rag-multiple-files 0.1.0__tar.gz

rag_multiple_files-0.1.0/.gitignore

@@ -0,0 +1,29 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# Distribution / packaging
+dist/
+build/
+*.egg-info/
+*.egg
+
+# Virtual environments
+.venv/
+venv/
+ENV/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# Environment variables
+.env
+.env.*
+
+# OS
+.DS_Store
+Thumbs.db

rag_multiple_files-0.1.0/BUILD.md

@@ -0,0 +1,6 @@
+# Build and Publish
+
+```bash
+uv build
+uv publish --token xxxxxxxxxxxxxxx
+```

rag_multiple_files-0.1.0/PKG-INFO

@@ -0,0 +1,61 @@
+Metadata-Version: 2.4
+Name: rag-multiple-files
+Version: 0.1.0
+Summary: MCP server for semantic PDF querying using RAG
+Project-URL: Homepage, https://codeberg.org/mionita1980/rag-multiple-files
+Project-URL: Source, https://codeberg.org/mionita1980/rag-multiple-files
+Author-email: Mihai Ionita <mionita1980@proton.me>
+Requires-Python: >=3.11
+Requires-Dist: faiss-cpu
+Requires-Dist: mcp[cli]
+Requires-Dist: sentence-transformers
+Description-Content-Type: text/markdown
+
+# RAG Multiple Files MCP Server
+
+A local MCP server that lets you semantically search multiple PDFs using natural language queries.
+
+## Use from Claude Code
+
+Configure the MCP server:
+
+```json
+"mcpServers": {
+    ...
+    "rag-multiple-files": {
+        "type": "stdio",
+        "command": "uvx",
+        "args": [
+            "rag-multiple-files"
+        ],
+        "env": {
+            "PDF_FOLDER": "/path/to/your/folder"
+        }
+    }
+}
+```
+
+### Sample Request
+
+Query:
+
+```
+YYYYYYYYYYYYYYY.pdf: What is the chain slack
+```
+
+Response:
+
+```
+Based on the YYYYYYYYYYYYYYY owner's manual:
+    Chain tension (slack): XX mm (ZZ in)
+```
+
+## How it works
+
+1. **Querying** — The query is embedded with the same model and searched against the FAISS index. The top-k most similar chunks are returned with their page numbers and similarity scores.
+
+Parameters:
+
+- query (str) - required - no default - the question or topic to search for
+- file_name (str) - required - no default - the name of the PDF file
+- top_k (int) - not required - 5 default - number of chunks to return

rag_multiple_files-0.1.0/README.md

@@ -0,0 +1,48 @@
+# RAG Multiple Files MCP Server
+
+A local MCP server that lets you semantically search multiple PDFs using natural language queries.
+
+## Use from Claude Code
+
+Configure the MCP server:
+
+```json
+"mcpServers": {
+    ...
+    "rag-multiple-files": {
+        "type": "stdio",
+        "command": "uvx",
+        "args": [
+            "rag-multiple-files"
+        ],
+        "env": {
+            "PDF_FOLDER": "/path/to/your/folder"
+        }
+    }
+}
+```
+
+### Sample Request
+
+Query:
+
+```
+YYYYYYYYYYYYYYY.pdf: What is the chain slack
+```
+
+Response:
+
+```
+Based on the YYYYYYYYYYYYYYY owner's manual:
+    Chain tension (slack): XX mm (ZZ in)
+```
+
+## How it works
+
+1. **Querying** — The query is embedded with the same model and searched against the FAISS index. The top-k most similar chunks are returned with their page numbers and similarity scores.
+
+Parameters:
+
+- query (str) - required - no default - the question or topic to search for
+- file_name (str) - required - no default - the name of the PDF file
+- top_k (int) - not required - 5 default - number of chunks to return

rag_multiple_files-0.1.0/pyproject.toml

@@ -0,0 +1,28 @@
+[project]
+name = "rag-multiple-files"
+version = "0.1.0"
+description = "MCP server for semantic PDF querying using RAG"
+authors = [
+    { name = "Mihai Ionita", email = "mionita1980@proton.me" }
+]
+readme = "README.md"
+requires-python = ">=3.11"
+dependencies = [
+    "mcp[cli]",
+    "sentence-transformers",
+    "faiss-cpu"
+]
+
+[project.urls]
+Homepage = "https://codeberg.org/mionita1980/rag-multiple-files"
+Source = "https://codeberg.org/mionita1980/rag-multiple-files"
+
+[project.scripts]
+rag-multiple-files = "ragmfiles.main:main"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/ragmfiles"]

rag_multiple_files-0.1.0/requirements.txt

@@ -0,0 +1,3 @@
+faiss-cpu==1.13.2
+mcp==1.26.0
+sentence-transformers==5.3.0

rag_multiple_files-0.1.0/src/ragmfiles/__init__.py

@@ -0,0 +1 @@
+"""Multiple files RAG MCP Server - Query info from multiple documents."""

