Building Custom AI Coding Assistants for Enterprise Codebases: RAG, AST Indexing, and Evals

from tree_sitter import Language, Parser
import tree_sitter_python as tspython

# Initialize Tree-Sitter Python AST Parser
PY_LANGUAGE = Language(tspython.language())
parser = Parser(PY_LANGUAGE)

def extract_ast_code_blocks(source_code: str) -> list[dict]:
    """Parses source code into semantic AST function and class blocks for AI vector indexing."""
    tree = parser.parse(bytes(source_code, "utf8"))
    root_node = tree.root_node
    
    code_blocks = []
    
    for child in root_node.children:
        if child.type in ["function_definition", "class_definition"]:
            start_line = child.start_point[0]
            end_line = child.end_point[0]
            block_name = child.child_by_field_name("name").text.decode("utf8")
            
            code_blocks.append({
                "block_type": child.type,
                "name": block_name,
                "start_line": start_line,
                "end_line": end_line,
                "content": source_code.splitlines()[start_line:end_line+1]
            })
            
    return code_blocks