npcpy 1.2.18__py3-none-any.whl → 1.2.20__py3-none-any.whl

npcpy/ft/memory_trainer.py

@@ -1,164 +1,171 @@
-import torch
-import torch.nn as nn
 try:
-    from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
     from torch.utils.data import Dataset
-except:
-    pass 
-import json
-from typing import List, Dict, Tuple
-import random
+    import torch
+    import torch.nn as nn
+    from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
 
-class MemoryDataset(Dataset):
-    def __init__(self, examples: List[Dict], tokenizer, max_length=512):
-        self.examples = examples
-        self.tokenizer = tokenizer
-        self.max_length = max_length
-    
-    def __len__(self):
-        return len(self.examples)
-    
-    def __getitem__(self, idx):
-        example = self.examples[idx]
-        
-        
-        text = f"Memory: {example['memory']}\nContext: {example.get('context', '')}"
-        
-        encoding = self.tokenizer(
-            text,
-            truncation=True,
-            padding='max_length',
-            max_length=self.max_length,
-            return_tensors='pt'
-        )
-        
-        return {
-            'input_ids': encoding['input_ids'].flatten(),
-            'attention_mask': encoding['attention_mask'].flatten(),
-            'labels': torch.tensor(example['label'], dtype=torch.long)
-        }
+    import json
+    from typing import List, Dict, Tuple
+    import random
 
-class MemoryTrainer:
-    def __init__(self, model_name="google/gemma-2b", device="cpu"):
-        self.device = device
-        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-        if self.tokenizer.pad_token is None:
-            self.tokenizer.pad_token = self.tokenizer.eos_token
-        
-        
-        self.model = AutoModelForSequenceClassification.from_pretrained(
-            model_name, 
-            num_labels=3
-        ).to(device)
-    
-    def prepare_training_data(self, approved_memories: List[Dict], 
-                            rejected_memories: List[Dict]) -> List[Dict]:
-        """Prepare training data from memory examples"""
-        examples = []
-        
-        
-        for memory in approved_memories:
-            examples.append({
-                "memory": memory.get("final_memory") or memory.get("initial_memory"),
-                "context": memory.get("context", ""),
-                "label": 1
-            })
-        
-        
-        for memory in rejected_memories:
-            examples.append({
-                "memory": memory.get("initial_memory"),
-                "context": memory.get("context", ""),
-                "label": 0
-            })
-        
-        
-        edited_examples = []
-        for memory in approved_memories[:len(rejected_memories)//2]:
-            if memory.get("final_memory") and memory.get("initial_memory"):
-                
-                edited_examples.append({
+    class MemoryDataset(Dataset):
+        def __init__(self, examples: List[Dict], tokenizer, max_length=512):
+            self.examples = examples
+            self.tokenizer = tokenizer
+            self.max_length = max_length
+        
+        def __len__(self):
+            return len(self.examples)
+        
+        def __getitem__(self, idx):
+            example = self.examples[idx]
+            
+            
+            text = f"Memory: {example['memory']}\nContext: {example.get('context', '')}"
+            
+            encoding = self.tokenizer(
+                text,
+                truncation=True,
+                padding='max_length',
+                max_length=self.max_length,
+                return_tensors='pt'
+            )
+            
+            return {
+                'input_ids': encoding['input_ids'].flatten(),
+                'attention_mask': encoding['attention_mask'].flatten(),
+                'labels': torch.tensor(example['label'], dtype=torch.long)
+            }
+
+    class MemoryTrainer:
+        def __init__(self, model_name="google/gemma-2b", device="cpu"):
+            self.device = device
+            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
+            if self.tokenizer.pad_token is None:
+                self.tokenizer.pad_token = self.tokenizer.eos_token
+            
+            
+            self.model = AutoModelForSequenceClassification.from_pretrained(
+                model_name, 
+                num_labels=3
+            ).to(device)
+        
+        def prepare_training_data(self, approved_memories: List[Dict], 
+                                rejected_memories: List[Dict]) -> List[Dict]:
+            """Prepare training data from memory examples"""
+            examples = []
+            
+            
+            for memory in approved_memories:
+                examples.append({
+                    "memory": memory.get("final_memory") or memory.get("initial_memory"),
+                    "context": memory.get("context", ""),
+                    "label": 1
+                })
+            
+            
+            for memory in rejected_memories:
+                examples.append({
                     "memory": memory.get("initial_memory"),
                     "context": memory.get("context", ""),
-                    "label": 2
+                    "label": 0
                 })
-        
-        examples.extend(edited_examples)
-        random.shuffle(examples)
-        return examples
+            
+            
+            edited_examples = []
+            for memory in approved_memories[:len(rejected_memories)//2]:
+                if memory.get("final_memory") and memory.get("initial_memory"):
+                    
+                    edited_examples.append({
+                        "memory": memory.get("initial_memory"),
+                        "context": memory.get("context", ""),
+                        "label": 2
+                    })
+            
+            examples.extend(edited_examples)
+            random.shuffle(examples)
+            return examples
 
