fixes

ai_client.py

@@ -1,4 +1,4 @@
-# ai_client.py
+# ai_client.py
 import tomllib
 import json
 import datetime
@@ -318,7 +318,7 @@ def _send_gemini(md_content: str, user_message: str, base_dir: str) -> str:
         full_message = f"<context>\n{md_content}\n</context>\n\n{user_message}"
 
         _append_comms("OUT", "request", {
-            "message": full_message,
+            "summary": f"[Gemini: context {len(md_content)} chars + user message {len(user_message)} chars]",
         })
 
         response = _gemini_chat.send_message(full_message)
@@ -546,6 +546,21 @@ def _build_files_context_blocks(
     return blocks
 
 
+
+def _strip_cache_controls(history: list[dict]):
+    """
+    Remove cache_control from all content blocks in the message history.
+    Anthropic allows a maximum of 4 cache_control blocks total across
+    system + tools + messages.  We reserve those slots for the stable
+    system/tools prefix and the current turn's context block, so all
+    older history entries must be clean.
+    """
+    for msg in history:
+        content = msg.get("content")
+        if isinstance(content, list):
+            for block in content:
+                if isinstance(block, dict):
+                    block.pop("cache_control", None)
 def _run_anthropic_loop(
     user_content: list[dict],
     user_message: str,
@@ -559,6 +574,7 @@ def _run_anthropic_loop(
     """
     global _anthropic_history
 
+    _strip_cache_controls(_anthropic_history)
     _anthropic_history.append({"role": "user", "content": user_content})
 
     _append_comms("OUT", "request", {
@@ -779,4 +795,5 @@ def send(
         return _send_gemini(md_content, user_message, base_dir)
     elif _provider == "anthropic":
         return _send_anthropic(md_content, user_message, base_dir, file_items)
-    raise ValueError(f"unknown provider: {_provider}")
+    raise ValueError(f"unknown provider: {_provider}")
+