Skip to content

History of mcp-local-rag in namespace io.github

io.github.shinpr/mcp-local-rag · Registry status: active

No name of its own reaches this register for this entry. The heading therefore carries the trailing segment of the identifier; the identifier itself is listed below as a coordinate.

Report data on this entry

Every published change to this entry, most recent first.

Changes
2026-10-11Declared version: 0.21.1 → 0.21.3
2026-10-11Last changed in the registry (raw): 2026-10-07 → 2026-10-10
2026-10-11First listed in the registry (raw): 2026-10-07 → 2026-10-10
2026-10-11Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.21.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"EMBED_HEADING_PREFIX","description":"Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.21.3","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"EMBED_HEADING_PREFIX","description":"Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-10-11Package versions (raw): 0.21.1 → 0.21.3
2026-10-08Declared version: 0.21.0 → 0.21.1
2026-10-08Last changed in the registry (raw): 2026-10-03 → 2026-10-07
2026-10-08First listed in the registry (raw): 2026-10-03 → 2026-10-07
2026-10-08Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.21.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"EMBED_HEADING_PREFIX","description":"Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.21.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"EMBED_HEADING_PREFIX","description":"Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-10-08Package versions (raw): 0.21.0 → 0.21.1
2026-10-04Declared version: 0.20.1 → 0.21.0
2026-10-04Last changed in the registry (raw): 2026-10-02 → 2026-10-03
2026-10-04First listed in the registry (raw): 2026-10-02 → 2026-10-03
2026-10-04Environment variable descriptions (raw): BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · EMBED_TITLE_PREFIX=Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest. · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false). → BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · EMBED_HEADING_PREFIX=Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it. · EMBED_TITLE_PREFIX=Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest. · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).
2026-10-04Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.20.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.21.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"EMBED_HEADING_PREFIX","description":"Add section headings to chunk embeddings when they fit (defaults to false). Independent of EMBED_TITLE_PREFIX. Re-ingest documents after changing it.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-10-04Package versions (raw): 0.20.1 → 0.21.0
2026-10-04Environment variables (raw): BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, EMBED_TITLE_PREFIX, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT, RAG_RERANK_CMD, RAG_RERANK_TIMEOUT_MS → BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, EMBED_TITLE_PREFIX, EMBED_HEADING_PREFIX, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT, RAG_RERANK_CMD, RAG_RERANK_TIMEOUT_MS
2026-10-03Declared version: 0.20.0 → 0.20.1
2026-10-03Last changed in the registry (raw): 2026-10-01 → 2026-10-02
2026-10-03First listed in the registry (raw): 2026-10-01 → 2026-10-02
2026-10-03Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.20.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.20.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-10-03Package versions (raw): 0.20.0 → 0.20.1
2026-10-02Declared version: 0.19.1 → 0.20.0
2026-10-02Last changed in the registry (raw): 2026-09-22 → 2026-10-01
2026-10-02First listed in the registry (raw): 2026-09-22 → 2026-10-01
2026-10-02Environment variable descriptions (raw): BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false). → BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · EMBED_TITLE_PREFIX=Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest. · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).
2026-10-02Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.19.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.20.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"EMBED_TITLE_PREFIX","description":"Embed each chunk together with its document title, which can help when passages don't restate the topic the title names (defaults to false). After changing it, use a new DB_PATH or delete the index and re-ingest.","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-10-02Package versions (raw): 0.19.1 → 0.20.0
2026-10-02Environment variables (raw): BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT, RAG_RERANK_CMD, RAG_RERANK_TIMEOUT_MS → BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, EMBED_TITLE_PREFIX, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT, RAG_RERANK_CMD, RAG_RERANK_TIMEOUT_MS
2026-09-23Declared version: 0.19.0 → 0.19.1
2026-09-23Last changed in the registry (raw): 2026-09-20 → 2026-09-22
2026-09-23First listed in the registry (raw): 2026-09-20 → 2026-09-22
2026-09-23Environment variable descriptions (raw): BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External command that reorders query_documents results (unset disables reranking). The search query and the matched document chunks, including their text, are passed to this command, so a command that contacts a remote service sends that content off this machine. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false). → BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).
2026-09-23Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.19.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External command that reorders query_documents results (unset disables reranking). The search query and the matched document chunks, including their text, are passed to this command, so a command that contacts a remote service sends that content off this machine.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.19.1","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External reranker command template. Use {query} and {top} for query text and result count; unset disables reranking.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-09-23Package versions (raw): 0.19.0 → 0.19.1
2026-09-21Declared version: 0.18.6 → 0.19.0
2026-09-21Last changed in the registry (raw): 2026-09-18 → 2026-09-20
2026-09-21First listed in the registry (raw): 2026-09-18 → 2026-09-20