-    def train(self, approved_memories: List[Dict], rejected_memories: List[Dict], 
-              output_dir: str = "./memory_model", epochs: int = 3):
-        """Train the memory classification model"""
-        
-        if len(approved_memories) < 10 or len(rejected_memories) < 10:
-            print("Not enough training data. Need at least 10 approved and 10 rejected memories.")
-            return False
-        
-        training_data = self.prepare_training_data(approved_memories, rejected_memories)
-        
-        
-        split_idx = int(0.8 * len(training_data))
-        train_data = training_data[:split_idx]
-        val_data = training_data[split_idx:]
-        
-        train_dataset = MemoryDataset(train_data, self.tokenizer)
-        val_dataset = MemoryDataset(val_data, self.tokenizer)
-        
-        training_args = TrainingArguments(
-            output_dir=output_dir,
-            num_train_epochs=epochs,
-            per_device_train_batch_size=4,
-            per_device_eval_batch_size=4,
-            warmup_steps=100,
-            weight_decay=0.01,
-            logging_dir='./logs',
-            evaluation_strategy="epoch",
-            save_strategy="epoch",
-            load_best_model_at_end=True,
-        )
-        
-        trainer = Trainer(
-            model=self.model,
-            args=training_args,
-            train_dataset=train_dataset,
-            eval_dataset=val_dataset,
-        )
-        
-        trainer.train()
-        trainer.save_model()
-        self.tokenizer.save_pretrained(output_dir)
-        
-        print(f"Model trained and saved to {output_dir}")
-        return True
+        def train(self, approved_memories: List[Dict], rejected_memories: List[Dict], 
+                output_dir: str = "./memory_model", epochs: int = 3):
+            """Train the memory classification model"""
+            
+            if len(approved_memories) < 10 or len(rejected_memories) < 10:
+                print("Not enough training data. Need at least 10 approved and 10 rejected memories.")
+                return False
+            
+            training_data = self.prepare_training_data(approved_memories, rejected_memories)
+            
+            
+            split_idx = int(0.8 * len(training_data))
+            train_data = training_data[:split_idx]
+            val_data = training_data[split_idx:]
+            
+            train_dataset = MemoryDataset(train_data, self.tokenizer)
+            val_dataset = MemoryDataset(val_data, self.tokenizer)
+            
+            training_args = TrainingArguments(
+                output_dir=output_dir,
+                num_train_epochs=epochs,
+                per_device_train_batch_size=4,
+                per_device_eval_batch_size=4,
+                warmup_steps=100,
+                weight_decay=0.01,
+                logging_dir='./logs',
+                evaluation_strategy="epoch",
+                save_strategy="epoch",
+                load_best_model_at_end=True,
+            )
+            
+            trainer = Trainer(
+                model=self.model,
+                args=training_args,
+                train_dataset=train_dataset,
+                eval_dataset=val_dataset,
+            )
+            
+            trainer.train()
+            trainer.save_model()
+            self.tokenizer.save_pretrained(output_dir)
+            
+            print(f"Model trained and saved to {output_dir}")
+            return True
 
-    def predict_memory_action(self, memory_content: str, context: str = "") -> Tuple[str, float]:
-        """Predict what action to take on a memory"""
-        text = f"Memory: {memory_content}\nContext: {context}"
-        
-        encoding = self.tokenizer(
-            text,
-            truncation=True,
-            padding=True,
-            max_length=512,
-            return_tensors='pt'
-        ).to(self.device)
-        
-        with torch.no_grad():
-            outputs = self.model(**encoding)
-            probabilities = torch.softmax(outputs.logits, dim=-1)
-            predicted_class = torch.argmax(probabilities, dim=-1).item()
-            confidence = probabilities[0][predicted_class].item()
-        
-        actions = {0: "model-rejected", 1: "model-approved", 2: "needs-editing"}
-        return actions[predicted_class], confidence
+        def predict_memory_action(self, memory_content: str, context: str = "") -> Tuple[str, float]:
+            """Predict what action to take on a memory"""
+            text = f"Memory: {memory_content}\nContext: {context}"
+            
+            encoding = self.tokenizer(
+                text,
+                truncation=True,
+                padding=True,
+                max_length=512,
+                return_tensors='pt'
+            ).to(self.device)
+            
+            with torch.no_grad():
+                outputs = self.model(**encoding)
+                probabilities = torch.softmax(outputs.logits, dim=-1)
+                predicted_class = torch.argmax(probabilities, dim=-1).item()
+                confidence = probabilities[0][predicted_class].item()
+            
+            actions = {0: "model-rejected", 1: "model-approved", 2: "needs-editing"}
+            return actions[predicted_class], confidence
 
