Update chunker.py

chunker.py

@@ -1,18 +1,72 @@
+import ast
 from langchain_core.documents import Document
 
-MAX_CHARS = 1200
-
 def chunk_code(file_path: str, code: str):
-    chunks = []
+    """
+    Chunk Python code by functions and classes using AST.
+    Falls back to whole-file chunk if parsing fails.
+    """
+    documents = []
 
-    for i in range(0, len(code), MAX_CHARS):
-        chunk = code[i:i + MAX_CHARS]
-        if len(chunk.strip()) > 100:
-            chunks.append(
+    try:
+        tree = ast.parse(code)
+    except Exception:
+        # Fallback: whole file as one chunk
+        if len(code.strip()) > 100:
+            documents.append(
                 Document(
-                    page_content=chunk,
-                    metadata={"file": file_path}
+                    page_content=code,
+                    metadata={
+                        "file": file_path,
+                        "type": "file"
+                    }
+                )
+            )
+        return documents
+
+    for node in ast.walk(tree):
+        # -------- FUNCTIONS --------
+        if isinstance(node, ast.FunctionDef):
+            source = ast.get_source_segment(code, node)
+            if source:
+                documents.append(
+                    Document(
+                        page_content=source,
+                        metadata={
+                            "file": file_path,
+                            "type": "function",
+                            "name": node.name,
+                            "line_start": node.lineno
+                        }
+                    )
                 )
+
+        # -------- CLASSES --------
+        elif isinstance(node, ast.ClassDef):
+            source = ast.get_source_segment(code, node)
+            if source:
+                documents.append(
+                    Document(
+                        page_content=source,
+                        metadata={
+                            "file": file_path,
+                            "type": "class",
+                            "name": node.name,
+                            "line_start": node.lineno
+                        }
+                    )
+                )
+
+    # If no functions/classes found, keep whole file
+    if not documents and len(code.strip()) > 100:
+        documents.append(
+            Document(
+                page_content=code,
+                metadata={
+                    "file": file_path,
+                    "type": "file"
+                }
             )
+        )
 
-    return chunks
+    return documents