refactor : memory clear section function logic

server/app.py

@@ -6,7 +6,6 @@ from flask import Flask, jsonify, request
 
 from sources.llamacpp_handler import LlamacppLLM
 from sources.ollama_handler import OllamaLLM
-from sources.vllm_handler import Vllm
 
 parser = argparse.ArgumentParser(description='AgenticSeek server script')
 parser.add_argument('--provider', type=str, help='LLM backend library to use. set to [ollama], [vllm] or [llamacpp]', required=True)
@@ -15,12 +14,11 @@ args = parser.parse_args()
 
 app = Flask(__name__)
 
-assert args.provider in ["ollama", "llamacpp", "vllm"], f"Provider {args.provider} does not exists. see --help for more information"
+assert args.provider in ["ollama", "llamacpp"], f"Provider {args.provider} does not exists. see --help for more information"
 
 handler_map = {
     "ollama": OllamaLLM(),
     "llamacpp": LlamacppLLM(),
-    "vllm": Vllm()
 }
 
 generator = handler_map[args.provider]

server/sources/vllm_handler.py

@@ -1,74 +0,0 @@
-from .generator import GeneratorLLM
-from vllm import LLM, SamplingParams
-import logging
-from typing import List, Dict
-
-class Vllm(GeneratorLLM):
-    def __init__(self):
-        """
-        Handle generation using vLLM.
-        """
-        super().__init__()
-        self.logger = logging.getLogger(__name__)
-        self.llm = None 
-        
-    def convert_history_to_prompt(self, history: List[Dict[str, str]]) -> str:
-        """
-        Convert OpenAI-format history to a single prompt string for vLLM.
-        """
-        prompt = ""
-        for message in history:
-            role = message["role"]
-            content = message["content"]
-            if role == "system":
-                prompt += f"System: {content}\n"
-            elif role == "user":
-                prompt += f"User: {content}\n"
-            elif role == "assistant":
-                prompt += f"Assistant: {content}\n"
-        prompt += "Assistant: "
-        return prompt
-
-    def generate(self, history: List[Dict[str, str]]):
-        """
-        Generate response using vLLM from OpenAI-format message history.
-        
-        Args:
-            history: List of dictionaries in OpenAI format [{"role": "user", "content": "..."}, ...]
-        """
-        self.logger.info(f"Using {self.model} for generation with vLLM")
-        if self.llm is None:
-            self.llm = LLM(model=self.model)
-        
-        try:
-            with self.state.lock:
-                self.state.is_generating = True
-                self.state.last_complete_sentence = ""
-                self.state.current_buffer = ""
-
-            prompt = self.convert_history_to_prompt(history)
-            
-            sampling_params = SamplingParams(
-                temperature=0.7,
-                max_tokens=512,
-                gpu_memory_utilization=0.5,
-                stream=True  # Enable streaming
-            )
-            outputs = self.llm.generate(prompt, sampling_params, use_tqdm=False)
-            for output in outputs:
-                content = output.outputs[0].text
-                with self.state.lock:
-                    if '.' in content:
-                        self.logger.info(self.state.current_buffer)
-                    self.state.current_buffer += content
-            with self.state.lock:
-                self.logger.info(f"Final output: {self.state.current_buffer}")
-
-        except Exception as e:
-            self.logger.error(f"Error during generation: {str(e)}")
-            raise e
-            
-        finally:
-            self.logger.info("Generation complete")
-            with self.state.lock:
-                self.state.is_generating = False

sources/agents/browser_agent.py

@@ -357,7 +357,7 @@ class BrowserAgent(Agent):
         mem_last_idx = self.memory.push('user', prompt)
         answer, reasoning = self.llm_request()
         pretty_print(answer, color="output")
-        self.memory.clear_section(mem_begin_idx, mem_last_idx)
+        self.memory.clear_section(mem_begin_idx+1, mem_last_idx-1)
         return answer, reasoning
 
 if __name__ == "__main__":

sources/browser.py

@@ -58,7 +58,7 @@ def create_driver(headless=False, stealth_mode=True) -> webdriver.Chrome:
         chrome_options.add_argument("--disable-gpu")
         chrome_options.add_argument("--disable-webgl")
     #ua = UserAgent()
-    #user_agent = ua.random # NOTE sometime return wrong user agent, investigate
+    #user_agent = ua.random # NOTE sometime return bad ua crash, investigate
     #chrome_options.add_argument(f'user-agent={user_agent}')
     user_data_dir = tempfile.mkdtemp()
     chrome_options.add_argument(f"--user-data-dir={user_data_dir}")

sources/memory.py

@@ -106,13 +106,23 @@ class Memory():
         return curr_idx-1
     
     def clear(self) -> None:
+        """Clear all memory except system prompt"""
         self.logger.info("Memory clear performed.")
-        self.memory = []
+        self.memory = self.memory[:1]
     
     def clear_section(self, start: int, end: int) -> None:
-        """Clear a section of the memory."""
+        """
-        self.logger.info(f"Memory section {start} to {end} cleared.")
+        Clear a section of the memory. Ignore system message index.
+        Args:
+            start (int): Starting bound of the section to clear.
+            end (int): Ending bound of the section to clear.
+        """
+        self.logger.info(f"Clearing memory section {start} to {end}.")
+        start = max(0, start) + 1
+        end = min(end, len(self.memory)-1) + 2
+        self.logger.info(f"Memory before: {self.memory}")
         self.memory = self.memory[:start] + self.memory[end:]
+        self.logger.info(f"Memory after: {self.memory}")
     
     def get(self) -> list:
         return self.memory
@@ -172,7 +182,10 @@ if __name__ == "__main__":
     sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
     memory = Memory("You are a helpful assistant.",
                     recover_last_session=False, memory_compression=True)
-    
+
+    memory.push('user', "hello")
+    memory.push('assistant', "how can i help you?")
+    memory.push('user', "why do i get this cuda error?")
     sample_text = """
 The error you're encountering:
 cuda.cu:52:10: fatal error: helper_functions.h: No such file or directory
@@ -190,11 +203,8 @@ Use #include "helper_functions.h" for local files.
 Use the -I flag to specify the directory containing helper_functions.h.
 Ensure the file exists in the specified location.
     """
-    
-    memory.push('user', "hello")
-    memory.push('assistant', "how can i help you?")
-    memory.push('user', "why do i get this cuda error?")
     memory.push('assistant', sample_text)
+    
     print("\n---\nmemory before:", memory.get())
     memory.compress()
     print("\n---\nmemory after:", memory.get())