-    def auto_approve_memory(self, memory_content: str, context: str = "", 
-                          confidence_threshold: float = 0.8) -> Dict:
-        """Auto-approve memory if confidence is high enough"""
-        action, confidence = self.predict_memory_action(memory_content, context)
-        
-        if confidence >= confidence_threshold:
-            return {"action": action, "confidence": confidence, "auto_processed": True}
-        else:
-            return {"action": "pending_approval", "confidence": confidence, "auto_processed": False}
+        def auto_approve_memory(self, memory_content: str, context: str = "", 
+                            confidence_threshold: float = 0.8) -> Dict:
+            """Auto-approve memory if confidence is high enough"""
+            action, confidence = self.predict_memory_action(memory_content, context)
+            
+            if confidence >= confidence_threshold:
+                return {"action": action, "confidence": confidence, "auto_processed": True}
+            else:
+                return {"action": "pending_approval", "confidence": confidence, "auto_processed": False}    
+except:
+    Dataset = None
+    nn = None
+    Trainer = None
+    TrainingArguments = None
+
+    MemoryDataset = None
+    MemoryTrainer = None

npcpy/memory/memory_processor.py

@@ -17,96 +17,6 @@ class MemoryItem:
     model: str
     provider: str
 
-class MemoryApprovalQueue:
-    def __init__(self, command_history):
-        self.command_history = command_history
-        self.pending_queue = queue.Queue()
-        self.approval_results = queue.Queue()
-        self.processing_thread = None
-        self.running = False
-
-    def add_memory(self, memory_item: MemoryItem):
-        """Add memory to processing queue (non-blocking)"""
-        self.pending_queue.put(memory_item)
-        
-    def start_background_processing(self):
-        """Start background thread for memory processing"""
-        if self.processing_thread and self.processing_thread.is_alive():
-            return
-            
-        self.running = True
-        self.processing_thread = threading.Thread(target=self._process_queue)
-        self.processing_thread.daemon = True
-        self.processing_thread.start()
-
-    def _process_queue(self):
-        """Background processing of memory queue"""
-        while self.running:
-            try:
-                
-                batch = []
-                try:
-                    
-                    memory = self.pending_queue.get(timeout=1.0)
-                    batch.append(memory)
-                    
-                    
-                    while len(batch) < 10:
-                        try:
-                            memory = self.pending_queue.get_nowait()
-                            batch.append(memory)
-                        except queue.Empty:
-                            break
-                    
-                    self._process_memory_batch(batch)
-                    
-                except queue.Empty:
-                    continue
-                    
-            except Exception as e:
-                print(f"Error in memory processing: {e}")
-                time.sleep(1)
-
-    def _process_memory_batch(self, memories: List[MemoryItem]):
-        """Process a batch of memories"""
-        for memory in memories:
-            
-            memory_id = self.command_history.add_memory_to_database(
-                message_id=memory.message_id,
-                conversation_id=memory.conversation_id,
-                npc=memory.npc,
-                team=memory.team,
-                directory_path=memory.directory_path,
-                initial_memory=memory.content,
-                status="pending_approval",
-                model=memory.model,
-                provider=memory.provider
-            )
-            
-            
-            self.approval_results.put({
-                "memory_id": memory_id,
-                "content": memory.content,
-                "context": memory.context,
-                "npc": memory.npc
-            })
-
-    def get_approval_batch(self, max_items: int = 5) -> List[Dict]:
-        """Get batch of memories ready for approval"""
-        batch = []
-        try:
-            while len(batch) < max_items:
-                item = self.approval_results.get_nowait()
-                batch.append(item)
-        except queue.Empty:
-            pass
-        return batch
-
-    def stop_processing(self):
-        """Stop background processing"""
-        self.running = False
-        if self.processing_thread:
-            self.processing_thread.join(timeout=2.0)
 
 def memory_approval_ui(memories: List[Dict]) -> List[Dict]:
     """Simple CLI interface for memory approval"""

