diff --git a/HISTORY.md b/HISTORY.md index d78f259d1..311ae502a 100644 --- a/HISTORY.md +++ b/HISTORY.md @@ -3,9 +3,9 @@ ### main branch -- [Use a pair of Senior/Junior models for improved coding](https://aider.chat/2024/09/26/senior-junior.html) - - Use a strong reasoning model like o1-preview as your Senior coder. - - Use a cheaper, faster model like gpt-4o as your Junior coder. +- [Use a pair of Architect/Editor models for improved coding](https://aider.chat/2024/09/26/senior-editor.html) + - Use a strong reasoning model like o1-preview as your Architect. + - Use a cheaper, faster model like gpt-4o as your Editor. - New `--o1-preview` and `--o1-mini` shortcuts. - New settings for completion menu colors, by @smh. - New `--voice-format` switch to send voice audio as wav/mp3/webm, by @mbailey. diff --git a/aider/args.py b/aider/args.py index 26ec41b57..f594d8896 100644 --- a/aider/args.py +++ b/aider/args.py @@ -198,11 +198,11 @@ def get_parser(default_config_files, git_root): help="Specify what edit format the LLM should use (default depends on model)", ) group.add_argument( - "--senior", + "--architect", action="store_const", dest="edit_format", - const="senior", - help="Use senior edit format for the main chat", + const="architect", + help="Use architect edit format for the main chat", ) group.add_argument( "--weak-model", @@ -214,16 +214,16 @@ def get_parser(default_config_files, git_root): ), ) group.add_argument( - "--junior-model", + "--editor-model", metavar="JUNIOR_MODEL", default=None, - help="Specify the model to use for junior tasks (default depends on --model)", + help="Specify the model to use for editor tasks (default depends on --model)", ) group.add_argument( - "--junior-edit-format", + "--editor-edit-format", metavar="JUNIOR_EDIT_FORMAT", default=None, - help="Specify the edit format for the junior model (default: depends on junior model)", + help="Specify the edit format for the editor model (default: depends on editor model)", ) group.add_argument( "--show-model-warnings", diff --git a/aider/coders/__init__.py b/aider/coders/__init__.py index edce98129..e9d334bc9 100644 --- a/aider/coders/__init__.py +++ b/aider/coders/__init__.py @@ -1,11 +1,11 @@ +from .architect_coder import ArchitectCoder from .ask_coder import AskCoder from .base_coder import Coder from .editblock_coder import EditBlockCoder from .editblock_fenced_coder import EditBlockFencedCoder +from .editor_editblock_coder import EditorEditBlockCoder +from .editor_whole_coder import EditorWholeFileCoder from .help_coder import HelpCoder -from .junior_editblock_coder import JuniorEditBlockCoder -from .junior_whole_coder import JuniorWholeFileCoder -from .senior_coder import SeniorCoder from .udiff_coder import UnifiedDiffCoder from .wholefile_coder import WholeFileCoder @@ -20,7 +20,7 @@ __all__ = [ WholeFileCoder, UnifiedDiffCoder, # SingleWholeFileFunctionCoder, - SeniorCoder, - JuniorEditBlockCoder, - JuniorWholeFileCoder, + ArchitectCoder, + EditorEditBlockCoder, + EditorWholeFileCoder, ] diff --git a/aider/coders/base_coder.py b/aider/coders/base_coder.py index 28b4dec64..772a16b17 100755 --- a/aider/coders/base_coder.py +++ b/aider/coders/base_coder.py @@ -180,10 +180,10 @@ class Coder: output += ", infinite output" lines.append(output) - if self.edit_format == "senior": + if self.edit_format == "architect": output = ( - f"Junior model: {main_model.junior_model.name} with" - f" {main_model.junior_edit_format} edit format" + f"Editor model: {main_model.editor_model.name} with" + f" {main_model.editor_edit_format} edit format" ) lines.append(output) diff --git a/aider/coders/junior_editblock_coder.py b/aider/coders/junior_editblock_coder.py index d2492222b..301ab14af 100644 --- a/aider/coders/junior_editblock_coder.py +++ b/aider/coders/junior_editblock_coder.py @@ -1,7 +1,7 @@ from .editblock_coder import EditBlockCoder -from .junior_editblock_prompts import JuniorEditBlockPrompts +from .editor_editblock_prompts import EditorEditBlockPrompts -class JuniorEditBlockCoder(EditBlockCoder): - edit_format = "junior-diff" - gpt_prompts = JuniorEditBlockPrompts() +class EditorEditBlockCoder(EditBlockCoder): + edit_format = "editor-diff" + gpt_prompts = EditorEditBlockPrompts() diff --git a/aider/coders/junior_editblock_prompts.py b/aider/coders/junior_editblock_prompts.py index a511a6b66..595e28bee 100644 --- a/aider/coders/junior_editblock_prompts.py +++ b/aider/coders/junior_editblock_prompts.py @@ -3,7 +3,7 @@ from .editblock_prompts import EditBlockPrompts -class JuniorEditBlockPrompts(EditBlockPrompts): +class EditorEditBlockPrompts(EditBlockPrompts): main_system = """Act as an expert software developer who edits source code. {lazy_prompt} Describe each change with a *SEARCH/REPLACE block* per the examples below. diff --git a/aider/coders/junior_whole_coder.py b/aider/coders/junior_whole_coder.py index d864cfab9..d3b4d278e 100644 --- a/aider/coders/junior_whole_coder.py +++ b/aider/coders/junior_whole_coder.py @@ -1,7 +1,7 @@ -from .junior_whole_prompts import JuniorWholeFilePrompts +from .editor_whole_prompts import EditorWholeFilePrompts from .wholefile_coder import WholeFileCoder -class JuniorWholeFileCoder(WholeFileCoder): - edit_format = "junior-whole" - gpt_prompts = JuniorWholeFilePrompts() +class EditorWholeFileCoder(WholeFileCoder): + edit_format = "editor-whole" + gpt_prompts = EditorWholeFilePrompts() diff --git a/aider/coders/junior_whole_prompts.py b/aider/coders/junior_whole_prompts.py index 4441b76be..7abd0ee53 100644 --- a/aider/coders/junior_whole_prompts.py +++ b/aider/coders/junior_whole_prompts.py @@ -3,7 +3,7 @@ from .wholefile_prompts import WholeFilePrompts -class JuniorWholeFilePrompts(WholeFilePrompts): +class EditorWholeFilePrompts(WholeFilePrompts): main_system = """Act as an expert software developer and make changes to source code. {lazy_prompt} Output a copy of each file that needs changes. diff --git a/aider/coders/senior_coder.py b/aider/coders/senior_coder.py index 918869e1c..983c0e11b 100644 --- a/aider/coders/senior_coder.py +++ b/aider/coders/senior_coder.py @@ -1,11 +1,11 @@ +from .architect_prompts import ArchitectPrompts from .ask_coder import AskCoder from .base_coder import Coder -from .senior_prompts import SeniorPrompts -class SeniorCoder(AskCoder): - edit_format = "senior" - gpt_prompts = SeniorPrompts() +class ArchitectCoder(AskCoder): + edit_format = "architect" + gpt_prompts = ArchitectPrompts() def reply_completed(self): content = self.partial_response_content @@ -15,11 +15,11 @@ class SeniorCoder(AskCoder): kwargs = dict() - # Use the junior_model from the main_model if it exists, otherwise use the main_model itself - junior_model = self.main_model.junior_model or self.main_model + # Use the editor_model from the main_model if it exists, otherwise use the main_model itself + editor_model = self.main_model.editor_model or self.main_model - kwargs["main_model"] = junior_model - kwargs["edit_format"] = self.main_model.junior_edit_format + kwargs["main_model"] = editor_model + kwargs["edit_format"] = self.main_model.editor_edit_format kwargs["suggest_shell_commands"] = False kwargs["map_tokens"] = 0 kwargs["total_cost"] = self.total_cost @@ -29,12 +29,12 @@ class SeniorCoder(AskCoder): new_kwargs = dict(io=self.io, from_coder=self) new_kwargs.update(kwargs) - junior_coder = Coder.create(**new_kwargs) - junior_coder.cur_messages = [] - junior_coder.done_messages = [] - junior_coder.show_announcements() + editor_coder = Coder.create(**new_kwargs) + editor_coder.cur_messages = [] + editor_coder.done_messages = [] + editor_coder.show_announcements() - junior_coder.run(with_message=content, preproc=False) + editor_coder.run(with_message=content, preproc=False) self.move_back_cur_messages("I made those changes to the files.") - self.total_cost = junior_coder.total_cost + self.total_cost = editor_coder.total_cost diff --git a/aider/coders/senior_prompts.py b/aider/coders/senior_prompts.py index ae64a30fa..e6798dbba 100644 --- a/aider/coders/senior_prompts.py +++ b/aider/coders/senior_prompts.py @@ -3,11 +3,11 @@ from .base_prompts import CoderPrompts -class SeniorPrompts(CoderPrompts): - main_system = """Act as an expert senior engineer and provide direction to your junior engineer. +class ArchitectPrompts(CoderPrompts): + main_system = """Act as an expert architect engineer and provide direction to your editor engineer. Study the change request and the current code. Describe how to modify the code to complete the request. -The junior engineer will rely solely on your instructions, so make them unambiguous and complete. +The editor engineer will rely solely on your instructions, so make them unambiguous and complete. Explain all needed code changes clearly and completely, but concisely. Just show the changes needed. diff --git a/aider/main.py b/aider/main.py index 8af56658e..5d3aeb39f 100644 --- a/aider/main.py +++ b/aider/main.py @@ -536,8 +536,8 @@ def main(argv=None, input=None, output=None, force_git_root=None, return_coder=F main_model = models.Model( args.model, weak_model=args.weak_model, - junior_model=args.junior_model, - junior_edit_format=args.junior_edit_format, + editor_model=args.editor_model, + editor_edit_format=args.editor_edit_format, ) if args.verbose: diff --git a/aider/models.py b/aider/models.py index 5a42f9e8f..40433647b 100644 --- a/aider/models.py +++ b/aider/models.py @@ -81,8 +81,8 @@ class ModelSettings: use_system_prompt: bool = True use_temperature: bool = True streaming: bool = True - junior_model_name: Optional[str] = None - junior_edit_format: Optional[str] = None + editor_model_name: Optional[str] = None + editor_edit_format: Optional[str] = None # https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo @@ -148,7 +148,7 @@ MODEL_SETTINGS = [ accepts_images=True, lazy=True, reminder="sys", - junior_edit_format="junior-diff", + editor_edit_format="editor-diff", ), ModelSettings( "openai/gpt-4o-2024-08-06", @@ -176,7 +176,7 @@ MODEL_SETTINGS = [ accepts_images=True, lazy=True, reminder="sys", - junior_edit_format="junior-diff", + editor_edit_format="editor-diff", ), ModelSettings( "gpt-4o-mini", @@ -263,8 +263,8 @@ MODEL_SETTINGS = [ "claude-3-5-sonnet-20240620", "diff", weak_model_name="claude-3-haiku-20240307", - junior_model_name="claude-3-5-sonnet-20240620", - junior_edit_format="junior-diff", + editor_model_name="claude-3-5-sonnet-20240620", + editor_edit_format="editor-diff", use_repo_map=True, examples_as_sys_msg=True, accepts_images=True, @@ -279,8 +279,8 @@ MODEL_SETTINGS = [ "anthropic/claude-3-5-sonnet-20240620", "diff", weak_model_name="claude-3-haiku-20240307", - junior_model_name="anthropic/claude-3-5-sonnet-20240620", - junior_edit_format="junior-diff", + editor_model_name="anthropic/claude-3-5-sonnet-20240620", + editor_edit_format="editor-diff", use_repo_map=True, examples_as_sys_msg=True, max_tokens=8192, @@ -314,8 +314,8 @@ MODEL_SETTINGS = [ "openrouter/anthropic/claude-3.5-sonnet", "diff", weak_model_name="openrouter/anthropic/claude-3-haiku-20240307", - junior_model_name="openrouter/anthropic/claude-3.5-sonnet", - junior_edit_format="junior-diff", + editor_model_name="openrouter/anthropic/claude-3.5-sonnet", + editor_edit_format="editor-diff", use_repo_map=True, examples_as_sys_msg=True, accepts_images=True, @@ -329,8 +329,8 @@ MODEL_SETTINGS = [ "vertex_ai/claude-3-5-sonnet@20240620", "diff", weak_model_name="vertex_ai/claude-3-haiku@20240307", - junior_model_name="vertex_ai/claude-3-5-sonnet@20240620", - junior_edit_format="junior-diff", + editor_model_name="vertex_ai/claude-3-5-sonnet@20240620", + editor_edit_format="editor-diff", use_repo_map=True, examples_as_sys_msg=True, accepts_images=True, @@ -466,14 +466,14 @@ MODEL_SETTINGS = [ accepts_images=True, lazy=True, reminder="sys", - junior_edit_format="junior-diff", + editor_edit_format="editor-diff", ), ModelSettings( "openai/o1-mini", "whole", weak_model_name="openai/gpt-4o-mini", - junior_model_name="openai/gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="openai/gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -484,8 +484,8 @@ MODEL_SETTINGS = [ "o1-mini", "whole", weak_model_name="gpt-4o-mini", - junior_model_name="gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -496,8 +496,8 @@ MODEL_SETTINGS = [ "openai/o1-preview", "diff", weak_model_name="openai/gpt-4o-mini", - junior_model_name="openai/gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="openai/gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -506,10 +506,10 @@ MODEL_SETTINGS = [ ), ModelSettings( "o1-preview", - "senior", + "architect", weak_model_name="gpt-4o-mini", - junior_model_name="gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -520,8 +520,8 @@ MODEL_SETTINGS = [ "openrouter/openai/o1-mini", "whole", weak_model_name="openrouter/openai/gpt-4o-mini", - junior_model_name="openrouter/openai/gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="openrouter/openai/gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -532,8 +532,8 @@ MODEL_SETTINGS = [ "openrouter/openai/o1-preview", "diff", weak_model_name="openrouter/openai/gpt-4o-mini", - junior_model_name="openrouter/openai/gpt-4o", - junior_edit_format="junior-diff", + editor_model_name="openrouter/openai/gpt-4o", + editor_edit_format="editor-diff", use_repo_map=True, reminder="user", use_system_prompt=False, @@ -616,11 +616,11 @@ def get_model_info(model): class Model(ModelSettings): - def __init__(self, model, weak_model=None, junior_model=None, junior_edit_format=None): + def __init__(self, model, weak_model=None, editor_model=None, editor_edit_format=None): self.name = model self.max_chat_history_tokens = 1024 self.weak_model = None - self.junior_model = None + self.editor_model = None self.info = self.get_model_info(model) @@ -641,10 +641,10 @@ class Model(ModelSettings): else: self.get_weak_model(weak_model) - if junior_model is False: - self.junior_model_name = None + if editor_model is False: + self.editor_model_name = None else: - self.get_junior_model(junior_model, junior_edit_format) + self.get_editor_model(editor_model, editor_edit_format) def get_model_info(self, model): return get_model_info(model) @@ -717,25 +717,25 @@ class Model(ModelSettings): def commit_message_models(self): return [self.weak_model, self] - def get_junior_model(self, provided_junior_model_name, junior_edit_format): - # If junior_model_name is provided, override the model settings - if provided_junior_model_name: - self.junior_model_name = provided_junior_model_name - if junior_edit_format: - self.junior_edit_format = junior_edit_format + 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: + self.editor_model_name = provided_editor_model_name + if editor_edit_format: + self.editor_edit_format = editor_edit_format - if not self.junior_model_name or self.junior_model_name == self.name: - self.junior_model = self + if not self.editor_model_name or self.editor_model_name == self.name: + self.editor_model = self else: - self.junior_model = Model( - self.junior_model_name, - junior_model=False, + self.editor_model = Model( + self.editor_model_name, + editor_model=False, ) - if not self.junior_edit_format: - self.junior_edit_format = self.junior_model.edit_format + if not self.editor_edit_format: + self.editor_edit_format = self.editor_model.edit_format - return self.junior_model + return self.editor_model def tokenizer(self, text): return litellm.encode(model=self.name, text=text) diff --git a/aider/website/HISTORY.md b/aider/website/HISTORY.md index 63ac05679..8168e8086 100644 --- a/aider/website/HISTORY.md +++ b/aider/website/HISTORY.md @@ -18,9 +18,9 @@ cog.out(text) ### main branch -- [Use a pair of Senior/Junior models for improved coding](https://aider.chat/2024/09/26/senior-junior.html) - - Use a strong reasoning model like o1-preview as your Senior coder. - - Use a cheaper, faster model like gpt-4o as your Junior coder. +- [Use a pair of Architect/Editor models for improved coding](https://aider.chat/2024/09/26/senior-editor.html) + - Use a strong reasoning model like o1-preview as your Architect. + - Use a cheaper, faster model like gpt-4o as your Editor. - New `--o1-preview` and `--o1-mini` shortcuts. - New settings for completion menu colors, by @smh. - New `--voice-format` switch to send voice audio as wav/mp3/webm, by @mbailey. diff --git a/aider/website/_data/senior.yml b/aider/website/_data/senior.yml index 253790657..9345af3a1 100644 --- a/aider/website/_data/senior.yml +++ b/aider/website/_data/senior.yml @@ -1,9 +1,9 @@ -- dirname: 2024-09-25-21-17-19--senior-sonnet-sonnet-diff +- dirname: 2024-09-25-21-17-19--architect-sonnet-sonnet-diff test_cases: 133 model: claude-3.5-sonnet - junior_model: claude-3.5-sonnet - junior_edit_format: diff - edit_format: senior + editor_model: claude-3.5-sonnet + editor_edit_format: diff + edit_format: architect commit_hash: c18d6a8-dirty pass_rate_1: 62.4 pass_rate_2: 80.5 @@ -47,12 +47,12 @@ seconds_per_case: 17.6 total_cost: 3.6346 -- dirname: 2024-09-25-21-25-01--senior-o1mini-4o-jr-diff +- dirname: 2024-09-25-21-25-01--architect-o1mini-4o-jr-diff test_cases: 133 model: o1-mini - junior_model: gpt-4o - junior_edit_format: diff - edit_format: senior + editor_model: gpt-4o + editor_edit_format: diff + edit_format: architect commit_hash: 3f682ed-dirty, 25e833b pass_rate_1: 51.1 pass_rate_2: 70.7 @@ -72,13 +72,13 @@ seconds_per_case: 23.7 total_cost: 9.3158 -- dirname: 2024-09-26-15-05-58--senior-o1mini-deep-jr-whole +- dirname: 2024-09-26-15-05-58--architect-o1mini-deep-jr-whole test_cases: 133 model: o1-mini - edit_format: senior + edit_format: architect commit_hash: 1676653-dirty - junior_model: deepseek - junior_edit_format: whole + editor_model: deepseek + editor_edit_format: whole pass_rate_1: 51.9 pass_rate_2: 71.4 percent_cases_well_formed: 100.0 @@ -97,12 +97,12 @@ seconds_per_case: 48.2 total_cost: 5.6069 -- dirname: 2024-09-25-21-33-40--senior-4o-4o-jr-diff +- dirname: 2024-09-25-21-33-40--architect-4o-4o-jr-diff test_cases: 133 model: gpt-4o - junior_model: gpt-4o - junior_edit_format: diff - edit_format: senior + editor_model: gpt-4o + editor_edit_format: diff + edit_format: architect commit_hash: 9f3cd92 pass_rate_1: 56.4 pass_rate_2: 75.2 @@ -145,12 +145,12 @@ seconds_per_case: 80.9 total_cost: 63.9190 -- dirname: 2024-09-25-21-39-05--senior-o1preview-4o-jr-diff +- dirname: 2024-09-25-21-39-05--architect-o1preview-4o-jr-diff