massgen 0.1.4__py3-none-any.whl → 0.1.6__py3-none-any.whl

massgen/memory/docs/QUICKSTART.md

@@ -0,0 +1,409 @@
+# MassGen Memory - Quick Start Guide
+
+Get up and running with MassGen memory in 5 minutes!
+
+## Installation
+
+```bash
+# Install mem0 for persistent memory support
+pip install mem0ai
+```
+
+## Basic Usage
+
+### 1. Short-term Conversation Memory
+
+Perfect for maintaining context during active chats:
+
+```python
+import asyncio
+from massgen.memory import ConversationMemory
+
+async def chat_example():
+    # Create memory
+    memory = ConversationMemory()
+
+    # Add messages
+    await memory.add({"role": "user", "content": "What's 2+2?"})
+    await memory.add({"role": "assistant", "content": "2+2 equals 4."})
+
+    # Get all messages
+    messages = await memory.get_messages()
+    print(f"Stored {len(messages)} messages")
+
+    # Get only recent messages
+    recent = await memory.get_messages(limit=10)
+
+asyncio.run(chat_example())
+```
+
+**When to use**: Current conversation, temporary storage, quick message access
+
+### 2. Long-term Persistent Memory
+
+For cross-session knowledge retention:
+
+```python
+import asyncio
+from massgen.memory import PersistentMemory
+
+async def persistent_example():
+    # NOTE: Replace with your actual backends
+    from massgen.backend import YourLLMBackend, YourEmbeddingBackend
+
+    llm = YourLLMBackend(model="gpt-4")
+    embedding = YourEmbeddingBackend(model="text-embedding-3-small")
+
+    # Create persistent memory
+    memory = PersistentMemory(
+        agent_name="my_assistant",
+        llm_backend=llm,
+        embedding_backend=embedding,
+        on_disk=True  # Persist across restarts
+    )
+
+    # Record conversation
+    await memory.record([
+        {"role": "user", "content": "I love hiking in mountains"},
+        {"role": "assistant", "content": "That sounds wonderful!"}
+    ])
+
+    # Later, retrieve relevant info
+    context = await memory.retrieve("user's hobbies")
+    print(f"Retrieved: {context}")  # "user loves hiking"
+
+asyncio.run(persistent_example())
+```
+
+**When to use**: Long-term facts, cross-session knowledge, semantic search
+
+## Common Patterns
+
+### Pattern 1: Simple Chat Agent
+
+```python
+class SimpleChatAgent:
+    def __init__(self, backend):
+        self.backend = backend
+        self.memory = ConversationMemory()
+
+    async def chat(self, user_message):
+        # Add user message
+        await self.memory.add({"role": "user", "content": user_message})
+
+        # Get context
+        context = await self.memory.get_messages()
+
+        # Generate response
+        response = await self.backend.chat(context)
+
+        # Save response
+        await self.memory.add({"role": "assistant", "content": response})
+
+        return response
+```
+
+### Pattern 2: Agent with Memory
+
+```python
+class SmartAgent:
+    def __init__(self, llm_backend, embedding_backend):
+        self.backend = llm_backend
+        self.short_term = ConversationMemory()
+        self.long_term = PersistentMemory(
+            agent_name="smart_agent",
+            llm_backend=llm_backend,
+            embedding_backend=embedding_backend
+        )
+
+    async def chat(self, user_message):
+        # 1. Add to short-term
+        await self.short_term.add({"role": "user", "content": user_message})
+
+        # 2. Get relevant long-term memories
+        context = await self.long_term.retrieve(user_message)
+
+        # 3. Build full context
+        messages = await self.short_term.get_messages()
+        if context:
+            messages.insert(0, {
+                "role": "system",
+                "content": f"Relevant background: {context}"
+            })
+
+        # 4. Generate response
+        response = await self.backend.chat(messages)
+
+        # 5. Save to short-term
+        await self.short_term.add({"role": "assistant", "content": response})
+
+        # 6. Optionally save to long-term
+        # (only if important)
+        if self._is_important(response):
+            await self.long_term.record(messages[-2:])
+
+        return response
+
+    def _is_important(self, message):
+        # Implement your logic
+        keywords = ["remember", "important", "fact", "name", "birthday"]
+        return any(kw in message.lower() for kw in keywords)
+```
+
+### Pattern 3: Agent-Controlled Memory (Tools)
+
+```python
+class AutonomousAgent:
+    def __init__(self, backend, embedding_backend):
+        self.backend = backend
+        self.memory = PersistentMemory(
+            agent_name="autonomous",
+            llm_backend=backend,
+            embedding_backend=embedding_backend
+        )
+
+        # Register memory as tools
+        self.tools = [
+            {
+                "name": "save_to_memory",
+                "description": "Save important facts for later",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "thinking": {"type": "string"},
+                        "content": {"type": "array", "items": {"type": "string"}}
+                    }
+                },
+                "function": self.memory.save_to_memory
+            },
+            {
+                "name": "recall_from_memory",
+                "description": "Search your memory for information",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "keywords": {"type": "array", "items": {"type": "string"}}
