diff --git a/src/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py b/src/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py
index 5551cb2..5c38226 100644
--- a/src/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py
+++ b/src/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py
@@ -17,6 +17,7 @@
     get_args,
     get_origin,
     get_type_hints,
+    Literal,
 )
 
 from docstring_parser import parse
@@ -65,6 +66,7 @@ class PydanticDataType(Enum):
     CUSTOM_CLASS = "custom-class"
     CUSTOM_DICT = "custom-dict"
     SET = "set"
+    LITERAL = "literal"
 
 
 def map_pydantic_type_to_gbnf(pydantic_type: type[Any]) -> str:
@@ -540,6 +542,11 @@ def generate_gbnf_rule_for_type(
         gbnf_type, rules = generate_gbnf_integer_rules(
             max_digit=max_digits, min_digit=min_digits
         )
+    elif get_origin(field_type) is Literal:
+        literal_types = get_args(field_type)
+        literal_types_str = [json.dumps(lt).replace('"', '\\"') for lt in literal_types]
+        literal_types_str = [f'"{lt}"' for lt in literal_types_str]
+        gbnf_type = f"({'|'.join(literal_types_str)})"
     else:
         gbnf_type, rules = gbnf_type, []
 
@@ -763,9 +770,9 @@ def generate_gbnf_grammar_from_pydantic_models(
                 model, processed_models, created_rules
             )
             if add_request_heartbeat and model.__name__ in request_heartbeat_models:
-                model_rules[
-                    0
-                ] += rf' "," ws "\"{request_heartbeat_field_name}\""  ":" ws boolean '
+                model_rules[0] += (
+                    rf' "," ws "\"{request_heartbeat_field_name}\""  ":" ws boolean '
+                )
             # if not has_special_string:
             #     model_rules[0] += r' ws "}"'