rag_multiple_files-0.1.0/src/ragmfiles/main.py

@@ -0,0 +1,116 @@
+"""MCP server that answers queries against pre-built PDF indexes.
+
+Indexes must already exist (run the indexer first for each PDF).
+"""
+
+import json
+import os
+from pathlib import Path
+import faiss
+import numpy as np
+from mcp.server.fastmcp import FastMCP
+from sentence_transformers import SentenceTransformer
+import logging
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+MODEL_NAME = "all-MiniLM-L6-v2"
+
+mcp = FastMCP("rag-multiple-files")
+
+_model: SentenceTransformer | None = None
+# Maps pdf filename (e.g. "doc.pdf") -> (faiss.Index, list[dict])
+_indexes: dict[str, tuple[faiss.Index, list[dict]]] = {}
+
+
+def get_model() -> SentenceTransformer:
+    global _model
+    if _model is None:
+        _model = SentenceTransformer(MODEL_NAME)
+    return _model
+
+
+def get_pdf_folder() -> Path:
+    folder = os.environ.get("PDF_FOLDER", "")
+    if not folder:
+        raise ValueError("PDF_FOLDER environment variable is not set")
+    path = Path(folder)
+    if not path.is_dir():
+        raise FileNotFoundError(f"Folder not found: {path}")
+    return path
+
+
+def load_all_indexes() -> None:
+    """Scan PDF_FOLDER for *.index subdirs and load each one."""
+    folder = get_pdf_folder()
+    for index_dir in sorted(folder.glob("*.index")):
+        faiss_file = index_dir / "index.faiss"
+        chunks_file = index_dir / "chunks.json"
+        if not faiss_file.exists() or not chunks_file.exists():
+            logger.warning(f"Skipping incomplete index at {index_dir}")
+            continue
+        # Reconstruct the original PDF filename from the index dir name
+        pdf_name = index_dir.stem + ".pdf"
+        if pdf_name in _indexes:
+            continue
+        logger.info(f"Loading index for {pdf_name}...")
+        index = faiss.read_index(str(faiss_file))
+        chunks = json.loads(chunks_file.read_text())
+        _indexes[pdf_name] = (index, chunks)
+    if not _indexes:
+        logger.warning("No indexes found in PDF_FOLDER. Run the indexer first.")
+
+
+@mcp.tool()
+def list_indexed_files() -> str:
+    """List all PDF files that have a loaded index available for querying."""
+    if not _indexes:
+        return "No indexes loaded."
+    return "\n".join(sorted(_indexes.keys()))
+
+
+@mcp.tool()
+def query_pdf(query: str, file_name: str, top_k: int = 5) -> str:
+    """Search the indexed PDF for chunks most relevant to the query.
+
+    Args:
+        query: The question or topic to search for.
+        file_name: The PDF filename to query (e.g. "document.pdf").
+        top_k: Number of chunks to return (default: 5).
+    """
+    logger.info(f"query_pdf file={file_name} query={query!r} top_k={top_k}")
+    if file_name not in _indexes:
+        available = ", ".join(sorted(_indexes.keys())) or "none"
+        return f"No index loaded for '{file_name}'. Available: {available}"
+
+    index, chunks = _indexes[file_name]
+    model = get_model()
+    query_vec = model.encode([query], normalize_embeddings=True)
+    query_vec = np.array(query_vec, dtype=np.float32)
+    scores, indices = index.search(query_vec, min(top_k, len(chunks)))
+    results = []
+    for score, idx in zip(scores[0], indices[0]):
+        if idx == -1:
+            continue
+        chunk = chunks[idx]
+        results.append(
+            f"[Page {chunk['page']}] (score: {score:.3f})\n{chunk['text']}"
+        )
+    if not results:
+        return "No matching chunks found."
+    return (
+        f"Found {len(results)} relevant chunks from: {file_name}\n\n"
+        + "\n\n---\n\n".join(results)
+    )
+
+
+def main():
+    """Entry point for the RAG MCP server."""
+    logger.info("Starting MCP server...")
+    load_all_indexes()
+    mcp.run(transport="stdio")
+
+
+if __name__ == "__main__":
+    main()