+                    }
+                },
+                "function": self.memory.recall_from_memory
+            }
+        ]
+
+    async def chat(self, user_message):
+        # Agent can now decide when to save/recall memories
+        response = await self.backend.chat(
+            messages=[{"role": "user", "content": user_message}],
+            tools=self.tools
+        )
+
+        # If agent called save_to_memory or recall_from_memory,
+        # those functions will be executed automatically
+
+        return response
+```
+
+## Configuration Options
+
+### Conversation Memory
+
+```python
+memory = ConversationMemory()
+
+# Truncate to last N messages
+await memory.truncate_to_size(100)
+
+# Get messages by role
+user_msgs = await memory.get_messages_by_role("user")
+
+# Save/restore state
+state = memory.state_dict()
+new_memory = ConversationMemory()
+new_memory.load_state_dict(state)
+```
+
+### Persistent Memory
+
+```python
+# Basic configuration
+memory = PersistentMemory(
+    agent_name="agent_1",
+    llm_backend=llm,
+    embedding_backend=embedding
+)
+
+# With user tracking
+memory = PersistentMemory(
+    agent_name="agent_1",
+    user_name="alice",
+    llm_backend=llm,
+    embedding_backend=embedding
+)
+
+# With session tracking
+memory = PersistentMemory(
+    agent_name="agent_1",
+    session_name="2024-01-15",
+    llm_backend=llm,
+    embedding_backend=embedding
+)
+
+# Custom vector store
+from mem0.vector_stores.configs import VectorStoreConfig
+
+vector_config = VectorStoreConfig(
+    provider="qdrant",
+    config={
+        "on_disk": True,
+        "path": "./my_memory_db"
+    }
+)
+
+memory = PersistentMemory(
+    agent_name="agent_1",
+    llm_backend=llm,
+    embedding_backend=embedding,
+    vector_store_config=vector_config
+)
+```
+
+## Tips & Best Practices
+
+### 1. Memory Management
+```python
+# Keep conversation memory manageable
+if await memory.size() > 100:
+    await memory.truncate_to_size(50)  # Keep last 50
+
+# Or clear when starting new topic
+await memory.clear()
+```
+
+### 2. Selective Persistence
+```python
+# Don't save everything to long-term memory
+# Only save important information
+
+if contains_important_info(message):
+    await long_term.record([message])
+```
+
+### 3. Efficient Retrieval
+```python
+# Use specific queries for better results
+context = await memory.retrieve("user's name and preferences")
+
+# Instead of vague queries
+# context = await memory.retrieve("information")  # Too broad
+```
+
+### 4. State Persistence
+```python
+import json
+
+# Save state to file
+state = memory.state_dict()
+with open("memory_state.json", "w") as f:
+    json.dump(state, f)
+
+# Restore state from file
+with open("memory_state.json", "r") as f:
+    state = json.load(f)
+new_memory = ConversationMemory()
+new_memory.load_state_dict(state)
+```
+
+## Troubleshooting
+
+### Problem: ImportError for mem0
+```bash
+# Solution: Install mem0
+pip install mem0ai
+```
+
+### Problem: Memory not persisting
+```python
+# Make sure on_disk=True
+memory = PersistentMemory(
+    agent_name="agent_1",
+    llm_backend=llm,
+    embedding_backend=embedding,
+    on_disk=True  # ← Important!
+)
+```
+
+### Problem: Out of memory
+```python
+# Regularly truncate conversation memory
+await conv_memory.truncate_to_size(100)
+
+# Or clear when no longer needed
+await conv_memory.clear()
+```
+
+### Problem: Slow retrieval
+```python
+# Reduce limit in retrieve calls
+context = await memory.retrieve(query, limit=3)  # Instead of default 5
+
+# Use more specific queries
+# Good: "user's favorite color"
+# Bad: "information about user"
+```
+
+## Next Steps
+
+1. **Read the full documentation**: Check `README.md` for detailed API reference
+2. **Explore examples**: Run `examples.py` to see more usage patterns
+3. **Understand design**: Read `DESIGN.md` for architecture details
+4. **Try it out**: Integrate memory into your agents!
+
+## Complete Example
+
+Here's a fully working example you can copy and run:
+
+```python
+import asyncio
+from massgen.memory import ConversationMemory
+
+async def complete_example():
+    # Create memory
+    memory = ConversationMemory()
+
+    # Simulate a conversation
+    print("Starting conversation...")
+
+    await memory.add({"role": "user", "content": "Hi, I'm Alice"})
+    await memory.add({"role": "assistant", "content": "Nice to meet you, Alice!"})
+    await memory.add({"role": "user", "content": "What's my name?"})
+    await memory.add({"role": "assistant", "content": "Your name is Alice."})
+
+    # Show stored messages
+    messages = await memory.get_messages()
+    print(f"\n📚 Stored {len(messages)} messages:")
+    for msg in messages:
+        print(f"  {msg['role']}: {msg['content']}")
+
+    # Show memory size
+    size = await memory.size()
+    print(f"\n📊 Memory size: {size} messages")
+
+    # Get last message
+    last = await memory.get_last_message()
+    print(f"\n💬 Last message: {last['role']}: {last['content']}")
+
+    # Save state
+    state = memory.state_dict()
+    print(f"\n💾 State saved: {len(state['messages'])} messages")
+
+    print("\n✅ Example completed!")
+
+if __name__ == "__main__":
+    asyncio.run(complete_example())
+```
+
+Run this with:
+```bash
+python -m massgen.memory.examples  # Or save the code above
+```
+
+Happy coding! 🚀