Add support for infinite output model fallback

aider/args.py

@@ -250,6 +250,12 @@ def get_parser(default_config_files, git_root):
         default=None,
         help="Specify the edit format for the editor model (default: depends on editor model)",
     )
+    group.add_argument(
+        "--infinite-output-model",
+        metavar="INFINITE_OUTPUT_MODEL",
+        default=None,
+        help="Specify the model to use for continuing long responses (default: None)",
+    )
     group.add_argument(
         "--show-model-warnings",
         action=argparse.BooleanOptionalAction,

aider/coders/base_coder.py

@@ -186,6 +186,7 @@ def get_announcements(self):
         # Model
         main_model = self.main_model
         weak_model = main_model.weak_model
+        infinite_output_model = main_model.infinite_output_model
 
         if weak_model is not main_model:
             prefix = "Main model"
@@ -210,6 +211,10 @@ def get_announcements(self):
             output = f"Weak model: {weak_model.name}"
             lines.append(output)
 
+        if infinite_output_model and infinite_output_model is not main_model:
+            output = f"Infinite output model: {infinite_output_model.name}"
+            lines.append(output)
+
         # Repo
         if self.repo:
             rel_repo_dir = self.repo.get_rel_repo_dir()
@@ -1275,9 +1280,14 @@ def send_message(self, inp):
                     break
                 except FinishReasonLength:
                     # We hit the output limit!
-                    if not self.main_model.info.get("supports_assistant_prefill"):
-                        exhausted = True
-                        break
+                    if self.main_model.info.get("supports_assistant_prefill"):
+                        use_model = self.main_model
+                    else:
+                        # Try to get an infinite output model
+                        use_model = self.main_model.infinite_output_model
+                        if not use_model or not use_model.info.get("supports_assistant_prefill"):
+                            exhausted = True
+                            break
 
                     self.multi_response_content = self.get_multi_response_content()
 
@@ -1287,6 +1297,10 @@ def send_message(self, inp):
                         messages.append(
                             dict(role="assistant", content=self.multi_response_content, prefix=True)
                         )
+
+                    # Switch to the infinite output model if needed
+                    if use_model != self.main_model:
+                        self.main_model = use_model
                 except Exception as err:
                     self.mdstream = None
                     lines = traceback.format_exception(type(err), err, err.__traceback__)

aider/main.py

@@ -717,6 +717,7 @@ def get_io(pretty):
         weak_model=args.weak_model,
         editor_model=args.editor_model,
         editor_edit_format=args.editor_edit_format,
+        infinite_output_model=args.infinite_output_model,
     )
 
     if args.copy_paste and args.edit_format is None:

aider/models.py

@@ -86,6 +86,7 @@ class ModelSettings:
     name: str
     edit_format: str = "whole"
     weak_model_name: Optional[str] = None
+    infinite_output_model_name: Optional[str] = None
     use_repo_map: bool = False
     send_undo_reply: bool = False
     lazy: bool = False
@@ -857,7 +858,7 @@ def get_model_info(self, model):
 
 
 class Model(ModelSettings):
-    def __init__(self, model, weak_model=None, editor_model=None, editor_edit_format=None):
+    def __init__(self, model, weak_model=None, editor_model=None, editor_edit_format=None, infinite_output_model=None):
         # Map any alias to its canonical name
         model = MODEL_ALIASES.get(model, model)
 
@@ -866,6 +867,7 @@ def __init__(self, model, weak_model=None, editor_model=None, editor_edit_format
         self.max_chat_history_tokens = 1024
         self.weak_model = None
         self.editor_model = None
+        self.infinite_output_model = None
 
         # Find the extra settings
         self.extra_model_settings = next(
@@ -896,6 +898,11 @@ def __init__(self, model, weak_model=None, editor_model=None, editor_edit_format
         else:
             self.get_editor_model(editor_model, editor_edit_format)
 
+        if infinite_output_model is False:
+            self.infinite_output_model_name = None
+        else:
+            self.get_infinite_output_model(infinite_output_model)
+
     def get_model_info(self, model):
         return model_info_manager.get_model_info(model)
 
@@ -1015,6 +1022,24 @@ def get_weak_model(self, provided_weak_model_name):
     def commit_message_models(self):
         return [self.weak_model, self]
 
+    def get_infinite_output_model(self, provided_infinite_output_model_name):
+        # If infinite_output_model_name is provided, override the model settings
+        if provided_infinite_output_model_name:
+            self.infinite_output_model_name = provided_infinite_output_model_name
+
+        if not self.infinite_output_model_name:
+            return None
+
+        if self.infinite_output_model_name == self.name:
+            return self
+
+        self.infinite_output_model = Model(
+            self.infinite_output_model_name,
+            weak_model=False,
+            editor_model=False,
+        )
+        return self.infinite_output_model
+
     def get_editor_model(self, provided_editor_model_name, editor_edit_format):
         # If editor_model_name is provided, override the model settings
         if provided_editor_model_name: