disable context manager

backend/agent/api.py

@@ -37,7 +37,8 @@ class AgentStartRequest(BaseModel):
     model_name: Optional[str] = "anthropic/claude-3-7-sonnet-latest"
     enable_thinking: Optional[bool] = False
     reasoning_effort: Optional[str] = 'low'
-    stream: Optional[bool] = False # Default stream to False for API
+    stream: Optional[bool] = True
+    enable_context_manager: Optional[bool] = False
 
 def initialize(
     _thread_manager: ThreadManager,
@@ -256,7 +257,7 @@ async def start_agent(
     user_id: str = Depends(get_current_user_id)
 ):
     """Start an agent for a specific thread in the background."""
-    logger.info(f"Starting new agent for thread: {thread_id} with config: model={body.model_name}, thinking={body.enable_thinking}, effort={body.reasoning_effort}, stream={body.stream}")
+    logger.info(f"Starting new agent for thread: {thread_id} with config: model={body.model_name}, thinking={body.enable_thinking}, effort={body.reasoning_effort}, stream={body.stream}, context_manager={body.enable_context_manager}")
     client = await db.client
     
     # Verify user has access to this thread
@@ -340,7 +341,8 @@ async def start_agent(
             model_name=MODEL_NAME_ALIASES.get(body.model_name, body.model_name), 
             enable_thinking=body.enable_thinking,
             reasoning_effort=body.reasoning_effort,
-            stream=body.stream # Pass stream parameter
+            stream=body.stream,
+            enable_context_manager=body.enable_context_manager
         )
     )
     
@@ -477,10 +479,11 @@ async def run_agent_background(
     model_name: str,
     enable_thinking: Optional[bool],
     reasoning_effort: Optional[str],
-    stream: bool # Add stream parameter
+    stream: bool,
+    enable_context_manager: bool
 ):
     """Run the agent in the background and handle status updates."""
-    logger.debug(f"Starting background agent run: {agent_run_id} for thread: {thread_id} (instance: {instance_id}) with model={model_name}, thinking={enable_thinking}, effort={reasoning_effort}, stream={stream}")
+    logger.debug(f"Starting background agent run: {agent_run_id} for thread: {thread_id} (instance: {instance_id}) with model={model_name}, thinking={enable_thinking}, effort={reasoning_effort}, stream={stream}, context_manager={enable_context_manager}")
     client = await db.client
     
     # Tracking variables
@@ -601,12 +604,13 @@ async def run_agent_background(
         agent_gen = run_agent(
             thread_id=thread_id,
             project_id=project_id,
-            stream=stream, # Pass stream parameter from API request
+            stream=stream,
             thread_manager=thread_manager,
             sandbox=sandbox,
-            model_name=model_name, # Pass model_name
-            enable_thinking=enable_thinking, # Pass enable_thinking
-            reasoning_effort=reasoning_effort # Pass reasoning_effort
+            model_name=model_name,
+            enable_thinking=enable_thinking,
+            reasoning_effort=reasoning_effort,
+            enable_context_manager=enable_context_manager
         )
         
         # Collect all responses to save to database

backend/agent/run.py

@@ -31,7 +31,8 @@ async def run_agent(
     max_iterations: int = 150,
     model_name: str = "anthropic/claude-3-7-sonnet-latest",
     enable_thinking: Optional[bool] = False,
-    reasoning_effort: Optional[str] = 'low'
+    reasoning_effort: Optional[str] = 'low',
+    enable_context_manager: bool = True
 ):
     """Run the development agent with specified configuration."""
     
@@ -142,7 +143,8 @@ async def run_agent(
             native_max_auto_continues=native_max_auto_continues,
             include_xml_examples=True,
             enable_thinking=enable_thinking,
-            reasoning_effort=reasoning_effort
+            reasoning_effort=reasoning_effort,
+            enable_context_manager=enable_context_manager
         )
             
         if isinstance(response, dict) and "status" in response and response["status"] == "error":
@@ -272,7 +274,8 @@ async def process_agent_response(
     stream: bool = True,
     model_name: str = "anthropic/claude-3-7-sonnet-latest",
     enable_thinking: Optional[bool] = False,
-    reasoning_effort: Optional[str] = 'low'
+    reasoning_effort: Optional[str] = 'low',
+    enable_context_manager: bool = True
 ):
     """Process the streaming response from the agent."""
     chunk_counter = 0
@@ -293,7 +296,8 @@ async def process_agent_response(
         native_max_auto_continues=25,
         model_name=model_name,
         enable_thinking=enable_thinking,
-        reasoning_effort=reasoning_effort
+        reasoning_effort=reasoning_effort,
+        enable_context_manager=enable_context_manager
     ):
         chunk_counter += 1
         

backend/agentpress/thread_manager.py

@@ -163,7 +163,8 @@ class ThreadManager:
         max_xml_tool_calls: int = 0,
         include_xml_examples: bool = False,
         enable_thinking: Optional[bool] = False,
-        reasoning_effort: Optional[str] = 'low'
+        reasoning_effort: Optional[str] = 'low',
+        enable_context_manager: bool = True
     ) -> Union[Dict[str, Any], AsyncGenerator]:
         """Run a conversation thread with LLM integration and tool execution.
         
@@ -183,6 +184,7 @@ class ThreadManager:
             include_xml_examples: Whether to include XML tool examples in the system prompt
             enable_thinking: Whether to enable thinking before making a decision
             reasoning_effort: The effort level for reasoning
+            enable_context_manager: Whether to enable automatic context summarization.
             
         Returns:
             An async generator yielding response chunks or error dict
@@ -249,7 +251,7 @@ Here are the XML tools available with examples:
                     logger.info(f"Thread {thread_id} token count: {token_count}/{token_threshold} ({(token_count/token_threshold)*100:.1f}%)")
                     
                     # If we're over the threshold, summarize the thread
-                    if token_count >= token_threshold:
+                    if token_count >= token_threshold and enable_context_manager:
                         logger.info(f"Thread token count ({token_count}) exceeds threshold ({token_threshold}), summarizing...")
                         
                         # Create summary using context manager
@@ -270,6 +272,8 @@ Here are the XML tools available with examples:
                             logger.info(f"After summarization: token count reduced from {token_count} to {new_token_count}")
                         else:
                             logger.warning("Summarization failed or wasn't needed - proceeding with original messages")
+                    else:
+                        logger.info("Automatic summarization disabled. Skipping token count check and summarization.")
                 except Exception as e:
                     logger.error(f"Error counting tokens or summarizing: {str(e)}")