | 50 tokens |\n| Prose\u002Farticles | 300-600 tokens | 75 tokens |\n| Code | 100-300 tokens | 25 tokens |\n\n# Indexing — Workflow Pattern\n\n```\nListFiles → LoadDocumentFile → Chunk → IndexTextChunk(collection)\n```\n\nPair every index\u002Fquery node with a `vector.Collection` node (or a collection\nname) so they target the same store. Use the **same embedding model** for\nindexing and querying.\n\n# Indexing — HTTP API\n\n```bash\n# Index a file into a collection\ncurl -X POST http:\u002F\u002Flocalhost:7777\u002Fapi\u002Fcollections\u002F\u003Cname>\u002Findex \\\n  -H \"Authorization: Bearer TOKEN\" \\\n  -H \"Content-Type: application\u002Fjson\" \\\n  -d '{\"file_path\": \"\u002Fpath\u002Fto\u002Fdocument.pdf\"}'\n```\n\nThe server resolves the collection, runs its ingestion workflow if one is\nregistered, otherwise falls back to split → embed → store.\n\n# Query — Workflow Pattern\n\n```\nChatInput → HybridSearch(collection, top_k) → FormatText → Agent → Output\n```\n\n| Node | Purpose |\n|------|---------|\n| `ChatInput` | User question |\n| `vector.HybridSearch` | Vector + keyword retrieval (best accuracy) |\n| `vector.QueryText` | Vector-only retrieval (faster) |\n| `FormatText` | Build the context string for the LLM |\n| `Agent` | Generate the answer from context + question |\n| `Output` | Return the answer |\n\n# Complete RAG Example\n\n## Index\n```\nListFiles(\"\u002Fdocs\u002F\") → LoadDocumentFile → Chunk(length=400, overlap=50)\n                                                  ↓\n                                  IndexTextChunk(collection=\"my-docs\")\n```\n\n## Query\n```\nChatInput(\"What is...?\") → HybridSearch(collection=\"my-docs\", top_k=5)\n                                  ↓\n                         FormatText(template=\"Context:\\n{documents}\\n\\nQuestion: {query}\")\n                                  ↓\n                         Agent(model=gpt-5.4, system=\"Answer using only the context provided.\")\n                                  ↓\n                         Output\n```\n\n# Environment Variables\n\n```bash\n# ChromaDB backend (only when using Chroma — SQLite-vec needs no config)\nCHROMA_URL=                          # Remote Chroma URL (empty = local)\nCHROMA_PATH=~\u002F.local\u002Fshare\u002Fnodetool\u002Fchroma  # Local storage path\nCHROMA_TOKEN=                        # Optional auth token\n```\n\nThe embedding model is chosen on the index\u002Fquery nodes via model selection\n(e.g. `text-embedding-3-small`, or a local sentence-transformers model).\n\n# Common Pitfalls\n\n- **Embedding model mismatch**: use the same embedding model for indexing and search.\n- **Chunks too large**: dilute the LLM context — keep to 200-500 tokens.\n- **Chunks too small**: sentences get fragmented; use 10-20% overlap.\n- **Empty collection**: index before querying — an unindexed collection returns nothing.\n- **Wrong node names**: it's `vector.IndexTextChunk` \u002F `vector.QueryText` \u002F `vector.HybridSearch`, not `IndexTextChunks` \u002F `TextSearch`, and there is no `vector.chroma.*`\u002F`vector.faiss.*` namespace.\n- **No `nodetool collections` CLI**: manage collections through the editor UI or the `\u002Fapi\u002Fcollections\u002F...` endpoints.\n",{"name":5,"description":7},"main","4ef72a828b24c99c49109033c7ec504c34f3bd17",".claude\u002Fskills\u002Fnodetool-rag-indexing",0,"https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas","https:\u002F\u002Fgithub.com\u002Fhanchenhao","2026-08-18 16:41:12",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add hanchenhao\u002Fai-canvas --skill nodetool-rag-indexing",{"tool":36,"command":37},"bunx","bunx skills add hanchenhao\u002Fai-canvas --skill nodetool-rag-indexing",{"tool":39,"command":40},"pnpm","pnpm dlx skills add hanchenhao\u002Fai-canvas --skill nodetool-rag-indexing","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fnodetool-rag-indexing\u002FSKILL.md",true,[],[45,52,59,66,73,80],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M7AE5NP5AWGV47WS3F37VPZ1","sandbox-zip","Read and write zip archives in a Code node or CodeAct action, with fflate running on the host under a 50 MB inflation cap","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-zip\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M7AE5JP0CG2S114Q246RC6MH","sandbox-yaml","Read and write YAML inside a Code node or CodeAct action, with js-yaml running in the guest","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-yaml\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":18},"01M7AE5GJ9AV36WM1DPW1878VM","sandbox-xml","Parse XML feeds and sitemaps in a Code node or CodeAct action, with fast-xml-parser running on the host","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-xml\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M7AE5DMYDAGJW27G8KKTZ2BN","sandbox-xlsx","Read and write Excel workbooks in a Code node or CodeAct action, with exceljs running on the host","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-xlsx\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":18},"01M7AE5BARJSFCT296SEMYDSJQ","sandbox-twilio","Send SMS and WhatsApp messages and look up numbers from a Code node or CodeAct action, with NodeTool's Twilio request builder running on the host","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-twilio\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M7AE57C5APRDF96NB3G0WQMW","sandbox-tfjs","Classify and embed images, detect objects, and answer from a passage in a Code node or CodeAct action, with TensorFlow.js running on the host","https:\u002F\u002Fgithub.com\u002Fhanchenhao\u002Fai-canvas\u002Fblob\u002Fmain\u002Fpackages\u002Fsandbox-packs\u002Fsandbox-tfjs\u002FSKILL.md",[],[]]