sonnet+deep got 60.9/82.0

aider/coders/ask_prompts.py

@@ -6,6 +6,7 @@ from .base_prompts import CoderPrompts
 class AskPrompts(CoderPrompts):
     main_system = """Act as an expert code analyst.
 Answer questions about the supplied code.
+Explain any needed code changes clearly but concisely.
 
 Always reply to the user in the same language they are using.
 """

aider/coders/wholefile_prompts.py

@@ -5,16 +5,8 @@ from .base_prompts import CoderPrompts
 
 class WholeFilePrompts(CoderPrompts):
     main_system = """Act as an expert software developer.
-Take requests for changes to the supplied code.
-If the request is ambiguous, ask questions.
-
-Always reply to the user in the same language they are using.
-
 {lazy_prompt}
-Once you understand the request you MUST:
-1. Determine if any code changes are needed.
-2. Explain any needed changes.
-3. If changes are needed, output a copy of each file that needs changes.
+Output a copy of each file that needs changes.
 """
 
     example_messages = [
@@ -24,10 +16,7 @@ Once you understand the request you MUST:
         ),
         dict(
             role="assistant",
-            content="""Ok, I will:
-
-1. Switch the greeting text from "Hello" to "Hey".
-
+            content="""
 show_greeting.py
 {fence[0]}
 import sys
@@ -42,7 +31,7 @@ if __name__ == '__main__':
         ),
     ]
 
-    system_reminder = """To suggest changes to a file you MUST return the entire content of the updated file.
+    system_reminder = """You MUST return the entire content of the updated file.
 You MUST use this *file listing* format:
 
 path/to/filename.js
@@ -57,7 +46,6 @@ Every *file listing* MUST use this format:
 - ... entire content of the file ...
 - Final line: closing {fence[1]}
 
-To suggest changes to a file you MUST return a *file listing* that contains the entire content of the file.
 *NEVER* skip, omit or elide content from a *file listing* using "..." or by adding comments like "... rest of code..."!
 Create a new file you MUST return a *file listing* which includes an appropriate filename, including any appropriate path.
 

benchmark/benchmark.py

@@ -549,18 +549,21 @@ def run_test_real(
         chat_history_file=history_fname,
     )
 
-    main_model = models.Model(model_name)
-    edit_format = edit_format or main_model.edit_format
+    # ask_model = models.Model("openrouter/anthropic/claude-3.5-sonnet")
+    # ask_model = models.Model("openrouter/anthropic/claude-3.5-sonnet")
+    ask_model = models.Model("openrouter/deepseek/deepseek-chat")
+    whole_model = models.Model("openrouter/deepseek/deepseek-chat")
+
+    main_model = ask_model
+    edit_format = "ask-whole"
 
     dump(main_model)
     dump(edit_format)
     show_fnames = ",".join(map(str, fnames))
     print("fnames:", show_fnames)
 
-    coder = Coder.create(
-        main_model,
-        edit_format,
-        io,
+    coder_kwargs = dict(
+        io=io,
         fnames=fnames,
         use_git=False,
         stream=False,
@@ -568,6 +571,12 @@ def run_test_real(
         # auto_lint=False,  # disabled for code-in-json experiments
         cache_prompts=True,
     )
+    coder = Coder.create(
+        main_model=ask_model,
+        edit_format="ask",
+        **coder_kwargs,
+    )
+
     coder.max_apply_update_errors = max_apply_update_errors
 
     timeouts = 0
@@ -592,7 +601,21 @@ def run_test_real(
 
             coder.apply_updates()
         else:
+            coder = Coder.create(
+                from_coder=coder,
+                main_model=ask_model,
+                edit_format="ask",
+                **coder_kwargs,
+            )
             response = coder.run(with_message=instructions, preproc=False)
+            coder = Coder.create(
+                from_coder=coder,
+                main_model=whole_model,
+                edit_format="whole",
+                **coder_kwargs,
+            )
+            response = coder.run(with_message="make those changes", preproc=False)
+
         dur += time.time() - start
 
         if not no_aider: