fix: minor memory issues in studio and added real weather api tool

[victorm-lc]

🐛 Fix: Memory Loading Issues in LangGraph Studio + Weather API Enhancement

Problem

When running agents with long-term memory in LangGraph Studio (langgraph dev), memory was not loading properly for users, and hot-reloading caused ModuleNotFoundError crashes.

Root Causes Identified

1. Namespace Type Mismatch

The load_memory() and create_memory() functions were using inconsistent types for the namespace:

• load_memory() used state["customer_id"] (integer)
• create_memory() used str(state["customer_id"]) (string)

This meant memory was stored at namespace ("memory_profile", "1") but loaded from ("memory_profile", 1) - two different namespaces!

2. Pydantic Model Serialization Issue

The create_memory() function was storing Pydantic model instances directly:

store.put(namespace, key, {"memory": updated_memory})  # ❌ Stores object

When LangGraph's disk-backed store pickles this, it saves class references (e.g., __agents__memory_enabled_music_store_supervisor_with_interrupt.UserProfile). During hot-reload, Python can't find these module references → crash.

3. Memory Format Handling

The format_user_memory() function only handled Pydantic models with attributes, but after retrieval from the store, data could be either a dict or a Pydantic model.

Changes Made

agents/memory_enabled_music_store_supervisor_with_interrupt.py

1. Fixed namespace consistency (line 142):

   user_id = str(state["customer_id"])  # Convert to string consistently

2. Fixed memory serialization (line 205):

   store.put(namespace, key, {"memory": updated_memory.model_dump()})  # Convert to dict

3. Fixed memory format handling (lines 128-132):

   # Handle both Pydantic model (attributes) and dict (keys) representations
   if isinstance(profile, dict):
       music_prefs = profile.get('music_preferences', [])
   else:
       music_prefs = getattr(profile, 'music_preferences', [])

agents/lg101-weather-agent.py

• Enhanced with real weather API integration (OpenWeatherMap or similar)

Testing

• ✅ Memory now stores and loads correctly across sessions
• ✅ Hot-reload works without ModuleNotFoundError
• ✅ Memory preferences persist properly for users
• ✅ Works in both notebook environment and LangGraph Studio

Impact

• Users' music preferences are now properly loaded and used for personalized recommendations
• Development workflow is smoother with working hot-reload
• No need to manually clear .langgraph_api folder after code changes

Related Issues

Fixes memory loading in LangGraph Studio for agents using long-term memory store.