Update agent.py

agent.py

@@ -366,42 +366,55 @@ class MyVectorStore:
 # -----------------------------
 # Define the URL where the JSON file is hosted
 
-import json
+from typing import TypedDict, Annotated, List
+import gradio as gr
 from langchain.schema import Document
-from typing import Any
-
-def process_content(raw_content: Any) -> str:
-    """Convert any input to a clean string, handling lists and reversed text"""
-    # Flatten nested lists
-    def flatten(nested):
-        if isinstance(nested, list):
-            return " ".join(flatten(e) for e in nested)
-        return str(nested)
-    
-    # Convert to string and clean
-    text = flatten(raw_content)
-    
-    # Fix reversed text patterns like ".rewsna..."
-    if text.startswith(('.', ',')):
-        text = text[::-1].replace("\\", "").strip()
-    
-    return text
+import json
+
+# 1. Type-Checked State for Gradio
+class ChatState(TypedDict):
+    messages: Annotated[
+        List[str], 
+        gr.State(render=False),
+        "Stores chat history as list of strings"
+    ]
 
-with open("questions.json", "r", encoding="utf-8") as f:
-    data = json.load(f)
+# 2. Content Processing Utilities
+def process_content(raw_content) -> str:
+    """Convert any input to a clean string"""
+    if isinstance(raw_content, list):
+        return " ".join(str(item) for item in raw_content)
+    return str(raw_content)
+
+def reverse_text(text: str) -> str:
+    """Fix reversed text patterns"""
+    return text[::-1].replace("\\", "").strip() if text.startswith(('.', ',')) else text
+
+# 3. Unified Document Creation
+def create_documents(data_source: str, data: List[dict]) -> List[Document]:
+    """Handle both Gradio chat and JSON questions"""
+    docs = []
+    
+    for item in data:
+        # Process different data sources
+        if data_source == "gradio":
+            content = "\n".join(item["messages"])
+        elif data_source == "json":
+            raw_question = item.get("question", "")
+            content = reverse_text(process_content(raw_question))
+        else:
+            continue
 
-docs = [
-    Document(
-        page_content=process_content(item["question"]),
-        metadata={
-            "task_id": item["task_id"],
-            "level": str(item.get("Level", "")),  # Force string conversion
+        # Ensure metadata type safety
+        metadata = {
+            "task_id": str(item.get("task_id", "")),
+            "level": str(item.get("Level", "")),
             "file_name": str(item.get("file_name", ""))
         }
-    )
-    for item in data
-    if "question" in item  # Skip items without questions
-]
+        
+        docs.append(Document(page_content=content, metadata=metadata))
+    
+    return docs
 
 texts = [doc.page_content for doc in docs]