knowledge-rag 3.4.2__tar.gz → 3.4.3__tar.gz

{knowledge_rag-3.4.2 → knowledge_rag-3.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: knowledge-rag
-Version: 3.4.2
+Version: 3.4.3
 Summary: Local RAG System for Claude Code — Hybrid search + Cross-encoder Reranking + 12 MCP Tools. Zero external servers.
 Project-URL: Homepage, https://github.com/lyonzin/knowledge-rag
 Project-URL: Repository, https://github.com/lyonzin/knowledge-rag
@@ -38,7 +38,7 @@ Description-Content-Type: text/markdown
 
 <div align="center">
 
-![Version](https://img.shields.io/badge/version-3.4.2-blue.svg)
+![Version](https://img.shields.io/badge/version-3.4.3-blue.svg)
 ![Python](https://img.shields.io/badge/python-3.11%20%7C%203.12-green.svg)
 ![License](https://img.shields.io/badge/license-MIT-yellow.svg)
 ![Platform](https://img.shields.io/badge/platform-Windows%20%7C%20Linux%20%7C%20macOS-lightgrey.svg)
@@ -61,7 +61,7 @@ Your documents become instantly searchable inside Claude Code — with reranking
 
 **12 MCP Tools** | **Hybrid Search + Cross-Encoder Reranking** | **Markdown-Aware Chunking** | **100% Local, Zero Cloud**
 
-[What's New](#whats-new-in-v342) | [Installation](#installation) | [Configuration](#configuration) | [API Reference](#api-reference) | [Architecture](#architecture)
+[What's New](#whats-new-in-v343) | [Installation](#installation) | [Configuration](#configuration) | [API Reference](#api-reference) | [Architecture](#architecture)
 
 </div>
 
@@ -108,9 +108,9 @@ After the initial rebuild, startup and queries are faster than v2.x because ther
 
 ---
 
-## What's New in v3.4.2
+## What's New in v3.4.3
 
-**MCP connection fix** — Stdout redirect now happens at module load (before any import), preventing `[INFO]` messages from corrupting the MCP stdio JSON-RPC stream. This was the root cause of "Failed to connect" on fresh installs.
+**MCP connection fix (correct)** — Stdout redirect now uses save/restore pattern: `__init__.py` saves real stdout and redirects print() to stderr during module load (protecting against `[INFO]` pollution from config/embedding imports), then `server.py main()` restores the original stdout right before `mcp.run()` so JSON-RPC responses flow correctly. v3.4.2 broke the response channel with a global redirect.
 
 **Reliable pip install** (v3.4.1) — Auto-detects project directory from venv location. No `cwd`/`cd /d` workarounds needed.
 
@@ -1238,9 +1238,13 @@ With ~200 documents, expect ~300-500MB RAM. The embedding model (~50MB) and rera
 
 ## Changelog
 
-### v3.4.2 (2026-04-16)
+### v3.4.3 (2026-04-16)
 
-- **FIX**: Stdout redirect moved to module-level (before imports) — fixes "Failed to connect" MCP error caused by `[INFO]` messages polluting the JSON-RPC stdio stream during server initialization
+- **FIX**: Correct stdout protection via save/restore pattern — `__init__.py` saves original stdout and redirects to stderr during init, `server.py main()` restores it before `mcp.run()`. v3.4.2's global redirect broke MCP JSON-RPC response channel.
+
+### v3.4.2 (2026-04-16) — BROKEN, use v3.4.3
+
+- **FIX**: Stdout redirect moved to module-level (before imports) — fixes "Failed to connect" MCP error caused by `[INFO]` messages polluting the JSON-RPC stdio stream during server initialization. **BUG**: Global redirect also blocked JSON-RPC responses.
 
 ### v3.4.1 (2026-04-16)
 

{knowledge_rag-3.4.2 → knowledge_rag-3.4.3}/README.md

@@ -2,7 +2,7 @@
 
 <div align="center">
 
-![Version](https://img.shields.io/badge/version-3.4.2-blue.svg)
+![Version](https://img.shields.io/badge/version-3.4.3-blue.svg)
 ![Python](https://img.shields.io/badge/python-3.11%20%7C%203.12-green.svg)
 ![License](https://img.shields.io/badge/license-MIT-yellow.svg)
 ![Platform](https://img.shields.io/badge/platform-Windows%20%7C%20Linux%20%7C%20macOS-lightgrey.svg)
@@ -25,7 +25,7 @@ Your documents become instantly searchable inside Claude Code — with reranking
 
 **12 MCP Tools** | **Hybrid Search + Cross-Encoder Reranking** | **Markdown-Aware Chunking** | **100% Local, Zero Cloud**
 
-[What's New](#whats-new-in-v342) | [Installation](#installation) | [Configuration](#configuration) | [API Reference](#api-reference) | [Architecture](#architecture)
+[What's New](#whats-new-in-v343) | [Installation](#installation) | [Configuration](#configuration) | [API Reference](#api-reference) | [Architecture](#architecture)
 
 </div>
 
@@ -72,9 +72,9 @@ After the initial rebuild, startup and queries are faster than v2.x because ther
 
 ---
 
-## What's New in v3.4.2
+## What's New in v3.4.3
 
-**MCP connection fix** — Stdout redirect now happens at module load (before any import), preventing `[INFO]` messages from corrupting the MCP stdio JSON-RPC stream. This was the root cause of "Failed to connect" on fresh installs.
+**MCP connection fix (correct)** — Stdout redirect now uses save/restore pattern: `__init__.py` saves real stdout and redirects print() to stderr during module load (protecting against `[INFO]` pollution from config/embedding imports), then `server.py main()` restores the original stdout right before `mcp.run()` so JSON-RPC responses flow correctly. v3.4.2 broke the response channel with a global redirect.
 
 **Reliable pip install** (v3.4.1) — Auto-detects project directory from venv location. No `cwd`/`cd /d` workarounds needed.
 
@@ -1202,9 +1202,13 @@ With ~200 documents, expect ~300-500MB RAM. The embedding model (~50MB) and rera
 
 ## Changelog
 
-### v3.4.2 (2026-04-16)
+### v3.4.3 (2026-04-16)
 
-- **FIX**: Stdout redirect moved to module-level (before imports) — fixes "Failed to connect" MCP error caused by `[INFO]` messages polluting the JSON-RPC stdio stream during server initialization
+- **FIX**: Correct stdout protection via save/restore pattern — `__init__.py` saves original stdout and redirects to stderr during init, `server.py main()` restores it before `mcp.run()`. v3.4.2's global redirect broke MCP JSON-RPC response channel.
+
+### v3.4.2 (2026-04-16) — BROKEN, use v3.4.3
+
+- **FIX**: Stdout redirect moved to module-level (before imports) — fixes "Failed to connect" MCP error caused by `[INFO]` messages polluting the JSON-RPC stdio stream during server initialization. **BUG**: Global redirect also blocked JSON-RPC responses.
 
 ### v3.4.1 (2026-04-16)
 

knowledge_rag-3.4.3/mcp_server/__init__.py

@@ -0,0 +1,17 @@
+"""Knowledge RAG MCP Server - Local Retrieval-Augmented Generation System"""
+
+import sys  # noqa: I001
+
+# MCP stdio uses stdout for JSON-RPC. print() on stdout corrupts the stream.
+# Save the real stdout for MCP, redirect print() to stderr during init.
+# server.py main() restores _original_stdout before mcp.run().
+_original_stdout = sys.stdout
+sys.stdout = sys.stderr
+
+__version__ = "3.4.3"
+__author__ = "Ailton Rocha (Lyon.)"
+
+from .config import Config  # noqa: E402
+from .ingestion import Document, DocumentParser  # noqa: E402
+
+__all__ = ["Config", "DocumentParser", "Document"]

{knowledge_rag-3.4.2 → knowledge_rag-3.4.3}/mcp_server/server.py

@@ -19,20 +19,14 @@ Features:
     - CRUD operations via MCP tools (add, update, remove docs)
 
 Autor:   Lyon (Ailton Rocha)
-Versao:  3.4.2
+Versao:  3.4.3
 Data:    2026-04-16
 """
 
-import sys
-
-# Redirect stdout to stderr IMMEDIATELY — before any import can print().
-# MCP stdio uses stdout for JSON-RPC; stray print() corrupts the stream.
-# This must happen before importing config, chromadb, fastembed, etc.
-sys.stdout = sys.stderr
-
 import hashlib
 import json
 import re
+import sys
 import threading
 import time
 from collections import OrderedDict
@@ -1878,8 +1872,6 @@ def _handle_init():
 
 def main():
     """Run the MCP server"""
-    import sys
-
     if len(sys.argv) > 1 and sys.argv[1] == "init":
         _handle_init()
         return
@@ -1917,6 +1909,10 @@ def main():
         print(f"[WARN] Failed to start file watcher: {e}")
         print("[WARN] Auto-reindexing disabled. Use reindex_documents tool manually.")
 
+    # Restore real stdout for MCP JSON-RPC, keep print() going to stderr
+    from . import _original_stdout
+
+    sys.stdout = _original_stdout
     mcp.run()
 
 

{knowledge_rag-3.4.2 → knowledge_rag-3.4.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "knowledge-rag"
-version = "3.4.2"
+version = "3.4.3"
 description = "Local RAG System for Claude Code — Hybrid search + Cross-encoder Reranking + 12 MCP Tools. Zero external servers."
 readme = "README.md"
 license = {text = "MIT"}

knowledge_rag-3.4.2/mcp_server/__init__.py

@@ -1,9 +0,0 @@
-"""Knowledge RAG MCP Server - Local Retrieval-Augmented Generation System"""
-
-__version__ = "3.4.2"
-__author__ = "Ailton Rocha (Lyon.)"
-
-from .config import Config
-from .ingestion import Document, DocumentParser
-
-__all__ = ["Config", "DocumentParser", "Document"]