Configurable conversation limit now set to 10 exchanges. This helps when you want to manually continue a thread of thought across different models manually.

.env.example

@@ -92,6 +92,12 @@ DEFAULT_THINKING_MODE_THINKDEEP=high
 # Defaults to 3 hours if not specified
 CONVERSATION_TIMEOUT_HOURS=3
 
+# Optional: Max conversation turns
+# Maximum number of turns allowed in an AI-to-AI conversation thread
+# Each exchange (Claude asks, Gemini responds) counts as 2 turns
+# So 20 turns = 10 exchanges. Defaults to 20 if not specified
+MAX_CONVERSATION_TURNS=20
+
 # Optional: Logging level (DEBUG, INFO, WARNING, ERROR)
 # DEBUG: Shows detailed operational messages for troubleshooting (default)
 # INFO: Shows general operational messages

.gitignore

@@ -166,3 +166,6 @@ test_simulation_files/.claude/
 
 # Temporary test directories
 test-setup/
+
+# Scratch feature documentation files
+FEATURE_*.md

README.md

@@ -470,7 +470,7 @@ This server enables **true AI collaboration** between Claude and multiple AI mod
 - **Asynchronous workflow**: Conversations don't need to be sequential - Claude can work on tasks between exchanges, then return to Gemini with additional context and progress updates
 - **Incremental updates**: Share only new information in each exchange while maintaining full conversation history
 - **Automatic 25K limit bypass**: Each exchange sends only incremental context, allowing unlimited total conversation size
-- Up to 5 exchanges per conversation with 3-hour expiry (configurable)
+- Up to 10 exchanges per conversation (configurable via `MAX_CONVERSATION_TURNS`) with 3-hour expiry (configurable via `CONVERSATION_TIMEOUT_HOURS`)
 - Thread-safe with Redis persistence across all tools
 
 **Cross-tool & Cross-Model Continuation Example:**

docker-compose.yml

@@ -41,6 +41,7 @@ services:
       - DEFAULT_MODEL=${DEFAULT_MODEL:-auto}
       - DEFAULT_THINKING_MODE_THINKDEEP=${DEFAULT_THINKING_MODE_THINKDEEP:-high}
       - CONVERSATION_TIMEOUT_HOURS=${CONVERSATION_TIMEOUT_HOURS:-3}
+      - MAX_CONVERSATION_TURNS=${MAX_CONVERSATION_TURNS:-20}
       # Model usage restrictions
       - OPENAI_ALLOWED_MODELS=${OPENAI_ALLOWED_MODELS:-}
       - GOOGLE_ALLOWED_MODELS=${GOOGLE_ALLOWED_MODELS:-}

tests/test_claude_continuation.py

@@ -146,8 +146,8 @@ class TestClaudeContinuationOffers:
             # Should still offer continuation since turns remain
             assert response_data["status"] == "continuation_available"
             assert "continuation_offer" in response_data
-            # 10 max - 2 existing - 1 new = 7 remaining
-            assert response_data["continuation_offer"]["remaining_turns"] == 7
+            # MAX_CONVERSATION_TURNS - 2 existing - 1 new = remaining
+            assert response_data["continuation_offer"]["remaining_turns"] == MAX_CONVERSATION_TURNS - 3
 
     @patch("utils.conversation_memory.get_redis_client")
     @patch.dict("os.environ", {"PYTEST_CURRENT_TEST": ""}, clear=False)
@@ -267,10 +267,10 @@ I'd be happy to examine the error handling patterns in more detail if that would
             # Parse response
             response_data = json.loads(response[0].text)
 
-            # Should offer continuation since there are remaining turns (9 remaining: 10 max - 0 current - 1)
+            # Should offer continuation since there are remaining turns (MAX - 0 current - 1)
             assert response_data["status"] == "continuation_available"
             assert response_data.get("continuation_offer") is not None
-            assert response_data["continuation_offer"]["remaining_turns"] == 9
+            assert response_data["continuation_offer"]["remaining_turns"] == MAX_CONVERSATION_TURNS - 1
 
     @patch("utils.conversation_memory.get_redis_client")
     @patch.dict("os.environ", {"PYTEST_CURRENT_TEST": ""}, clear=False)
@@ -465,8 +465,8 @@ class TestContinuationIntegration:
             # Should still offer continuation if there are remaining turns
             assert response_data2["status"] == "continuation_available"
             assert "continuation_offer" in response_data2
-            # 10 max - 1 existing - 1 new = 8 remaining
-            assert response_data2["continuation_offer"]["remaining_turns"] == 8
+            # MAX_CONVERSATION_TURNS - 1 existing - 1 new = remaining
+            assert response_data2["continuation_offer"]["remaining_turns"] == MAX_CONVERSATION_TURNS - 2
 
 
 if __name__ == "__main__":

utils/conversation_memory.py

@@ -57,7 +57,17 @@ from pydantic import BaseModel
 logger = logging.getLogger(__name__)
 
 # Configuration constants
-MAX_CONVERSATION_TURNS = 10  # Maximum turns allowed per conversation thread
+# Get max conversation turns from environment, default to 20 turns (10 exchanges)
+try:
+    MAX_CONVERSATION_TURNS = int(os.getenv("MAX_CONVERSATION_TURNS", "20"))
+    if MAX_CONVERSATION_TURNS <= 0:
+        logger.warning(f"Invalid MAX_CONVERSATION_TURNS value ({MAX_CONVERSATION_TURNS}), using default of 20 turns")
+        MAX_CONVERSATION_TURNS = 20
+except ValueError:
+    logger.warning(
+        f"Invalid MAX_CONVERSATION_TURNS value ('{os.getenv('MAX_CONVERSATION_TURNS')}'), using default of 20 turns"
+    )
+    MAX_CONVERSATION_TURNS = 20
 
 # Get conversation timeout from environment (in hours), default to 3 hours
 try: