move to gpt-3.5-turbo-1106

aider/coders/base_coder.py

@@ -53,10 +53,10 @@ class Coder:
     @classmethod
     def create(
         self,
-        client,
         main_model=None,
         edit_format=None,
         io=None,
+        client=None,
         skip_model_availabily_check=False,
         **kwargs,
     ):
@@ -67,12 +67,13 @@ class Coder:
 
         if not skip_model_availabily_check and not main_model.always_available:
             if not check_model_availability(io, client, main_model):
+                fallback_model = models.GPT35_1106
                 if main_model != models.GPT4:
                     io.tool_error(
                         f"API key does not support {main_model.name}, falling back to"
-                        f" {models.GPT35_16k.name}"
+                        f" {fallback_model.name}"
                     )
-                main_model = models.GPT35_16k
+                main_model = fallback_model
 
         if edit_format is None:
             edit_format = main_model.edit_format
@@ -163,7 +164,9 @@ class Coder:
 
         if use_git:
             try:
-                self.repo = GitRepo(self.io, fnames, git_dname, aider_ignore_file)
+                self.repo = GitRepo(
+                    self.io, fnames, git_dname, aider_ignore_file, client=self.client
+                )
                 self.root = self.repo.root
             except FileNotFoundError:
                 self.repo = None
@@ -950,7 +953,8 @@ class Coder:
 
 def check_model_availability(io, client, main_model):
     available_models = client.models.list()
-    model_ids = sorted(model.id for model in available_models["data"])
+    dump(available_models)
+    model_ids = sorted(model.id for model in available_models)
     if main_model.name in model_ids:
         return True
 

aider/main.py

@@ -157,12 +157,13 @@ def main(argv=None, input=None, output=None, force_git_root=None):
         default=False,
         help="Override to skip model availability check (default: False)",
     )
+    default_3_model = models.GPT35_1106
     core_group.add_argument(
         "-3",
         action="store_const",
         dest="model",
-        const=models.GPT35_16k.name,
+        const=default_3_model.name,
-        help=f"Use {models.GPT35_16k.name} model for the main chat (gpt-4 is better)",
+        help=f"Use {default_3_model.name} model for the main chat (gpt-4 is better)",
     )
     core_group.add_argument(
         "--voice-language",
@@ -509,11 +510,11 @@ def main(argv=None, input=None, output=None, force_git_root=None):
 
     try:
         coder = Coder.create(
-            client,
+            main_model=main_model,
-            main_model,
+            edit_format=args.edit_format,
-            args.edit_format,
+            io=io,
-            io,
+            skip_model_availabily_check=args.skip_model_availability_check,
-            args.skip_model_availability_check,
+            client=client,
             ##
             fnames=fnames,
             git_dname=git_dname,

aider/models/__init__.py

@@ -4,6 +4,7 @@ from .openrouter import OpenRouterModel
 
 GPT4 = Model.create("gpt-4")
 GPT35 = Model.create("gpt-3.5-turbo")
+GPT35_1106 = Model.create("gpt-3.5-turbo-1106")
 GPT35_16k = Model.create("gpt-3.5-turbo-16k")
 
 __all__ = [

aider/models/model.py

@@ -33,11 +33,11 @@ class Model:
 
     @staticmethod
     def weak_model():
-        return Model.create("gpt-3.5-turbo")
+        return Model.create("gpt-3.5-turbo-1106")
 
     @staticmethod
     def commit_message_models():
-        return [Model.create("gpt-3.5-turbo"), Model.create("gpt-3.5-turbo-16k")]
+        return [Model.weak_model()]
 
     def token_count(self, messages):
         if not self.tokenizer:

aider/models/openai.py

@@ -55,6 +55,7 @@ class OpenAIModel(Model):
         if self.is_gpt35():
             self.edit_format = "whole"
             self.always_available = True
+            self.send_undo_reply = False
 
             if self.name == "gpt-3.5-turbo-1106":
                 self.prompt_price = 0.001

aider/models/openrouter.py

@@ -7,12 +7,8 @@ cached_model_details = None
 
 class OpenRouterModel(Model):
     def __init__(self, client, name):
-        if name == "gpt-4":
+        if name.startswith("gpt-4") or name.startswith("gpt-3.5-turbo"):
-            name = "openai/gpt-4"
+            name = "openai/" + name
-        elif name == "gpt-3.5-turbo":
-            name = "openai/gpt-3.5-turbo"
-        elif name == "gpt-3.5-turbo-16k":
-            name = "openai/gpt-3.5-turbo-16k"
 
         self.name = name
         self.edit_format = edit_format_for_model(name)

aider/repo.py

@@ -110,9 +110,7 @@ class GitRepo:
             raise OpenAIClientNotProvided
 
         if len(diffs) >= 4 * 1024 * 4:
-            self.io.tool_error(
+            self.io.tool_error("Diff is too large to generate a commit message.")
-                f"Diff is too large for {models.GPT35.name} to generate a commit message."
-            )
             return
 
         diffs = "# Diffs:\n" + diffs

tests/test_models.py

@@ -12,6 +12,9 @@ class TestModels(unittest.TestCase):
         model = Model.create("gpt-3.5-turbo-16k")
         self.assertEqual(model.max_context_tokens, 16 * 1024)
 
+        model = Model.create("gpt-3.5-turbo-1106")
+        self.assertEqual(model.max_context_tokens, 16 * 1024)
+
         model = Model.create("gpt-4")
         self.assertEqual(model.max_context_tokens, 8 * 1024)