2026-09-21Environment variable descriptions (raw): BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false). → BASE_DIR=Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set. · BASE_DIRS=JSON array of base directories (e.g. '["/a","/b"]'). Takes precedence over BASE_DIR. · CACHE_DIR=Directory where Transformers.js models are cached (defaults to ./models/) · CHUNK_MIN_LENGTH=Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion. · DB_PATH=Path to LanceDB database directory (defaults to ./lancedb/) · HF_ENDPOINT=Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co). · MAX_FILE_SIZE=Maximum file size in bytes (defaults to 104857600 / 100MB) · MODEL_NAME=Embedding model name (defaults to Xenova/all-MiniLM-L6-v2) · RAG_DEVICE=Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error. · RAG_DTYPE=Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data. · RAG_GROUPING=Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter · RAG_HYBRID_WEIGHT=Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score. · RAG_MAX_DISTANCE=Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only) · RAG_MAX_FILES=Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering. · RAG_RERANK_CMD=External command that reorders query_documents results (unset disables reranking). The search query and the matched document chunks, including their text, are passed to this command, so a command that contacts a remote service sends that content off this machine. · RAG_RERANK_TIMEOUT_MS=Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned. · STORE_IMAGES=Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).
2026-09-21Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.18.6","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.19.0","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_CMD","description":"External command that reorders query_documents results (unset disables reranking). The search query and the matched document chunks, including their text, are passed to this command, so a command that contacts a remote service sends that content off this machine.","format":"string","required":false,"secret":false},{"name":"RAG_RERANK_TIMEOUT_MS","description":"Time budget per rerank call in milliseconds (100-600000, defaults to 10000). On timeout the spawned command is killed and the pre-rerank ordering is returned.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-09-21Package versions (raw): 0.18.6 → 0.19.0
2026-09-21Environment variables (raw): BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT → BASE_DIR, BASE_DIRS, DB_PATH, CACHE_DIR, HF_ENDPOINT, MODEL_NAME, MAX_FILE_SIZE, RAG_MAX_DISTANCE, RAG_GROUPING, RAG_MAX_FILES, CHUNK_MIN_LENGTH, STORE_IMAGES, RAG_DEVICE, RAG_DTYPE, RAG_HYBRID_WEIGHT, RAG_RERANK_CMD, RAG_RERANK_TIMEOUT_MS
2026-09-19Declared version: 0.18.5 → 0.18.6
2026-09-19Last changed in the registry (raw): 2026-09-15 → 2026-09-18
2026-09-19First listed in the registry (raw): 2026-09-15 → 2026-09-18
2026-09-19Connection paths (source structure): {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.18.5","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]} → {"packages":[{"registryType":"npm","identifier":"mcp-local-rag","version":"0.18.6","transport":"stdio","environment":[{"name":"BASE_DIR","description":"Base directory for document storage (defaults to current working directory). Ignored when BASE_DIRS is set.","format":"string","required":false,"secret":false},{"name":"BASE_DIRS","description":"JSON array of base directories (e.g. '[\"/a\",\"/b\"]'). Takes precedence over BASE_DIR.","format":"string","required":false,"secret":false},{"name":"DB_PATH","description":"Path to LanceDB database directory (defaults to ./lancedb/)","format":"string","required":false,"secret":false},{"name":"CACHE_DIR","description":"Directory where Transformers.js models are cached (defaults to ./models/)","format":"string","required":false,"secret":false},{"name":"HF_ENDPOINT","description":"Hugging Face model download endpoint. Set this to a mirror URL when direct downloads are blocked (defaults to https://huggingface.co).","format":"string","required":false,"secret":false},{"name":"MODEL_NAME","description":"Embedding model name (defaults to Xenova/all-MiniLM-L6-v2)","format":"string","required":false,"secret":false},{"name":"MAX_FILE_SIZE","description":"Maximum file size in bytes (defaults to 104857600 / 100MB)","format":"string","required":false,"secret":false},{"name":"RAG_MAX_DISTANCE","description":"Maximum distance threshold for filtering search results. Results with distance greater than this value will be excluded. Lower values mean stricter filtering (e.g., 0.5 for high relevance only)","format":"string","required":false,"secret":false},{"name":"RAG_GROUPING","description":"Grouping mode for quality filtering. 'similar' returns only the most similar group (stops at first distance jump). 'related' includes related groups (stops at second distance jump). Unset means no grouping filter","format":"string","required":false,"secret":false},{"name":"RAG_MAX_FILES","description":"Maximum number of files to keep in search results. Results are filtered to include only chunks from the top N best-scoring files. For example, 1 returns only the single best-matching file's chunks. Unset means no file filtering.","format":"string","required":false,"secret":false},{"name":"CHUNK_MIN_LENGTH","description":"Minimum chunk length in characters (1-10000, defaults to 50). Chunks shorter than this threshold are filtered out during ingestion.","format":"string","required":false,"secret":false},{"name":"STORE_IMAGES","description":"Store supported PDF and DOCX images during ingestion and return them with matched chunks (defaults to false).","format":"string","required":false,"secret":false},{"name":"RAG_DEVICE","description":"Execution device for the embedder (defaults to cpu). Passed straight to ONNX Runtime; see the Transformers.js device source for the supported backend names. If the requested device fails to initialize, the server throws an error.","format":"string","required":false,"secret":false},{"name":"RAG_DTYPE","description":"Embedding quantization dtype for the embedder (defaults to fp32). Opt-in and pass-through; accepts any dtype the chosen model provides (fp32, fp16, q8, int8, ...). If the model has no variant for the requested dtype, the server throws an error. Changing this changes the embedding space — re-ingest existing data.","format":"string","required":false,"secret":false},{"name":"RAG_HYBRID_WEIGHT","description":"Keyword boost factor for hybrid search (0.0-1.0, defaults to 0.6). 0 means semantic similarity only; higher values increase the keyword-match contribution to the final score.","format":"string","required":false,"secret":false}],"additional_arguments_declared":false}],"remotes":[]}
2026-09-19Package versions (raw): 0.18.5 → 0.18.6
2026-09-16Declared version: 0.18.4 → 0.18.5
2026-09-16Last changed in the registry (raw): 2026-09-07 → 2026-09-15
2026-09-16First listed in the registry (raw): 2026-09-07 → 2026-09-15

This view pages forward and states no total: individual rows are dropped only at output time when their evidence is missing. A counted total would be larger than what is shown.

tracevero · https://tracevero.com/mcp/io-github-shinpr-mcp-local-rag/aenderungen