feat: comprehensive v2 upgrade — streaming, error KB, file upload, layout analysis

Major changes:
- Streaming: LLM统一 _BaseLLM 接口 (invoke + stream), generate/modify/correct
  节点使用 get_stream_writer() 实现逐字输出, UI 节点平铺展开自动折叠
- Prompt外部化: 7个prompt拆分到 prompts/*.md, loader.py 支持热重载
- 错误自增长: backend/error_kb.py — 指纹去重 + ChromaDB持久化,
  correct_jrxml→validate 通过时自动入库, retrieve同时搜索错误KB
- 文件上传: backend/file_parser.py — PDF/DOCX/图片/文本解析,
  侧边栏多文件上传, 文本自动注入下一条消息
- A4模板识别: backend/layout_analyzer.py — 三种模式(完整A4/行片段修改/行片段新建),
  PaddleOCR元素提取 + 行分组 + JRXML section匹配
- 会话历史下载: jrxml_versions版本追踪 + 侧边栏历史版本下载按钮
- 预览修复: route_after_save跳过预览/导出意图的验证循环
- Ctrl+C修复: JS注入拦截Streamlit裸c键清缓存

Docs: CLAUDE.md (完整项目文档), ROADMAP.md (改进路线图)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

backend/llm.py

@@ -8,13 +8,33 @@ from dotenv import load_dotenv
 load_dotenv()
 
 
+class _BaseLLM:
+    """LLM 统一接口基类 — 所有后端都提供 invoke() 和 stream()。"""
+
+    def invoke(self, prompt: str) -> Any:
+        raise NotImplementedError
+
+    def stream(self, prompt: str):
+        raise NotImplementedError
+
+
 def get_llm():
     backend = os.getenv("LLM_BACKEND", "cloud")
     if backend == "local":
         from langchain_ollama import ChatOllama
 
         model = os.getenv("LOCAL_LLM_MODEL", "qwen2.5-coder:7b")
-        return ChatOllama(model=model, temperature=0.1)
+        raw = ChatOllama(model=model, temperature=0.1)
+
+        class OllamaWrapper(_BaseLLM):
+            def invoke(self, prompt):
+                return raw.invoke(prompt)
+
+            def stream(self, prompt):
+                for chunk in raw.stream(prompt):
+                    yield chunk.content
+
+        return OllamaWrapper()
 
     provider = os.getenv("LLM_PROVIDER", "openai")
     if provider == "anthropic":
@@ -30,7 +50,7 @@ def get_llm():
 
         client = Anthropic(api_key=api_key, base_url=base_url, timeout=120)
 
-        class MiniMaxLLM:
+        class MiniMaxLLM(_BaseLLM):
             def invoke(self, prompt: str) -> Any:
                 resp = client.messages.create(
                     model=model,
@@ -43,20 +63,44 @@ def get_llm():
                         return type("Response", (), {"content": block.text})()
                 return type("Response", (), {"content": ""})()
 
+            def stream(self, prompt: str):
+                with client.messages.stream(
+                    model=model,
+                    max_tokens=max_tokens,
+                    temperature=temperature,
+                    messages=[{"role": "user", "content": [{"type": "text", "text": prompt}]}],
+                ) as s:
+                    for text in s.text_stream:
+                        yield text
+
             def get_num_tokens(self, text: str) -> int:
-                return client.count_tokens(text)
+                resp = client.messages.count_tokens(
+                    model=model,
+                    messages=[{"role": "user", "content": [{"type": "text", "text": text}]}],
+                )
+                return resp.input_tokens
 
         return MiniMaxLLM()
     else:
         from langchain_openai import ChatOpenAI
 
-        return ChatOpenAI(
+        raw = ChatOpenAI(
             model=os.getenv("LLM_MODEL", "gpt-4o"),
             api_key=os.getenv("OPENAI_API_KEY"),
             base_url=os.getenv("OPENAI_BASE_URL", "https://api.openai.com/v1"),
             temperature=0.1,
         )
 
+        class OpenAIWrapper(_BaseLLM):
+            def invoke(self, prompt):
+                return raw.invoke(prompt)
+
+            def stream(self, prompt):
+                for chunk in raw.stream(prompt):
+                    yield chunk.content
+
+        return OpenAIWrapper()
+
 
 def get_llm_for_correction():
     return get_llm()