test_cases: 133 model: o1-preview - junior_model: gpt-4o - junior_edit_format: diff - edit_format: senior + editor_model: gpt-4o + editor_edit_format: diff + edit_format: architect commit_hash: 9f3cd92 pass_rate_1: 63.2 pass_rate_2: 80.5 @@ -170,14 +170,14 @@ seconds_per_case: 42.3 total_cost: 39.3766 -- dirname: 2024-09-25-21-52-42--senior-o1preview-sonnet-jr-diff +- dirname: 2024-09-25-21-52-42--architect-o1preview-sonnet-jr-diff test_cases: 133 model: o1-preview - junior_model: claude-3.5-sonnet - junior_edit_format: diff - edit_format: senior + editor_model: claude-3.5-sonnet + editor_edit_format: diff + edit_format: architect commit_hash: 9f3cd92 - junior_model: claude-3-5-sonnet + editor_model: claude-3-5-sonnet pass_rate_1: 60.9 pass_rate_2: 82.7 percent_cases_well_formed: 100.0 @@ -219,13 +219,13 @@ seconds_per_case: 26.7 total_cost: 2.4226 -- dirname: 2024-09-25-23-12-14--senior-o1mini-deep-jr-diff +- dirname: 2024-09-25-23-12-14--architect-o1mini-deep-jr-diff test_cases: 133 model: o1-mini - edit_format: senior + edit_format: architect commit_hash: 9f3cd92-dirty - junior_model: deepseek - junior_edit_format: diff + editor_model: deepseek + editor_edit_format: diff pass_rate_1: 48.9 pass_rate_2: 69.2 percent_cases_well_formed: 100.0 @@ -244,13 +244,13 @@ seconds_per_case: 52.2 total_cost: 5.7927 -- dirname: 2024-09-25-23-18-16--senior-o1preview-deep-jr-diff +- dirname: 2024-09-25-23-18-16--architect-o1preview-deep-jr-diff test_cases: 133 model: o1-preview - edit_format: senior + edit_format: architect commit_hash: 9f3cd92-dirty - junior_model: deepseek - junior_edit_format: diff + editor_model: deepseek + editor_edit_format: diff pass_rate_1: 64.7 pass_rate_2: 80.5 percent_cases_well_formed: 100.0 @@ -269,13 +269,13 @@ seconds_per_case: 73.2 total_cost: 35.7887 -- dirname: 2024-09-25-23-30-36--senior-o1preview-deep-jr-whole +- dirname: 2024-09-25-23-30-36--architect-o1preview-deep-jr-whole test_cases: 133 model: o1-preview - edit_format: senior + edit_format: architect commit_hash: 9f3cd92-dirty - junior_model: deepseek - junior_edit_format: whole + editor_model: deepseek + editor_edit_format: whole pass_rate_1: 63.9 pass_rate_2: 85.0 percent_cases_well_formed: 100.0 @@ -294,13 +294,13 @@ seconds_per_case: 67.4 total_cost: 35.3152 -- dirname: 2024-09-26-15-15-17--senior-sonnet-deep-jr-whole +- dirname: 2024-09-26-15-15-17--architect-sonnet-deep-jr-whole test_cases: 133 model: claude-3.5-sonnet - edit_format: senior + edit_format: architect commit_hash: bc1559f-dirty - junior_model: deepseek - junior_edit_format: whole + editor_model: deepseek + editor_edit_format: whole pass_rate_1: 61.7 pass_rate_2: 78.9 percent_cases_well_formed: 100.0 @@ -342,13 +342,13 @@ seconds_per_case: 9.7 total_cost: 3.8088 -- dirname: 2024-09-26-15-41-08--senior-4o-deep-jr-whole +- dirname: 2024-09-26-15-41-08--architect-4o-deep-jr-whole test_cases: 133 model: gpt-4o - edit_format: senior + edit_format: architect commit_hash: 89aa385-dirty - junior_model: deepseek - junior_edit_format: whole + editor_model: deepseek + editor_edit_format: whole pass_rate_1: 60.9 pass_rate_2: 73.7 percent_cases_well_formed: 100.0 @@ -367,13 +367,13 @@ seconds_per_case: 38.0 total_cost: 2.4737 -- dirname: 2024-09-26-15-54-08--senior-4o-deep-jr-diff +- dirname: 2024-09-26-15-54-08--architect-4o-deep-jr-diff test_cases: 133 model: gpt-4o - edit_format: senior + edit_format: architect commit_hash: 89aa385-dirty - junior_model: deepseek - junior_edit_format: diff + editor_model: deepseek + editor_edit_format: diff pass_rate_1: 57.1 pass_rate_2: 74.4 percent_cases_well_formed: 100.0 @@ -392,13 +392,13 @@ seconds_per_case: 44.0 total_cost: 2.5498 -- dirname: 2024-09-26-16-06-39--senior-sonnet-deep-jr-diff +- dirname: 2024-09-26-16-06-39--architect-sonnet-deep-jr-diff test_cases: 133 model: claude-3.5-sonnet - edit_format: senior + edit_format: architect commit_hash: 89aa385-dirty - junior_model: deepseek - junior_edit_format: diff + editor_model: deepseek + editor_edit_format: diff pass_rate_1: 61.7 pass_rate_2: 78.9 percent_cases_well_formed: 100.0 diff --git a/aider/website/_posts/2024-09-26-senior-junior.md b/aider/website/_posts/2024-09-26-senior-junior.md index 4264dd29f..2201aae94 100644 --- a/aider/website/_posts/2024-09-26-senior-junior.md +++ b/aider/website/_posts/2024-09-26-senior-junior.md @@ -1,7 +1,7 @@ --- title: Separating code reasoning and editing -excerpt: A Senior model describes how to solve the coding problem, and a Junior model translates that into file edits. This Senior/Junior approach produces SOTA benchmark results. -highlight_image: /assets/senior.jpg +excerpt: A Architect model describes how to solve the coding problem, and a Editor model translates that into file edits. This Architect/Editor approach produces SOTA benchmark results. +highlight_image: /assets/architect.jpg draft: true nav_exclude: true --- @@ -13,8 +13,8 @@ nav_exclude: true Aider now has experimental support for using two models to complete each coding task: -- A Senior model is asked to describe how to solve the coding problem. -- A Junior model is given the Senior's solution and asked to produce specific code editing instructions to apply those changes to source files. +- A Architect model is asked to describe how to solve the coding problem. +- A Editor model is given the Architect's solution and asked to produce specific code editing instructions to apply those changes to source files. Splitting up "code reasoning" and "code editing" has produced SOTA results on [aider's code editing benchmark](/docs/benchmarks.html#the-benchmark). @@ -70,9 +70,9 @@ top coding models, as compared to their previous "solo" scores (striped bars). {% assign grouped_data = sorted_data | group_by: "model" %} {% for group in grouped_data %} {% for item in group.items %} - labels.push("{{ item.junior_model | default: "(No Junior)" }} {{ item.junior_edit_format | default: item.edit_format }}"); + labels.push("{{ item.editor_model | default: "(No Editor)" }} {{ item.editor_edit_format | default: item.edit_format }}"); data.push({{ item.pass_rate_2 }}); - if ("{{ item.junior_model }}" == "") { + if ("{{ item.editor_model }}" == "") { backgroundColors.push(patterns["{{ item.model }}"]); } else { backgroundColors.push(colorMapping["{{ item.model }}"]); @@ -114,7 +114,7 @@ top coding models, as compared to their previous "solo" scores (striped bars). x: { title: { display: true, - text: 'Junior model and edit format', + text: 'Editor model and edit format', font: { size: 18 } @@ -201,7 +201,7 @@ They are strong at reasoning, but often fail to output properly formatted code editing instructions. It helps to instead let them describe the solution however they prefer and then pass that output to a more traditional LLM. -This Junior LLM can then interpret the solution description and +This Editor LLM can then interpret the solution description and produce the code editing instructions needed to update the existing source code file. @@ -209,7 +209,7 @@ Traditional frontier models like gpt-4o and Sonnet also seem to benefit from separating code reasoning and editing like this. A pair of gpt-4o or a pair of Sonnet models -in Senior/Junior configuration outperform their previous solo benchmark results. +in Architect/Editor configuration outperform their previous solo benchmark results. Another reason why this approach is newly viable is that the speed and costs of frontier models have been rapidly improving. @@ -233,41 +233,41 @@ But this all happens in a single prompt/response round trip to the LLM, and the model has to split its attention between solving the coding problem and confirming to the edit format. -The Senior/Junior approach splits this into two round trips, possibly +The Architect/Editor approach splits this into two round trips, possibly using two different LLMs: -- Ask how to solve the coding problem (Senior). -- Turn the proposed solution into a series of well formed code edits (Junior). +- Ask how to solve the coding problem (Architect). +- Turn the proposed solution into a series of well formed code edits (Editor). -The Senior/Junior approach allows the Senior to focus on solving the coding problem +The Architect/Editor approach allows the Architect to focus on solving the coding problem and describe the solution however comes naturally to it. -This gives the Senior more reasoning capacity to focus just on solving the coding +This gives the Architect more reasoning capacity to focus just on solving the coding task. -We can also assign the Senior task to a strong reasoning model like o1-preview, +We can also assign the Architect task to a strong reasoning model like o1-preview, and give the editing task to an appropriate model based on cost, editing skill, etc. -Similarly, the Junior can focus all of its attention on properly formatting the edits +Similarly, the Editor can focus all of its attention on properly formatting the edits without needing to reason much about how to solve the coding problem. ## Results The graph above and the table below show the [aider's code editing benchmark](/docs/benchmarks.html#the-benchmark) -score for various combinations of Senior and Junior models. +score for various combinations of Architect and Editor models. Some noteworthy observations: -- Pairing o1-preview as Senior with Deepseek as Junior sets a SOTA significantly above the previous best score. This result is obtained with Deepseek using the "whole" editing format, requiring it to output a full update copy of each edited source file. Both of these steps are therefore quite slow, so probably not practical for interactive use with aider. -- Pairing OpenAI's o1-preview with Anthropic's Sonnet as the Junior produces the second best result. This is an entirely practical configuration for users able to work with both providers. +- Pairing o1-preview as Architect with Deepseek as Editor sets a SOTA significantly above the previous best score. This result is obtained with Deepseek using the "whole" editing format, requiring it to output a full update copy of each edited source file. Both of these steps are therefore quite slow, so probably not practical for interactive use with aider. +- Pairing OpenAI's o1-preview with Anthropic's Sonnet as the Editor produces the second best result. This is an entirely practical configuration for users able to work with both providers. - Pairing Sonnet/Sonnet and GPT-4o/GPT-4o provides significant lift for both models compared to their solo results, especially for GPT-4o. -- Deepseek is surprisingly effective as a Junior model. It seems remarkably capable at turning proposed coding solutions into new, updated versions of the source files. Using the efficient "diff" editing format, Deepseek helps all the Senior models except for Sonnet. +- Deepseek is surprisingly effective as a Editor model. It seems remarkably capable at turning proposed coding solutions into new, updated versions of the source files. Using the efficient "diff" editing format, Deepseek helps all the Architect models except for Sonnet. ## Try it! The development version of aider -has built in defaults to support Senior/Junior coding with +has built in defaults to support Architect/Editor coding with OpenAI's o1 models, gpt-4o and Anthropic's Claude 3.5 Sonnet. -Run aider with `--senior` or get started quickly like this: +Run aider with `--architect` or get started quickly like this: ``` pip install -U git+https://github.com/paul-gauthier/aider.git @@ -275,15 +275,15 @@ pip install -U git+https://github.com/paul-gauthier/aider.git # Change directory into a git repo cd /to/your/git/repo -# Work with Claude 3.5 Sonnet as the Senior and Junior +# Work with Claude 3.5 Sonnet as the Architect and Editor export ANTHROPIC_API_KEY=your-key-goes-here -aider --sonnet --senior +aider --sonnet --architect -# Work with OpenAI models, using gpt-4o as the Junior +# Work with OpenAI models, using gpt-4o as the Editor export OPENAI_API_KEY=your-key-goes-here -aider --4o --senior -aider --o1-mini --senior -aider --o1-preview --senior +aider --4o --architect +aider --o1-mini --architect +aider --o1-preview --architect ``` ## Full results @@ -292,8 +292,8 @@ aider --o1-preview --senior - - + + @@ -304,8 +304,8 @@ aider --o1-preview --senior {% for item in group.items %} - - + + diff --git a/aider/website/assets/sample.aider.conf.yml b/aider/website/assets/sample.aider.conf.yml index f8316ad8a..0f3d173bf 100644 --- a/aider/website/assets/sample.aider.conf.yml +++ b/aider/website/assets/sample.aider.conf.yml @@ -89,17 +89,17 @@ ## Specify what edit format the LLM should use (default depends on model) #edit-format: xxx -## Use senior edit format for the main chat -#senior: false +## Use architect edit format for the main chat +#architect: false ## Specify the model to use for commit messages and chat history summarization (default depends on --model) #weak-model: xxx -## Specify the model to use for junior tasks (default depends on --model) -#junior-model: xxx +## Specify the model to use for editor tasks (default depends on --model) +#editor-model: xxx -## Specify the edit format for the junior model (default: depends on junior model) -#junior-edit-format: xxx +## Specify the edit format for the editor model (default: depends on editor model) +#editor-edit-format: xxx ## Only work with models that have meta-data available (default: True) #show-model-warnings: true diff --git a/aider/website/assets/sample.env b/aider/website/assets/sample.env index 556dd8221..12447e181 100644 --- a/aider/website/assets/sample.env +++ b/aider/website/assets/sample.env @@ -93,17 +93,17 @@ ## Specify what edit format the LLM should use (default depends on model) #AIDER_EDIT_FORMAT= -## Use senior edit format for the main chat -#AIDER_SENIOR= +## Use architect edit format for the main chat +#AIDER_ARCHITECT= ## Specify the model to use for commit messages and chat history summarization (default depends on --model) #AIDER_WEAK_MODEL= -## Specify the model to use for junior tasks (default depends on --model) -#AIDER_JUNIOR_MODEL= +## Specify the model to use for editor tasks (default depends on --model) +#AIDER_EDITOR_MODEL= -## Specify the edit format for the junior model (default: depends on junior model) -#AIDER_JUNIOR_EDIT_FORMAT= +## Specify the edit format for the editor model (default: depends on editor model) +#AIDER_EDITOR_EDIT_FORMAT= ## Only work with models that have meta-data available (default: True) #AIDER_SHOW_MODEL_WARNINGS=true diff --git a/aider/website/docs/benchmarks.md b/aider/website/docs/benchmarks.md index b7d68af96..3b7849328 100644 --- a/aider/website/docs/benchmarks.md +++ b/aider/website/docs/benchmarks.md @@ -55,7 +55,7 @@ about prompting GPT for complex tasks like coding. It's beneficial to minimize the "cognitive overhead" of formatting the response, allowing GPT to concentrate on the coding task at hand. -As a thought experiment, imagine a slack conversation with a junior developer where +As a thought experiment, imagine a slack conversation with a editor developer where you ask them to write the code to add some new feature to your app. They're going to type the response back to you by hand in the chat. Should they type out the diff --git a/aider/website/docs/config/adv-model-settings.md b/aider/website/docs/config/adv-model-settings.md index b0c6c9015..65c6a58dd 100644 --- a/aider/website/docs/config/adv-model-settings.md +++ b/aider/website/docs/config/adv-model-settings.md @@ -85,11 +85,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-3.5-turbo @@ -104,11 +104,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-3.5-turbo-0125 @@ -123,11 +123,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-3.5-turbo-1106 @@ -142,11 +142,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-3.5-turbo-0613 @@ -161,11 +161,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-3.5-turbo-16k-0613 @@ -180,11 +180,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: udiff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4-turbo-2024-04-09 @@ -199,11 +199,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: udiff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4-turbo @@ -218,11 +218,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: null lazy: true max_tokens: null name: openai/gpt-4o @@ -237,11 +237,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: openai/gpt-4o-2024-08-06 @@ -256,11 +256,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4o-2024-08-06 @@ -275,11 +275,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: null lazy: true max_tokens: null name: gpt-4o @@ -294,11 +294,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4o-mini @@ -313,11 +313,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: openai/gpt-4o-mini @@ -332,11 +332,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: udiff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4-0125-preview @@ -351,11 +351,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: udiff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: true max_tokens: null name: gpt-4-1106-preview @@ -370,11 +370,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-4-vision-preview @@ -389,11 +389,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-4-0314 @@ -408,11 +408,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-4-0613 @@ -427,11 +427,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gpt-4-32k-0613 @@ -446,11 +446,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: claude-3-opus-20240229 @@ -465,11 +465,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: openrouter/anthropic/claude-3-opus @@ -484,11 +484,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: claude-3-sonnet-20240229 @@ -503,12 +503,12 @@ cog.out("```\n") cache_control: true caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: claude-3-5-sonnet-20240620 examples_as_sys_msg: true extra_body: null extra_headers: anthropic-beta: prompt-caching-2024-07-31 - junior_edit_format: junior-diff - junior_model_name: claude-3-5-sonnet-20240620 lazy: false max_tokens: 8192 name: claude-3-5-sonnet-20240620 @@ -523,12 +523,12 @@ cog.out("```\n") cache_control: true caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: anthropic/claude-3-5-sonnet-20240620 examples_as_sys_msg: true extra_body: null extra_headers: anthropic-beta: prompt-caching-2024-07-31 - junior_edit_format: junior-diff - junior_model_name: anthropic/claude-3-5-sonnet-20240620 lazy: false max_tokens: 8192 name: anthropic/claude-3-5-sonnet-20240620 @@ -543,12 +543,12 @@ cog.out("```\n") cache_control: true caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: anthropic-beta: prompt-caching-2024-07-31 - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: anthropic/claude-3-haiku-20240307 @@ -563,12 +563,12 @@ cog.out("```\n") cache_control: true caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: anthropic-beta: prompt-caching-2024-07-31 - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: claude-3-haiku-20240307 @@ -583,11 +583,11 @@ cog.out("```\n") cache_control: true caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: openrouter/anthropic/claude-3.5-sonnet examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: openrouter/anthropic/claude-3.5-sonnet lazy: false max_tokens: 8192 name: openrouter/anthropic/claude-3.5-sonnet @@ -602,11 +602,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: vertex_ai/claude-3-5-sonnet@20240620 examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: vertex_ai/claude-3-5-sonnet@20240620 lazy: false max_tokens: 8192 name: vertex_ai/claude-3-5-sonnet@20240620 @@ -621,11 +621,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: vertex_ai/claude-3-opus@20240229 @@ -640,11 +640,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: vertex_ai/claude-3-sonnet@20240229 @@ -659,11 +659,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: command-r-plus @@ -678,11 +678,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: command-r-08-2024 @@ -697,11 +697,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: command-r-plus-08-2024 @@ -716,11 +716,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: groq/llama3-70b-8192 @@ -735,11 +735,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: openrouter/meta-llama/llama-3-70b-instruct @@ -754,11 +754,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-pro-002 @@ -773,11 +773,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-flash-002 @@ -792,11 +792,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff-fenced + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-pro @@ -811,11 +811,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff-fenced + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-pro-latest @@ -830,11 +830,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff-fenced + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-pro-exp-0827 @@ -849,11 +849,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: gemini/gemini-1.5-flash-exp-0827 @@ -868,11 +868,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: 8192 name: deepseek/deepseek-chat @@ -887,11 +887,11 @@ cog.out("```\n") cache_control: false caches_by_default: true edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: 8192 name: deepseek/deepseek-coder @@ -906,11 +906,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: 8192 name: deepseek-chat @@ -925,11 +925,11 @@ cog.out("```\n") cache_control: false caches_by_default: true edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: 8192 name: deepseek-coder @@ -944,11 +944,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: null + editor_model_name: null examples_as_sys_msg: true extra_body: null extra_headers: null - junior_edit_format: null - junior_model_name: null lazy: false max_tokens: null name: openrouter/deepseek/deepseek-coder @@ -963,11 +963,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: null examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: null lazy: true max_tokens: null name: openrouter/openai/gpt-4o @@ -982,11 +982,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: editor-diff + editor_model_name: openai/gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: openai/gpt-4o lazy: false max_tokens: null name: openai/o1-mini @@ -1001,11 +1001,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: editor-diff + editor_model_name: gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: gpt-4o lazy: false max_tokens: null name: o1-mini @@ -1020,11 +1020,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: openai/gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: openai/gpt-4o lazy: false max_tokens: null name: openai/o1-preview @@ -1038,12 +1038,12 @@ cog.out("```\n") - accepts_images: false cache_control: false caches_by_default: false - edit_format: senior + edit_format: architect + editor_edit_format: editor-diff + editor_model_name: gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: gpt-4o lazy: false max_tokens: null name: o1-preview @@ -1058,11 +1058,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: whole + editor_edit_format: editor-diff + editor_model_name: openrouter/openai/gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: openrouter/openai/gpt-4o lazy: false max_tokens: null name: openrouter/openai/o1-mini @@ -1077,11 +1077,11 @@ cog.out("```\n") cache_control: false caches_by_default: false edit_format: diff + editor_edit_format: editor-diff + editor_model_name: openrouter/openai/gpt-4o examples_as_sys_msg: false extra_body: null extra_headers: null - junior_edit_format: junior-diff - junior_model_name: openrouter/openai/gpt-4o lazy: false max_tokens: null name: openrouter/openai/o1-preview diff --git a/aider/website/docs/config/aider_conf.md b/aider/website/docs/config/aider_conf.md index f67fefdf8..b11a4249a 100644 --- a/aider/website/docs/config/aider_conf.md +++ b/aider/website/docs/config/aider_conf.md @@ -137,17 +137,17 @@ cog.outl("```") ## Specify what edit format the LLM should use (default depends on model) #edit-format: xxx -## Use senior edit format for the main chat -#senior: false +## Use architect edit format for the main chat +#architect: false ## Specify the model to use for commit messages and chat history summarization (default depends on --model) #weak-model: xxx -## Specify the model to use for junior tasks (default depends on --model) -#junior-model: xxx +## Specify the model to use for editor tasks (default depends on --model) +#editor-model: xxx -## Specify the edit format for the junior model (default: depends on junior model) -#junior-edit-format: xxx +## Specify the edit format for the editor model (default: depends on editor model) +#editor-edit-format: xxx ## Only work with models that have meta-data available (default: True) #show-model-warnings: true diff --git a/aider/website/docs/config/dotenv.md b/aider/website/docs/config/dotenv.md index 716ab707c..6f55ced83 100644 --- a/aider/website/docs/config/dotenv.md +++ b/aider/website/docs/config/dotenv.md @@ -135,17 +135,17 @@ cog.outl("```") ## Specify what edit format the LLM should use (default depends on model) #AIDER_EDIT_FORMAT= -## Use senior edit format for the main chat -#AIDER_SENIOR= +## Use architect edit format for the main chat +#AIDER_ARCHITECT= ## Specify the model to use for commit messages and chat history summarization (default depends on --model) #AIDER_WEAK_MODEL= -## Specify the model to use for junior tasks (default depends on --model) -#AIDER_JUNIOR_MODEL= +## Specify the model to use for editor tasks (default depends on --model) +#AIDER_EDITOR_MODEL= -## Specify the edit format for the junior model (default: depends on junior model) -#AIDER_JUNIOR_EDIT_FORMAT= +## Specify the edit format for the editor model (default: depends on editor model) +#AIDER_EDITOR_EDIT_FORMAT= ## Only work with models that have meta-data available (default: True) #AIDER_SHOW_MODEL_WARNINGS=true diff --git a/aider/website/docs/config/options.md b/aider/website/docs/config/options.md index 92bcd5c24..bcf5d4e1f 100644 --- a/aider/website/docs/config/options.md +++ b/aider/website/docs/config/options.md @@ -33,8 +33,8 @@ usage: aider [-h] [--openai-api-key] [--anthropic-api-key] [--model] [--openai-organization-id] [--model-settings-file] [--model-metadata-file] [--verify-ssl | --no-verify-ssl] [--edit-format] - [--senior] [--weak-model] [--junior-model] - [--junior-edit-format] + [--architect] [--weak-model] [--editor-model] + [--editor-edit-format] [--show-model-warnings | --no-show-model-warnings] [--max-chat-history-tokens] [--env-file] [--cache-prompts | --no-cache-prompts] @@ -196,21 +196,21 @@ Aliases: - `--edit-format EDIT_FORMAT` - `--chat-mode EDIT_FORMAT` -### `--senior` -Use senior edit format for the main chat -Environment variable: `AIDER_SENIOR` +### `--architect` +Use architect edit format for the main chat +Environment variable: `AIDER_ARCHITECT` ### `--weak-model WEAK_MODEL` Specify the model to use for commit messages and chat history summarization (default depends on --model) Environment variable: `AIDER_WEAK_MODEL` -### `--junior-model JUNIOR_MODEL` -Specify the model to use for junior tasks (default depends on --model) -Environment variable: `AIDER_JUNIOR_MODEL` +### `--editor-model JUNIOR_MODEL` +Specify the model to use for editor tasks (default depends on --model) +Environment variable: `AIDER_EDITOR_MODEL` -### `--junior-edit-format JUNIOR_EDIT_FORMAT` -Specify the edit format for the junior model (default: depends on junior model) -Environment variable: `AIDER_JUNIOR_EDIT_FORMAT` +### `--editor-edit-format JUNIOR_EDIT_FORMAT` +Specify the edit format for the editor model (default: depends on editor model) +Environment variable: `AIDER_EDITOR_EDIT_FORMAT` ### `--show-model-warnings` Only work with models that have meta-data available (default: True) diff --git a/benchmark/benchmark.py b/benchmark/benchmark.py index 0ff0900ef..2d21ff8e7 100755 --- a/benchmark/benchmark.py +++ b/benchmark/benchmark.py @@ -125,8 +125,8 @@ def main( graphs: bool = typer.Option(False, "--graphs", help="Generate graphs"), model: str = typer.Option("gpt-3.5-turbo", "--model", "-m", help="Model name"), edit_format: str = typer.Option(None, "--edit-format", "-e", help="Edit format"), - junior_model: str = typer.Option(None, "--junior-model", help="Junior model name"), - junior_edit_format: str = typer.Option(None, "--junior-edit-format", help="Junior edit format"), + editor_model: str = typer.Option(None, "--editor-model", help="Junior model name"), + editor_edit_format: str = typer.Option(None, "--editor-edit-format", help="Junior edit format"), replay: str = typer.Option( None, "--replay", @@ -245,8 +245,8 @@ def main( commit_hash, replay, max_apply_update_errors, - junior_model, - junior_edit_format, + editor_model, + editor_edit_format, ) all_results.append(results) @@ -266,8 +266,8 @@ def main( commit_hash, replay, max_apply_update_errors, - junior_model, - junior_edit_format, + editor_model, + editor_edit_format, ) all_results = run_test_threaded.gather(tqdm=True) @@ -378,7 +378,7 @@ def summarize_results(dirname): res.syntax_errors += results.get("syntax_errors", 0) res.indentation_errors += results.get("indentation_errors", 0) - for key in "model edit_format commit_hash junior_model junior_edit_format".split(): + for key in "model edit_format commit_hash editor_model editor_edit_format".split(): val = results.get(key) if val: variants[key].add(val) @@ -524,8 +524,8 @@ def run_test_real( commit_hash, replay, max_apply_update_errors, - junior_model, - junior_edit_format, + editor_model, + editor_edit_format, ): if not os.path.isdir(testdir): print("Not a dir:", testdir) @@ -585,8 +585,8 @@ def run_test_real( main_model = models.Model( model_name, weak_model=weak_model_name, - junior_model=junior_model, - junior_edit_format=junior_edit_format, + editor_model=editor_model, + editor_edit_format=editor_edit_format, ) edit_format = edit_format or main_model.edit_format @@ -700,8 +700,8 @@ def run_test_real( ) if edit_format == "senior": - results["junior_model"] = main_model.junior_model.name if main_model.junior_model else None - results["junior_edit_format"] = main_model.junior_edit_format + results["editor_model"] = main_model.editor_model.name if main_model.editor_model else None + results["editor_edit_format"] = main_model.editor_edit_format dump(results) results_fname.write_text(json.dumps(results, indent=4))
SeniorJuniorArchitectEditor Edit Format Pass Rate
{{ item.model }}{{ item.junior_model }}{{ item.junior_edit_format | default: item.edit_format }}{{ item.editor_model }}{{ item.editor_edit_format | default: item.edit_format }} {{ item.pass_rate_2 }}%