Fix: Add missing langchain dependencies for text processing

🐛 Problem:
- Backend startup fails with: No module named 'langchain_community'
- text_chunking_strategies.py imports langchain modules
- langchain/langchain-community not in pyproject.toml dependencies

✅ Solution:
Added to dependencies:
- langchain>=0.3.0 (for RecursiveCharacterTextSplitter)
- langchain-community>=0.3.0 (for JSONLoader)

🔧 Used By:
- agentgraph/input/text_processing/text_chunking_strategies.py
- backend/services/processing_service.py
- backend/routers/traces.py
- agentgraph/extraction/graph_processing/knowledge_graph_processor.py

🎯 Result:
- Backend should start successfully
- Text chunking functionality available
- No more import errors on startup

pyproject.toml

@@ -43,6 +43,8 @@ dependencies = [
     "litellm>=1.60.0",           # ✅ 核心
     "langfuse>=3.0.0",           # ✅ 核心 (15+ 使用)
     "langsmith>=0.3.38",         # ✅ 核心 (12+ 使用)
+    "langchain>=0.3.0",          # ✅ 核心 (text splitting)
+    "langchain-community>=0.3.0", # ✅ 核心 (document loaders)
     "crewai>=0.108.0",           # ✅ 功能性 (2+ 使用)
     "crewai-tools>=0.38.1",      # ✅ 功能性
     "langgraph>=0.5.3",          # ✅ 功能性