fix: Add cache control headers to chat messages

aider/coders/base_coder.py

@@ -66,14 +66,39 @@ class ChatChunks:
         return (
             self.system
             + self.examples
-            + self.done
             + self.repo
             + self.readonly_files
+            + self.done
             + self.chat_files
             + self.cur
             + self.reminder
         )
 
+    def add_cache_control_headers(self):
+        if self.examples:
+            self.add_cache_control(self.examples)
+        else:
+            self.add_cache_control(self.system)
+
+        if self.readonly_files:
+            self.add_cache_control(self.readonly_files)
+        else:
+            self.add_cache_control(self.repo)
+
+        self.add_cache_control(self.chat_files)
+
+    def add_cache_control(self, messages):
+        if not messages:
+            return
+
+        content = messages[-1]
+        if type(content) is str:
+            content = dict(
+                type="text",
+                text=content,
+            )
+        content["cache_control"] = {"type": "ephemeral"}
+
 
 class Coder:
     abs_fnames = None
@@ -257,12 +282,12 @@ class Coder:
         total_cost=0.0,
         map_refresh="auto",
         cache_prompts=False,
-        cache_prompts=False,
     ):
         self.commit_before_message = []
         self.aider_commit_hashes = set()
         self.rejected_urls = set()
         self.abs_root_path_cache = {}
+        self.cache_prompts = cache_prompts
 
         if not fnames:
             fnames = []
@@ -991,6 +1016,9 @@ class Coder:
 
     def format_messages(self):
         chunks = self.format_chat_chunks()
+        if self.cache_prompts and self.main_model.cache_control:
+            chunks.add_cache_control_headers()
+
         return chunks.all_messages()
 
     def send_message(self, inp):

aider/models.py

@@ -74,7 +74,7 @@ class ModelSettings:
     examples_as_sys_msg: bool = False
     extra_headers: Optional[dict] = None
     max_tokens: Optional[int] = None
-    cache_control_enabled: bool = False
+    cache_control: bool = False
 
 
 # https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo
@@ -274,6 +274,7 @@ MODEL_SETTINGS = [
         extra_headers={
             "anthropic-beta": ANTHROPIC_BETA_HEADER,
         },
+        cache_control=True,
     ),
     ModelSettings(
         "anthropic/claude-3-5-sonnet-20240620",
@@ -285,6 +286,7 @@ MODEL_SETTINGS = [
         extra_headers={
             "anthropic-beta": ANTHROPIC_BETA_HEADER,
         },
+        cache_control=True,
     ),
     ModelSettings(
         "openrouter/anthropic/claude-3.5-sonnet",
@@ -296,8 +298,8 @@ MODEL_SETTINGS = [
         max_tokens=8192,
         extra_headers={
             "anthropic-beta": "max-tokens-3-5-sonnet-2024-07-15",
-            "HTTP-Referer": "https://aider.chat",
-            "X-Title": "Aider",
+            "HTTP-Referer": AIDER_SITE_URL,
+            "X-Title": AIDER_APP_NAME,
         },
     ),
     # Vertex AI Claude models