Export
🧠 Neode as Memory for OpenAI
This notebook demonstrates using Neode as a persistent knowledge graph that functions as memory for OpenAI's GPT models. The LLM can store facts and retrieve them later using Neode as a tool.
How It Works
- Store Memory: When the LLM learns new information, it calls Neode to extract and store triples
- Recall Memory: When asked questions, the LLM searches Neode for relevant knowledge
- Persistent: Knowledge persists across conversations and sessions
Use Cases
- Personal AI assistants that remember user preferences
- Research assistants that accumulate domain knowledge
- Chatbots with long-term memory
[ ]
Define Memory Tools
We'll give the LLM two tools:
store_memory: Save information to the knowledge graphrecall_memory: Search for relevant information
[2]
✅ Memory tools defined: • store_memory: Store information in long-term memory. Use this when you lea... • recall_memory: Search long-term memory for relevant information. Use this w...
Implement Memory Functions
These functions connect the LLM tools to Neode:
[8]
✅ Memory functions implemented!
Create the Chat Assistant
A chat function that handles tool calls automatically:
[9]
✅ Chat function ready!
Demo: Teaching the Assistant
Let's have a conversation where we teach the assistant some facts:
[5]
👤 User: Hi! My name is Alex and I'm a software engineer. I work at TechCorp and I love building AI applications.
🔧 Calling store_memory({'information': "User's name is Alex. They are a software engineer at TechCorp and love building AI applications."})
📋 Result: Stored 11 facts:
Alex → has_occupation → software engineer
Alex → has_occupation → software engineer...
🤖 Assistant: Nice to meet you, Alex — software engineer at TechCorp, and into building AI applications.
I’ve stored that info so I can remember it for later.
[6]
👤 User: I'm currently working on a project using Python and FastAPI. My favorite programming language is Rust though!
🔧 Calling store_memory({'information': 'Alex is currently working on a project using Python and FastAPI. Their favorite programming language is Rust.'})
📋 Result: Stored 10 facts:
Alex → is_working_on_project_using → Python
Alex → is_working_on_project_using → Fa...
🤖 Assistant: Got it — you’re currently working with Python + FastAPI, and your favorite language is Rust.
I’ve stored that for later.
Demo: Recalling Information
Now let's test if the assistant can recall what we told it:
[12]
👤 User: What do you know about me, Alex?
🔧 Calling recall_memory({'query': 'Alex user details preferences personal info'})
📋 Result: Found memories:
• Alex has occupation software engineer
• Alex has occupation software engineer
• Al...
🤖 Assistant: From what I have saved in memory about you:
- Your name is **Alex**.
- You’re a **software engineer**.
- You’re **employed by TechCorp**.
That’s all I currently have stored.
[13]
👤 User: What programming languages do I use?
🔧 Calling recall_memory({'query': 'programming languages Alex uses'})
📋 Result: Found memories:
• Alex favorite programming language Rust
• Alex favorite programming language Rust
...
🤖 Assistant: I have stored that your favorite programming language is **Rust**.
Interactive Chat
Run this cell to have an interactive conversation with memory:
[15]
🧠 Chat with Memory (type 'quit' to exit) ================================================== 👤 You: quit Goodbye!
View Stored Knowledge
Check what's been stored in your Neode knowledge graph:
[16]
📊 Knowledge Graph Contents: ============================================================ Alex → is_working_on_project_using → Python Alex → is_working_on_project_using → FastAPI Alex → favorite_programming_language → Rust Python → is_a → programming language FastAPI → is_a → web framework for building APIs with Python Rust (programming language) → is_a → programming language FastAPI → is_based_on → Starlette FastAPI → uses → Pydantic FastAPI → is_authored_by → Sebastián Ramírez Rust (programming language) → first_released_in → 2010 Alex → is_working_on_project_using → Python Alex → is_working_on_project_using → FastAPI Alex → favorite_programming_language → Rust Python → is_a → programming language FastAPI → is_a → web framework for building APIs with Python Rust (programming language) → is_a → programming language FastAPI → is_based_on → Starlette FastAPI → uses → Pydantic FastAPI → is_authored_by → Sebastián Ramírez Rust (programming language) → first_released_in → 2010 Alex → has_occupation → software engineer Alex → has_occupation → software engineer Alex → employed_by → TechCorp Alex → interested_in → building AI applications Techcorp LLC → provides → information technology solutions Techcorp LLC → provides → consulting services Techcorp LLC → supports → Artificial Intelligence Techcorp LLC → located_at → 3703 Green St, Claymont, Delaware 19703 Techcorp Information Systems Technology Company LLC → located_in → Riyadh 12621, Kingdom of Saudi Arabia TECHCORP → located_at → PO Box 17868, Fountain Hills, Arizona 85269 TechCorp → founded_in → 2015 Alex → has_occupation → software engineer Alex → has_occupation → software engineer Alex → employed_by → TechCorp Alex → interested_in → building AI applications Techcorp LLC → provides → information technology solutions Techcorp LLC → provides → consulting services Techcorp LLC → supports → Artificial Intelligence Techcorp LLC → located_at → 3703 Green St, Claymont, Delaware 19703 Techcorp Information Systems Technology Company LLC → located_in → Riyadh 12621, Kingdom of Saudi Arabia TECHCORP → located_at → PO Box 17868, Fountain Hills, Arizona 85269 TechCorp → founded_in → 2015 Alex → is_working_on_project_using → FastAPI FastAPI → is_a → web framework FastAPI → programming_language → Python FastAPI → builds → APIs FastAPI → uses → Python type hints Rust → is_a → programming language Rust → current_version → 1.90.0 Rust → has_package_manager → Cargo
[ ]