Upload 2 files

src/agent.py

@@ -10,7 +10,7 @@ from src.prompts import Prompts  # Import system prompts
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
 
 class Agent:
-    def __init__(self, llm: Llama, db_path: str, system_prompt: str = "", max_tokens: int = 512, temperature: float = 0.7, top_p: float = 0.95):
+    def __init__(self, llm: Llama, db_path: str, system_prompt: str = "", max_tokens: int = 500, temperature: float = 0.7, top_p: float = 0.95):
         self.llm = llm
         self.memory = MemoryManager(db_path)
         self.prompts = {
@@ -122,7 +122,7 @@ class Agent:
         full_context = " ".join([memory['description'] for memory, _, _ in all_memories])
 
         # Truncate the context if it exceeds the token limit
-        max_context_length = 30000  # Adjust this based on your LLM's token limit
+        max_context_length = 500  # Adjust this based on your LLM's token limit
         if len(full_context) > max_context_length:
             full_context = full_context[:max_context_length]
             logging.info(f"Truncated full context to {max_context_length} characters.")
@@ -147,7 +147,7 @@ class Agent:
         combined_context = f"Initial Response: {initial_response}\nHigh-Level Summary: {high_level_summary}"
 
         # Truncate the combined context to fit within the model's context window
-        max_context_length = 30000  # Adjust this based on your LLM's token limit
+        max_context_length = 500  # Adjust this based on your LLM's token limit
         if len(combined_context) > max_context_length:
             combined_context = combined_context[:max_context_length]
             logging.info(f"Truncated combined context to {max_context_length} characters.")

src/main.py

@@ -34,13 +34,13 @@ def download_model(repo_id, filename, save_path):
 
 # Download the model if it doesn't exist
 if not os.path.exists(model_path):
-    download_model("PurpleAILAB/Llama3.2-3B-uncensored-SQLi-Q4_K_M-GGUF", filename, model_path)
+    download_model("adeptusnull/llama3.2-1b-wizardml-vicuna-uncensored-finetune-test", filename, model_path)
 
 def main():
     model_path = "models/unsloth.Q4_K_M.gguf"  # Path to the downloaded model
     db_path = "agent.db"
     system_prompt = "Vous êtes l'assistant intelligent de Les Chronique MTC. Votre rôle est d'aider les visiteurs en expliquant le contenu des Chroniques, Flash Infos et Chronique-FAQ de Michel Thomas. Utilisez le contexte fourni pour améliorer vos réponses et veillez à ce qu'elles soient précises et pertinentes."
-    max_tokens = 512
+    max_tokens = 500
     temperature = 0.7
     top_p = 0.95
 
@@ -53,7 +53,7 @@ def main():
     # Load the model
     llm = Llama(
         model_path=model_path,
-        n_ctx=5072,  # Set the maximum context length
+        n_ctx=572,  # Set the maximum context length
         max_tokens=max_tokens  # Control the maximum number of tokens generated in the response
     )