{npcpy-1.2.18.dist-info → npcpy-1.2.20.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: npcpy
-Version: 1.2.18
+Version: 1.2.20
 Summary: npcpy is the premier open-source library for integrating LLMs and Agents into python systems.
 Home-page: https://github.com/NPC-Worldwide/npcpy
 Author: Christopher Agostino

{npcpy-1.2.18.dist-info → npcpy-1.2.20.dist-info}/RECORD

@@ -17,7 +17,7 @@ npcpy/data/web.py,sha256=ARGoVKUlQmaiX0zJbSvvFmRCwOv_Z7Pcan9c5GxYObQ,5117
 npcpy/ft/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 npcpy/ft/diff.py,sha256=R3Qo6v0-6M1iI0wiXhUzyuYI2ja0q_0i9bE0z3coxzU,28
 npcpy/ft/ge.py,sha256=my5LtGyVTT40V0i1h9FR-tFFA1FHSga-PeCCgUX1UUI,61
-npcpy/ft/memory_trainer.py,sha256=Ne-qoDnCUl-tRQ_iSZjsUy8C8_-FoL89I9ovo7jqCBU,5896
+npcpy/ft/memory_trainer.py,sha256=QZPznxEEwXbOGroHdMUMa5xpqlNwgV6nqOazI2xgrnQ,6635
 npcpy/ft/rl.py,sha256=l3RUkEJe4b2yB6pildveu2LJymtNq0F17COwf_CCq3U,34
 npcpy/ft/sft.py,sha256=i4ENygRPArbLWN4XZZuBnPWaehs8M-J68JB_mewGJHI,62
 npcpy/gen/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -30,7 +30,7 @@ npcpy/memory/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 npcpy/memory/command_history.py,sha256=Ww7vZTSjQDuElQXuOjsvu7NTljOLAg07QIFrfKARpVg,45562
 npcpy/memory/kg_vis.py,sha256=TrQQCRh_E7Pyr-GPAHLSsayubAfGyf4HOEFrPB6W86Q,31280
 npcpy/memory/knowledge_graph.py,sha256=2XpIlsyPdAOnzQ6kkwP6MWPGwL3P6V33_3suNJYMMJE,48681
-npcpy/memory/memory_processor.py,sha256=nK5vWTLeevAQDXM15wF9MSTK6Yr-qYCjgiwmCKXWTSI,5288
+npcpy/memory/memory_processor.py,sha256=bLfzT-uDgwNegs1hVBqW3Hl2fYtdmFQbdc5To_f4i5E,2106
 npcpy/memory/search.py,sha256=glN6WYzaixcoDphTEHAXSMX3vKZGjR12Jx9YVL_gYfE,18433
 npcpy/mix/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 npcpy/mix/debate.py,sha256=lQXxC7nl6Rwyf7HIYrsVQILMUmYYx55Tjt2pkTg56qY,9019
@@ -41,8 +41,8 @@ npcpy/work/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 npcpy/work/desktop.py,sha256=F3I8mUtJp6LAkXodsh8hGZIncoads6c_2Utty-0EdDA,2986
 npcpy/work/plan.py,sha256=QyUwg8vElWiHuoS-xK4jXTxxHvkMD3VkaCEsCmrEPQk,8300
 npcpy/work/trigger.py,sha256=P1Y8u1wQRsS2WACims_2IdkBEar-iBQix-2TDWoW0OM,9948
-npcpy-1.2.18.dist-info/licenses/LICENSE,sha256=j0YPvce7Ng9e32zYOu0EmXjXeJ0Nwawd0RA3uSGGH4E,1070
-npcpy-1.2.18.dist-info/METADATA,sha256=uewXn2ASHxuVuCjL3UmFjyVUCV6LuWaEsSmx1uogluo,26084
-npcpy-1.2.18.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-npcpy-1.2.18.dist-info/top_level.txt,sha256=g1pbSvrOOncB74Bg5-J0Olg4V0A5VzDw-Xz5YObq8BU,6
-npcpy-1.2.18.dist-info/RECORD,,
+npcpy-1.2.20.dist-info/licenses/LICENSE,sha256=j0YPvce7Ng9e32zYOu0EmXjXeJ0Nwawd0RA3uSGGH4E,1070
+npcpy-1.2.20.dist-info/METADATA,sha256=P5knjysgVTcYCKlMkbJT-iiw_t9cy1SlskD2YuyknHE,26084
+npcpy-1.2.20.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+npcpy-1.2.20.dist-info/top_level.txt,sha256=g1pbSvrOOncB74Bg5-J0Olg4V0A5VzDw-Xz5YObq8BU,6
+npcpy-1.2.20.dist-info/RECORD,,